Hybrid Sources Powered Electric Vehicle Configuration and Integrated Optimal Power Management Strategy

Internal Combustion Engine based transportation rapidly increases the impact on the environment. The burning of fossil fuels in the industries and transportation sector breadth of global warming. However, Hybrid Electric Vehicle is not a permanent solution for emission-less road vehicles. Therefore, electric vehicles are the most feasible technologies to attain the goals of energy savings and zero-emission road vehicles. This paper sequentially surveys the key point of vehicles like; Powertrain strategy, configuration, fuel economy, reduced emission, and control strategy expounds in terms of its basic principles, pros, and cons. These key points make the energy management strategy more conclusive for the more efficient vehicle. In addition, this paper review systematically qualitative and quantitative algorithm in all type of EMS used in HEV and compare them with existing approaches in terms of pros & cons through a comprehensive analysis. Furthermore, addressed the potential research gap and provide the directives for further development in power train and ems in every respect.


I. INTRODUCTION
This Concern over air contamination, the hydrocarbon-based conveyance has been raising worldwide concerns. International Energy Agency (IEA) estimated 298 Mtoe biofuel consumption in the transportation sector by 2030. This sector consumes 49% of the oil resources and it seems that world resources depleted by 2038 [1], and the U.S. EPA (Environmental Protection Agency) reported that 29% of greenhouse gas emissions (CO 2 , NO 2 , CO, NO) from the burning of fossil fuel in transportation activities.
The associate editor coordinating the review of this manuscript and approving it for publication was Huaqing Li . Around 28% of all carbon dioxide (CO2) emissions are attributed to the transportation industry, with road transport responsible for more than 70% of those emissions, according to studies by the European Union and other sources. In which main contribution of light commercial vehicles 59%, intermediate and heavy commercial trucks 23%, aircraft 9%, ships & boats 3 %, rail 2% other 4%.
Therefore, the governments of the majority of developed nations are encouraging the use of electric cars in order to minimize the concentration of air pollutants, including CO2 and other greenhouse gases [2]. These days, public consciousness of weather variation and importation of the power savings is increasing the development of innovative technologies for the green vehicle to the utilization of Plug-in Hybrid Electric Vehicle (HEV) and full electric vehicles (FEV), which is a possible eco-friendly and economical solution [3]. A wide range of automakers are working hard to create new EVs [4]: • General Motors intends to switch to all-electric vehicles by 2023 • Ford will offer seven electrified and customized plug-in hybrids in the future • Mazda, Denso, and Toyota are working together to develop technologies for EVs • Renault, Nissan, and Mitsubishi are working to create pure electric vehicles with the goal of releasing 12 EVs by 2023 • 300 vehicles from the Volkswagen group, which also owns the brands Audi and Porsche, will be available in electric and hybrid versions by 2030. At present HEV has thrived as a solution [5] due to the discontinuity of renewable energy and battery life span. There are some reluctances for public interest in Electric Vehicles (EVs) as: the cost of EVs is higher than fossil-fueled vehicles, EVs do not have much range, refueling of the fossil-fueled vehicle is easy in comparison to EVs, and reluctant to spend in current EVs technology while much-advanced technology may be available in next 2-3 Years.
Existing electric vehicles having more than a single source of power, hybrid powertrains provide a large design space for the system and increase the complexity of the control algorithm [6]. The objective function of the Energy Management Strategy (EMS) optimization problem is generally coupled with powertrain topology collection, while technology and the size of components are treated as optimization constraints. Power management strategy will play a vital role in the expansion of the new generation of green vehicles. The highest challenges of a power management strategy are to power split in optimum mode to provide intended performance under system limitations.
A significant amount of research into energy management strategy has been directed over the last era, not only for HEV [7], [5] but also for FEV. However, with the improvement and introduction of novel approaches in automotive technology, the author perceives EMS for HEV and FEV as a constantly developing area that will continue to draw fresh ideas for many upcoming years.
The major goals of this review work are to contribute to an emergent point of discussion about the latest EMS approaches, as well as to provide an inclusive outline of EMS generated for controlling power management in HEV or FEV.
This research paper is structured as follows: a collection of data sources in section 2, the architecture of the vehicle and a range of powertrain systems & electric alignment of HESS in FEV in section 3, and classification of optimization techniques in power management represent in section 4, latest developing trends & prospective opportunities for future research trends in energy management strategies discussed in Section 5 and some remarkable conclusion in section 6.

II. SOURCES OF DATA COLLECTIONS
All documents employed in this review paper were from the database of reputed journals and conference papers in which we found energy management strategies have been used in the past and current. There was a total of article 267 in this review paper have been used. The contribution of articles is given in Fig.1.

III. ARCHITECTURE OF HYBRID ELECTRIC VEHICLE
The primary powertrain-controlled structures fo HEV and FEV HEV and FEV are discussed in this segment, along with their main properties. The functional mode of a powertrain topology is necessary to understand before expressing an EMS optimization issue. There is numerous topology on the powertrain of various competencies that can be implemented via modification of the power source connections. These connections may be mechanical or electrical links. HEV powertrain has three major configurations, (1) Series Connection; (2) Parallel Connection, and (3) Series-Parallel, although FEV is further divided into three parts according to the source of energy, battery, and solar-based or fuel cellbased [3]. The Vehicle power train strategy of HEV and FEV is given below in Fig.2.

A. HYBRID ELECTRIC VEHICLES
In the HEV, engine power is transferred across the ring gear which is mechanically joined through the drive shaft while another part of engine power is converted into electrical power to drive the motor, which is hinged with gear, connected into mechanical power again. The prior arrangement is called as a parallel path and the second is called as a series path [8]. The primary aim for HEVs development is to decrease fuel intake and tailpipe emission [5]. According to vehicle powertrain arrangement, HEV can be separated into series, parallel, power-split/ series-parallel HEV, and Plug-in HEV, which can make energy more efficient and relatively high fuel economy [9], [10]. The vehicle configuration and VOLUME 10, 2022 architecture are shown in Fig.2 (a & b)., while the summary of HEV& FEV, and their application are shown in Table 1.

1) SERIES HYBRID ELECTRIC VEHICLES (S-HEV)
A series-HEV topology gives the best performance in a stop-and-go driving pattern. There, ICE and wheels don't have a mechanical connection. The ICE is basically used to generate the electrical power by driving the generator, which is combined with an output power of electrical storage and transmits that power by DC bus to an electric motor to drive the wheels [3]. In this strategy, the engine runs efficiently in varying vehicle speeds [6]. S-HEV are suitable for urban and buses for highway only, while buses are not suitable for urban area driving due to high conversion loss [7].

2) PARALLEL HYBRID ELECTRIC VEHICLES (P-HEV)
In Parallel HEVs, an electric motor is used alone at low speeds while the engine and wheels are mechanically connected directly [11], such that their combined torque is transferred to the wheels via a standard moving shaft and probably a different gear. In this method energy loss is minimum but they are less suitable for quick change stop-and-go traffic in compared to the series HEV topology [6], [12], [3].

3) SERIES-PARALLEL HYBRID ELECTRIC VEHICLES (SP-HEVS)
A Series-Parallel HEVs, as well power-split HEV, have an additional mechanical connection in between the motor and generator through the transmission. This arrangement provides the complementary benefits of series and parallel HEVs [3], [6]. As a result, one of the main problems of SP-HEV is power flow regulation of splitting power because it combines functioning components from both series and parallel systems, increasing the system's complexity [13].

4) PLUG-IN HYBRID ELECTRIC VEHICLES (P-HEV)
Plug-in HEVs differ from conventional HEVs but have the same configuration with additional electric charging plug, and higher capacity electrical components shown in Fig.3. In this way, the Energy Storage System (ESS) is considered as the prime source, which provided a new dimension to the EMS method in PHEV for a superior fuel economy by operating in two modes as charge depleting (CD) and charge sustaining(CS) modes [5]. PHEV can be run in full-electric mode for a long time period due to high capacity electrical components [1]. The plug-in HEV is a good initiative towards reducing worldwide emissions, by which it proposed high performance and fuel efficiency in both electric and hybrid mode [3], [12]. In 2011, Nissan introduced, the company's first plug-in electric vehicle 'Nissan Leaf', while A Ford Fusion Energi is a plug-in hybrid car that debuted in 2013.

B. FULL ELECTRIC VEHICLE
Presently, FEV has seven types of power transfer topology as shown in Fig.4, in which only three types of topologies are prominent for use via an auto-industrialist [14]. The comparative configuration of various HESS is shown in Table 2.
In general, fully electric vehicles are divided into two categories based on the energy source, it may be a fuel cell or battery. Now a day, PV-based vehicle has captured a considerable amount of interest from researcher to enhance VOLUME 10, 2022  the utilization of renewable resources which are available in abundant amount in nature. The PV-based vehicle was also used in my research project.

1) BATTERY-BASED FEVS
In this FEV, the battery is used as a primary source that has high-energy content. It is combined with another high-density power device like, Supercapacitor, known as ultra-capacitor also to form the HESS. It is also known as an electric double-layer capacitor (EDLC) [15]. In comparison--to a supercapacitor, batteries have a high energy density but a poor power density. So, HESS store enough energy to satisfy abrupt power demands to achieve the required vehicle performance. H. He et al., presented the seven battery model with enough precision and less complexity and provide the optimal performance on experimental results [16]. An FEV can be divided into two categories.

2) FUEL CELL-BASED FEVS
In this FEV, FC worked as a primary source, which uses H 2 and O 2 to produce electricity. The FC's specific energy and power are similar, but not identical to the gasoline [3]. Fuel cells have a slow response due to chemical reactions, so it's not good to provide the frequent changing load. To mitigate this problem, it has hybridized with battery /UC. The FC-based FEVs configuration are shown in Fig.5

3) PV BASED FEV
The architecture of a solar power-based FEVs (PV-FEV) is similar to the Plug-in HEV except for an additional photovoltaic (PV) panel, which provides the current for battery charging in a day time. And the maximum power point tracking (MPPT) control algorithms are applied to achieve the maximum power through PV Panels. The PV-based FEV configuration is shown in Fig.6.

IV. ENERGY MANAGEMENT STRATEGIES FOR VEHICLE
FEV and (P)HEVs are complex electro-mechanical drive systems. The choice of the circuit configuration and EMS have decided the flow of power, fuel economy, and emission reduction [17]. The main purpose of an EMS is to control the power flow for obtaining the improved fuel economy, emissions reduction, ensured drivability, and maintain the state of charge (SOC) and life span of the ESS via advertent the restrictions. A general outline of the objectives of EMS for both FEV and HEVs is shown in Fig.7.
In the past, lots of diversity of research are available for the usage of EMS in Hybrid-EV, Fully-EV, and Plug-in HEV applications. Even though lots of classification can be found in the literature, mainly divided into three parts: rule-based, optimization-based, and learning-based algorithms, and these are further divided into subparts. The taxonomy of the primary parts and subparts are shown in Fig.8 for HEV and FEV technologies. In this article optimization algorithms which are used in Intelligent Transportation System (ITS), additionally included besides of EMS categorization shown in fig.8.

A. RULE-BASED CONTROL METHOD
Rule-Based (RB) control techniques are heuristic control techniques in which the control method well-defined as a set of ''if-then'' procedures to regulate the control action [18]. The rules-based method is determined by using humanoid intelligence, intuitions, or mathematical models and mostly without pre-information of a drive cycle. They are required low computation so that they are commonly used in many commercial vehicles like Honda Insight and the Toyota Prius [7]. Peng et al. [19] recalibrated the rule-based EMS results to locate the optimum power train by applying dynamic programming (DP) and reduced the fuel consumption by about 10.45 % as well as the electricity consumption up to 4.75%. Ceraolo et al. [20] discussed the optimal energy consumption technique to resolve the energy problem occurring in the designing process. Ali et al. [21] presented an optimized situation-based power management strategy for multi-source EVs. Here proposed methodology obtained the optimal results and improve the energy efficiency 11.9-18.98% while prediction accuracy about 63.9-65.2% respectively dynamic programming and rulebased algorithm. Even though a rule-based EMS may not provide the best result; it has attracted interest due to its ease of deployment in real-time. This strategy is more classified into fuzzy and deterministic rule-based EMSs.

1) DETERMINISTIC RULE-BASED METHODS
The Deterministic Rules-Based (RB) methods are developed with the support of fuel economy or emission data, power split, operating point of ICE, and power flow in the drive train. Rules execution is performed on the basis of a lookup table to share the power in between IC Engine and motor. Hofman et al. [18] discussed an RB-ECMS, in which the control strategies are defined based on ''if-then'' rules for control action, compared these results with dynamic programming, and increase the accuracy of 1%.

a: THERMOSTAT CONTROL STRATEGY
This control method uses the ICE and generator to produce electrical energy via vehicle. ICE works at its maximum efficiency point when it starts, although the SOC of the battery is always maintained in between its pre-defined upper and lower levels by simply turning on or off ICE.
Ali et al. [22] implemented an RB energy management optimization technique to activate the ICE at optimal fuel saving mode during different standard drive cycles, to improve the vehicle efficiency then it reduces the total trip estimated cost by around 46%. Jalil et al. [23] applied an RB scheme to control the split power among the battery and ICE for a series HEV, in which the 'Thermostat' strategy improved the fuel economy up to 11% in the urban driving cycle and 6% in the highway driving cycle. The parameter designing optimization is proposed [24] to improve the vehicle economy and transmission ratio of the vehicle. Badjate et al. [25] developed a split power-based efficient control strategy in which fuzzy logic (FL) was applied to the interpretation of driver command and driving situations. Gao et al. [26] proposed an equivalent fuel consumption optimal control of a series (EFCOCS) type of power split management strategy [27], the combination of traction control system (TCS) and power factor correction (PFC), in which TCS offer the highest efficiency in the engine generator set while PFC improves the battery durability by controlling SOC. Kim et al. [28] proposed based on equivalent specific fuel consumption (ESFC) with the continuously varying transmission. They determine the maximum system efficiency with optimal values of parameters. These methods are generally used in Series HEV.

b: POWER FOLLOWER (Baseline) CONTROL STRATEGY
This control strategy is reformulated on/off control algorithm to deliver the added power and sustain battery SOC. In this strategy, ICE is taken as the main Source. This strategy is appropriate for both parallel and series-parallel HEVs. Luo et al. [29] uses a combination of two control strategies power follower control and DC-link voltage control to minimize the fuel economy of Series HEV. This method gives better performance than individual control strategies.

c: MODIFIED POWER FOLLOWER-ADAPTIVE RB (ARB)
In order to improve the thermostat and power follower technique, proposed an adaptive rule-based strategy. In this method, a decision is performed stepwise. Johnson et al. [30] proposed the adaptive-based real-time control method to optimize the efficiency and discharges of a parallel HEV. This strategy reduced the 23% NOx and 13% particulate matter (PM) discharges at an expense of 1.4% in fuel economy. Wipke et al. [31] proposed a modified baseline controller to focus on its combination of forwarding and backward-facing methodologies, and evaluates the model in terms of its design objectives.

d: FREQUENCY-BASED APPROACH
Frequency-based strategy approaches for split power requirement at low and high-level frequency components to fulfill the load demand. Kim et al. [32] proposed a frequencydomain power distribution (FDPD) method to improve the fuel budget via 5.9% and shrinkage the shoot emission via 62.7% for the engine and also reduced 23 % ineffective Ah for improving the life span of the battery. Tani et al. [33] focused on power management according to the dynamics performance of the hybrid sources using polynomial correctors to mitigate the transients of dynamic load.

e: OPTIMAL POINTS TRACKING
This method corresponds to the baseline control methods in which the functioning point of the IC engine can be adjusted easily. So that engine optimal functioning point, operation line, efficiency portion, and system optimal operation point are projected for series-parallel HEVs. Park et al. [8] applied the direct search method to optimize the losses in power flow and select the optimal power flow [34] point according to the efficiency and emission. This strategy has the advantage to control the battery power since the maps take charging or discharging power, which is used as one input data. Ahn et al. [35] formulated power-split architectures and applied two control strategies, power split configuration and mode switching configuration [36], for improving the efficiency of series-parallel HEV. They compare the results and verify them by simulation [37]. Andriollo et al. [38] present the optimum design of an electrodynamic suspension magnetic levitation (EDS-MAGLEV) transport system by using global objective function at an analytical aspect of system performance. The new definition is based on possible object oriented language (OOL) transition in HEV during a simulation. The power-split solution gives qualitative fuel depletion improvement in comparison to the conventional line tracking scheme.

2) FUZZY RULE-BASED METHOD
The fuzzy Rule-based control method is presented to energy management in HEV. This strategy is more advantageous due to its toughness to inaccurate dimension and component modeling inconsistency besides its adoption. This strategy is more applicable to multi-domain, non-linear time-changing systems for example HEVs. Fuzzy logic has decision-making properties, it is accepted to the analysis of a real-time & suboptimal split power [39]. Basically, this is the addition of deterministic RB-EMS. For example, Won et al. [40], implemented fuzzy rule bases control strategies for torque distribution and charge sustained in traffic situations for intelligent energy management. Hannoun et al. [41] designed a fuzzy controller [42], [43] according to the energy demand by vehicle speed and SOC of the battery optimize the energy consumption and condensed the pollutant emission. Mohan et al. [44] worked on a fuzzy proportional differential controller with respectably two input velocity & acceleration and single output for incremental control effort. This FL strategy is further divided as follows:

3) CONVENTIONAL FUZZY METHOD
In this method, the FL controller is used to perform basic steps of fuzzy logic. These are tuned via an optimization algorithm to full fill the control purposes for power management like reduced fuel consumption, emission reduction, and maintaining the state of charge of ESS to enhance the driving performance. For example, Lee et al. [45] Implemented FL control through the driving cycle strategy to accelerate the pedal stroke and reduced the hydrogen up to 22% in FCHEV [46]. While, Farrall et al. [47] control the powertrain through legislation in the heat engine used in the hybrid vehicle. Montazeri-Gh et al. [48] improve the fuel economy and reduced its consumption upto 21% by using multi-input FL control strategies. Pan et al. [49] and [50] applied a wavelet-based FLC for multi-input EMS in a hybrid system for a tracked bulldozer and verified the result in real-time. Yuan et al. [51] proposed the power management through stochasticity and fuzziness in solar power generation as well as ship power load demand in his research and improve the solar performance by reducing fuel consumption and CO 2 emission. Chen et al. [52] proposed an FLC in EMS for a certain driving cycle and applied the multi-objective optimization evolutionary algorithm to optimize the parameter [53] of fuzzy function trims and semi-trapmf to improve the FLC efficiency. Wu et al. [54] recognize driving cycle patterns to improve the fuel economy through FL-based EMS. This pattern is recognized by the learning vector quantization method. Schouten et al. [42] and Baumann et al. [55] proposed a novel control strategy for FLC-based power management [56], [57] in HEV and optimize all its components related to the efficiency of the vehicle and also show that the strategy justify the highly nonlinear multi-domain and time-varying plant.

4) ADAPTIVE FUZZY METHOD
The performance of the conventional fuzzy method can be further improved if the control parameters are adaptive for the present operating point. Regarding the HEV, Tong, et al. [58] applied adaptive fuzzy decentralized control technique to estimate the unmeasurable non-linear function, which is designed by back stepping technique and, the Lyapunov function and average dwell time method are used for stability. Chen et al. [59] considered the problem observer-based adaptive fuzzy control for non-linear time-delay system and the signals in closed-loop system are uniformly bounded. Tian et al. [60] proposed adaptive FL used to adhere to the deviation of SOC online and speed of the vehicle to find out the degree of the engine's output powers. Bathaee et al. [61] proposed an FL-based torque controller [62] and optimize the energy flow, generation, and conversion in the individual component [56] of the parallel hybrid vehicle. Li et al. [63] proposed strategy combined the logic threshold and fuzzy control in which fuzzy control design based on Improve Quantum Genetic Algorithm (IQGA) optimization technique to improve the fuel efficiency. Fuel consumption is decreased by 5.17 % by applying IQGA which gives a better response than GA and QGA. Wang et al. [64] implemented a novel real-time evolutionary FL-based EMS then applied the GA for fine-tuning and optimization same. Wu et al. [65] described a method to Optimize control strategy and fuel economies of HEV using multi-objective self-adaptive differential evolution. Wang et al. [66] used a fuzzy control method and DP method for reducing the time period of cold start and energy consumption of warm-up process respectively. Li and Liu [67] show overall efficiency of an FC/battery hybrid vehicle is maximum for given driving cycles and results show that optimally controlled HEV which can provide better fuel economy and enhance system efficiency.

5) PREDICTIVE FUZZY METHOD
Predictive Fuzzy Logic controller works on prior knowledge of driving a trip on a planned route to perform in realtime but mainly drawback of its incapability to accomplish real-time control task. Hajimiri et al. [68] used predictive and protective algorithm (PPA) with FLC to extend the battery life. Montazeri-Gh et al. [54] described a predictive optimized intelligent fuzzy control strategy based on traffic condition recognition for fuel consumption and emission. Niu et al. [69] presented a machine learning framework for real-time driving cycle and trends, which is named as neural network standard driving cycle (NN-SDC) and neural network driving trends (NN-DT). It is developed an intelligent FLC strategy based on micro-controller frame in HEV for determining the power consumption and emission to improve the efficiency. To extend the battery range of BEV here, Mohd et al. [70] applied integrated multimode driving using FL enabled the adapting driving, which select the parameter automatically through speed and reduced the energy consumption by about 32.25%, and increase the driving range up to 4.21%. Yin et al. [71], analyzed a power split mechanism and transmission efficiency of the EV based on control strategies. Where torque distribution is realized by FLC which is optimized by PSO. Ippolito et al. [72] design a fuzzy clustering criterion-based controller with GA to reduce the computational effort and improve efficiency in energy management. Kamal et al. [73] investigated a robust FLC tuned with NN for energy management with battery fault detection and power distribution management. Tao and Taur [74] designed a flexible difficulty-reduced PIDlike fuzzy controller which reduced complexity by reducing the number of input variables. Chen et al. [75] developed a machine learning (ML) algorithm ''LOPPS'' to study optimal power combination in an EVs load variation then applied FL according to LOPPS results to reduce power losses.

B. OPTIMIZATION BASED POWER MANAGEMENT CONTROL METHOD
The main goal of the optimization-based power management method is to find the optimal control consequences to minimize the process cost above the particular time period. The optimization-based methods can be divided into two categories: offline mode method and online mode method. Earlier bibliometric expose that Optimization Based (OB) methods clasp additional courtesy in the research field with a 56.7% in compared to Rule-Based (RB) methods 32.9% [7]. More details about each category are given below.

1) OFFLINE MODE METHOD
An offline Optimization-based method is belonging to the non-causal and worldwide optimization-based method because it needs a pre-information of upcoming driving cycles. The significance of this type of strategy is that the optimal solution is providing a standard solution for another causal method compared to the other modified online strategies. The offline strategies based on the problem-solving approach can be distributed into four parts: direct algorithm, indirect algorithm, gradient, and derivative-free Algorithms.

a: DIRECT ALGORITHM
The direct algorithm is used to solve the static optimization problems by discretization. The commonly used algorithm to solve the problem of energy management in the offline strategy is dynamic programming (DP), which is originated via Bellman in the 1950s. It is also called deterministic dynamic programming (DDP) because it required prior knowledge of the driving cycle.

b: DYNAMIC PROGRAMMING
Dynamic programming is proposed to resolve optimal control problems in a non-linear system. It decays dynamic optimization problems in an order of sub-problems by discretizing original optimization time. DP applied by Pei et al. [76] to find the equivalent marginal cost factor in ECMS. Zahraeia et al. [77] considered temperature noise factor for optimal EMS and improve fuel efficiency [78] and emission. Sinoquet et al. [79] presented a real-time control strategy in HEV to minimize fuel consumption and reduction in pollutant emission. Marano et al. [80] and Tulpule et al. [81] reduced the 1% fuel consumption in PHEV through DP in comparison to ECMS. Skugor et al. [82] use fleet charging method and show the DP optimization more successfully by reducing the charging cost by more than 10 %. Pan et al. [83] applied two methods RB projection partition method for system efficiency and DP based optimization method to explore the energy-saving factor in HEV. In which RB strategy reduced 13.4 % while DP reduced 17.6 % energy consumption. Zhang et al. [84] applied dynamic programming in distribution to utilize all power users at variable load requirement. The author also shows a 20 % improvement in fuel economy by DP [85]. Chen et al. [86] applied DP for design standards and real-time implementation strategy in power management to evaluate fuel-energy-loss-oriented (FE-LO) and battery-energy-loss-oriented (BE-LO) in which battery protection and fuel economy are used as a cost function. Lin et al. [87] proposed a novel battery model and dynamic programming-based energy management (EM) algorithm for reconfigurable battery packs and optimal power distribution to increase the battery lifetime in BEV [88]. Kessels et al. [89] reduce the mathematical complexity and prior information of driving cycle in DP and produce the optimal solution without using prior road information, and improve the fuel economy up to 25 %. Wu et al. [90] proposed the driving cyclebased DP optimization technique for energy flow in rangeextended electric buses (REEB). It reduced the computation time by 96.85 % but power consumption is 0.47% greater than the traditional DP. Asus et al. [91] applied to optimize the control parameter and found an operating point of the engine to achieve longer durability. Sundstrom et al. [92] discussed the hybridization ratio in torque assist and fuel consumption is obtained by using DP for different hybridization ratios. VOLUME 10, 2022 c: DETERMINISTIC DYNAMIC PROGRAMMING Deterministic dynamic programming algorithm is based on the concept of subdividing a non-linear dynamic optimization problem into a discrete temporal sub problem, where a cost go function is created at every step time. And the sub-problems solve via a backward recursive technique or a forward DP method to find the best control policy. Vinot et al. [93] developed a global optimum design method for parameters and component sizing in EV by discrete dynamic programming. Extracting DP for optimal energy management [94] design strategy to reduce the fuel consumption [95] and optimal speed and power split strategy [96], [97] in HEV. Wang et al. [85] proposed an optimal control method in PHEV and develop a mechanism based on discretization resolution variables and boundary issues and found 20% development in fuel economy in comparison to the traditional control strategy. Lin et al. [98] applied to extract DP rules to design the power management control strategy in HEV for optimal fuel consumption, reduction in emission and battery constraint for SOC and provide the 45% higher fuel economy than ICE truck.
The main issues of DDP are the high computation required due to the quantization of conditions and control variables, the essential expletive of dimensionality, and the reliance on the driving cycle. These disadvantages make DDP incapable of real-time application.

d: STOCHASTIC DYNAMIC PROGRAMMING (SDP)
The DDP-derived control law can only function as a detailed driving cycle, and it may not assurance a degree of optimality or a constant charge in different driving cycles. Additionally, they advised that DDP is not implemented directly and the rule extracting is a time taking process. To reduce these problems C.C. Lin et al. [99] applied SDD to model for Markov chain process and optimal control firstly and Zeng et al. [100] solved the formulated problem as a finite-horizon Markov decision. Romans et al. [101] optimize the size of the storage system & reduced the power distribution losses up to 24%. Wegmann et al. [102] found 2.4 -3.4% less battery energy losses calculated in SDP than EECMS. Moura et al. [103] applying the SDP for optimal power management in PHEV to sustain the battery charging and configure the charging station equipments as queuing theory based technique [104] to improve the engine efficiency and reduce charging time. Lust [105] approach iterative dynamic programming for a relatively coarse grid for optimum and vector controls to find the optimal policies for the next iteration. Gao et al. [106] presented the direct heuristic DP with filtered tracking error, to provide a solution for the optimum tracking control problem in the Henon Mapping chaotic system. Tate et al. [107] optimize the consumption of fuel and tailpipe emission for furnished HEV with a dual-mode electric variable transmission (EVT) and a catalytic converter. The SP-SDP controller was capable to provide significantly better performance and trade-offs between emission and fuel consumption [108].
Liu et al. [109] presented an optimal control strategy in Hybrid electric high mobility multipurpose wheeled vehicle (HMMWV) by using SDP, implemented in engine-in-loop setup, to analyze the effect of transient on engine emission. Opila et al. [110] developed a real-time energy management controller to optimal performance in between fuel efficiency and drivability for HEV. The SP-SDP-based controller is 11% additional efficient in comparison to other controllers.

e: INDIRECT ALGORITHM
Pontryagin's Minimum Principle (PMP) is known as an indirect algorithm to solve the optimum control issues, which is derived by Russian mathematician Lev Pontrygain in 1956 to resolve the global optimization problem. It is the expansion of calculus and the Euler Lagrange equation. The main advantage of PMP is that the starting costate is the only calibration parameter for a given driving cycle, which has a significant impact on battery condition. But, it is not suitable in realtime implementation because of starting costate is linked to the driving cycle, and different driving cycles necessitate different optimal initial costate values by which the size of the look-up table grows exponentially and increase the high computational load [111]. Lee et al. [112] developed a control strategy based on PMP and found the best output result in total energy consumption. Later revised Pontryagin's minimum principle algorithm was applied by T. Wu et al. [113] for optimal control in minimizing fuel consumption to prolong the life of lithium-ion batteries. Pérez et al. [114] discuss the finite-dimensional optimization problem to solve the driving cycle equation by PMP and resolve by a programming tool the direct transcription approaches. The size of the table is depending on the dimensions. Therefore, Hou et al. [115] introduced the approximate PMP algorithm based on engine fuel consumption rate, streamlining the Hamiltonian optimization problem into convex optimization problem which applied in-vehicle controller. Onori et al. [9] applied PMP based adaptive supervisory control strategy in less driving information to resolve the energy management problem and improve the fuel consumption by about 20 % in comparison to Optimal PMP, A-PMP, and CD/CS in the vehicle. Rousseau et al. [116] presented the PMP-based heuristic method in the free state for optimal control optimization, which provide the approximately same result as DP in less time. Zhang et al. [117] applied model predictive control (MPC) scheme for real-time optimization in receding horizon optimal problem. Kim et al. [118], describe the mathematical analysis for inequality state constraints in necessary conditions and provide a unique solution for HEV. In [119], PMP algorithms based on instantaneous minimization of the Hamiltonian are applied for real-time optimal control. And provide the closely optimal power solution in HEV via prior knowledge of future driving conditions and suggest to keep proper costate in SOC of battery at desired and predefined level [120]. Stockar et al. [121], minimize the CO 2 emission and utilize the optimal energy in PHEV by PMP. Chasse & Sciarretta [122], presented a chain of tools to develop the EMS for hybrid power trains, to optimize the energy consumption in a real-time environment. Xiao et al. [123] studied the comparison of different energy management methods for a parallel P-HEV and control methods are inferred at different initial SOC of the battery. In terms of computing efficiency, PMP-MPC approach shows a sizable advantage over DP-MPC. In contrast to dynamic programming, PMP-MPC generates solutions with total costs that are equivalent to those of the globally optimum solutions (6.1% and 6.6% departures from DP and PMP, respectively). In light of this, the suggested PMP-MPC by Xie et al. [124] emerges as a practical and appealing substitute for online predictive energy management of plug-in HEVs.
Xie et al. [125] proposed an integrated control strategy for optimizing power distribution in between the auxiliary power unit (APU) and the battery. It is also compared with different techniques to measure its performance in terms of time efficiency and computational accuracy.

f: GRADIENT ALGORITHM
The gradient algorithm is to decreases the calculation time and increases the toughness of the vehicle. This algorithm is more sophisticated with a non-linear model of EVs and HEVs. This algorithm is used as a derivative analytical approach for an objective function. It solves the optimization problem under mathematical conditions, for example by satisfying the Lipschitz condition. In the Linear Programming structure, the algorithms provide the key to optimizing the problems with linear objectives functions, and constraints. Ripaccioli et al. [126] developed a linear and piece-wise affine identification methods-based hybrid dynamical model to illustrate the use of hybrid modeling and MPC for advanced powertrain-based vehicles. Wu et al. [127], proposed mixed-integer linear programming (MILP) based EMS, which synthesized the velocity trajectory via prior knowledge of the real-time traffic condition for optimizing the fuel consumption. In this strategy fuel saved around 10-15% over the binary mode strategy. The MILP is a powerful tool for modeling and resolving problems with continuous and integer variables [128].

h: QUADRATIC PROGRAMMING (QP)
A Quadratic Programming (QP) based EMS is also used to approximate the powertrain model, resulting in a QP structure that is determined by a quadratic cost criterion subject to linear restrictions. It also starts in a quadratically constrained multiple instance learning (MIL) algorithm. The quadratic programming is applied in energy management over DP to reduce the processing time and global solution in a large driving cycle [129]. It is tested on Urban Dynamometer Driving Schedule (UDDS) and Highway Fuel Economy Test (HWFET) driving cycle via Zhou et al. [130] and found better vehicle performance through optimal power management. Constraint QP is applied by Gonsrang et al. [131] to solve the power management problem and found its performance based on nonlinear MPC in power management. Xia et al. [132] presented a quadratic performance index-based control approach in split power HEV to reduce the fuel consumption and it is also restricting the fluctuation of battery SOC.

i: SEQUENTIAL QUADRATIC PROGRAMMING (SQP)
SQP is an iterative approach for nonlinear controlled optimization. SQP tactics on mathematical problems on which the objective function and constraints can be separated twice regularly. Oh et al. [133], design an SQP based control strategy for multi variable optimization to find optimal value in control parameters. The solution of the embedded optimal control problem offers the non-linear characteristics solved via a Sequential quadratic programming algorithm [134]. The optimization problem, which includes a function of cost and inequity restrictions, may be addressed in both convex and affinity forms. The optimization of fuel economy is seen as a nonlinear convex problem (CP) to find the fuel efficiency and analysis the system capabilities [135]. Said et al. [136] described the CP and PMP for the energy management and validate the analytical solution [137] by comparing the obtained results by DP-based original model. Lu et al. [138], proposes the multi-objective optimization problem to solve the device power loss, battery current ripples, and quick charge /discharge ability of ultra-capacitor to stable the dc connection voltage by weighted method and no-preference approach into a convex optimization problem with Karush-Kuhn-Tucker (KKT) conditions.

k: DERIVATIVE-FREE ALGORITHM (DFA)
DFA algorithm applied to solve the optimization problem in which derivative information is unavailable for optimal power management. It can cover the global solution in comparison with the gradient algorithm. The DFA for EMS found in the literature is described below, which is mostly a metaheuristic algorithm.

l: SIMULATING ANNEALING (SA)
SA emerged in 1983 via Kirkpatrick, influenced by the method of annealing the metal. This algorithm uses the stochastic search method to provide a better solution, in which it selects the parameters after changing the objective function. There is very little security to find the global solution. Furthermore, repetitive annealing is exceedingly sluggish, especially when dealing with computationally expensive objective functions. Therefore, this algorithm is used with the combination of another corresponding algorithm to overcome these disadvantages. Delprat et al. discuss the drawback of SA and DP and design an algorithm to reduce the drawback of this optimization technique based on parameter control for the powertrain of HEV [139]. Hui et al. [140] presented an adaptable Simulated Annealing-Genetic Algorithm (SA-GA) to boost the performance of the vehicle's with enhancing fuel efficiency.

m: GENETIC ALGORITHM (GA)
GA is a metaheuristic technique stimulated by the dynamics of evolution. As the first population, it firstly suggests a set of solutions (chromosomes). The results achieved from this first population are evaluated against an objective fitness function. The finest solutions are required more time to develop. Deb et al. [141] proposed the mutation and crossover-based non dominated sorting genetic algorithm applied to find the optimal solution in a computationally complex problem. The sorting mechanism subpopulation of parents Pi & Offspring Qi are evaluated via rank which indicates the conjunction to the crowding distance and optimal Pareto set, which reflects the diversification in solution [142]. The power tracking via GA was adopted in the ADVISOR simulator to optimize the fuel consumption and reduced 17.6% and 9.7% under UDDS and HWFET driving cycle [143]. It provides the solution precisely and avoids to trapped in a local abscissa. In [144], GA optimized the component size of PHEV and reduced the fuel consumption up to 24.38 %. The equivalent fuel consumption minimization tactic discussed to find proper value of conversion factor [122], Optimization in the propagation of abrupt power fluctuation [145] and component size [146], and parameter in EV [93], while data optimize for charging station [147] for the entire region by GA to reduce the problem of excess driving distance. Li et al. [148], proposed the combination of GA with ACA (Ant Colony Algorithm) to acquire the optimal control parameters exactly and efficiently remains an unresolved problem. Ma et al. [149], proposed a robust optimization method for distribution path in EVs to reduce the computational time [150] by GA, which resolves the problem associated with the uncertainty factor in battery charging and distribution of power in EVs.

n: MULTI-OBJECTIVE GENETIC ALGORITHM(MOGA)
The GA with an optimal Pareto result like MOGA can be used to resolve the multi-objective optimization problems. A MOGA was applied to design parameters with respect to fuel consumption, driving cycle performance [151], and operating cost [152], [153]. The fuzzy clustering condition with GA is applied to reduce the computational effort and improve efficiency [72] and electric-assist control strategy (EACS) to curtail fuel utilization and emission, with maintaining the vehicle performance requirement [154]. Poursamad et al. [155] minimize the fuel uses and discharge as well as enhance the driving performance of the vehicle by applying genetic fuzzy control strategy and performed on New European Driving Cycle (NEDC), Federal Test Procedure (FTP), and the car driving cycle. Shahi et al. [156], design a method for optimal control hybridization via Pareto set pursuing (PSP) multi-objective optimization algorithm and powertrain system analysis toolkit (PSAT) on a Toyota Prius PHEV. This algorithm's key advantage is that it takes much less time than an exhaustive search.

o: PARTICLE SWARM OPTIMISATION(PSO)
PSO was developed in 1995 by Kennedy and Eberhart and is concerned with the behavior of community creatures that travel in clusters, for example, ant colonies and flocks of birds in the wild. Participants of this group will exchange knowledge and communicate with all others nearby, reviewing their final finest location and the preeminent solution for the group to achieve an optimal solution. Rule-based control strategies are applied to optimize the fuel consumption for the decision of driving torque demand by a DCWPSO based algorithm and improve the fuel economy by 15.8% European driving cycle while 14.5% worldwide [157]. The effectiveness of Unified PSO justifies by comparing the result with the standard PSO algorithm [158]. Wu et al. [159] proposed a Learning Vector Quantization (LVQ) method based on driving cycle recognition for fuzzy energy management controller optimized by PSO. It is also used to optimize the power-sharing in between the source and component sizing [160] and optimal design variables on a multi-route environment to wireless charging for EV [161]. The fuzzy membership function and fuzzy rules are optimized by PSO in torque distribution in EV for split power management [71]. The various parameter of HEV optimizes to perform a case study on the smaller size of engine & motor, which provide the 22% improvement in fuel economy [162]. Chen et al. optimize the threshold parameter by PSO and reduce up to 1.76% energy loss in uncertain driving cycles [163]. A predictive EMS is applied for EVs to optimize the problems i.e. the minimization of battery consumption, maximizing the temperature comfortable for the driver cabin, and minimizing the travel time by PSO [164]. PSO provides the optimal solution for different cycles in the revised RB strategy [165]. It is a faster, easy, inexpensive, and robust stochastic global optimization technique [166]. The extension of PSO is deal with a multi-objective optimization problem, this method uses the concept of Pareto Dominance and gives effective results as compared to another existing multi-objective optimization [167].

p: OTHER ALGORITHM
The Game Theory (GT) method is used for economic energy management by Dextreit et al. [168]. This method depends on human nature for learning, understanding then action. Yin et al. [169] proposed control strategies based on Game Theory because the energy management is formulated as a non-cooperative current control game, so the Nash Equilibrium analytical method originate for a stable solution to reduce the challenges in energy management by multi-source hybrid energy system (HES). Younis et al. [170] proposed a spreading sampling point-based SEUMRE method to explore the optimal global solution. This method is faster to get the solution in the highly nonlinear problem than GA.

C. ON-LINE BASED STRATEGIES
The offline optimization-based method is not applied in a straightforward way for an online (real-time) control strategy.
The online control strategy is also called causal and local optimization for the reason that these are not required preinformation of the driving cycle. EMS strategies for real-time optimization problems are simply because of limited computation costs and memory resources. In addition, you can avoid manual adjustment of control parameters. The realization of an EMS for real-time optimization can be accomplished in numerous ways. The ECMS and MPC are the famous EMSs for real-time analysis and these have been widely used in various applications.

1) EQUIVALENT CONSUMPTION MINIMIZATION STRATEGIES (ECMS)
Paganelli et al. propose the renowned real-time optimization EMSs known as ECMS which is the realization of offline PMP. The global optimization problem reformulated into limited optimization issues by decreasing the total fuel consumption. The equivalent fuel factor was evaluated by ECMS for the analysis of the fuel consumption to charge the batteries. The EF in ECMS has a similar character as costate in offline PMP. EF is the main key point of ECMS because the control performance of HEV is dependent on it.
So the researcher has focused on the estimation of EF in EVs, dependent randomly on three factors: (i) SOC limits of Battery, (ii) Information of driving cycle (iii) ESS Charging and discharging. Paganelli et al. [171], presented the ECMS for PHEV, which provides the instantaneous power split strategy to optimize the fuel consumption in between ICE and electric machine by charge sustaining mode [172]. This is a real-time minimization strategy for estimating the future driving conditions and can be improved the fuel economy by up to 1% [173]. Pisu et al. [174], analyzed two different energy sources by power split algorithm in modified instantaneous ECMS for a Series Hybrid Vehicle. The comparative result analysis between ECMS and DP-based control strategy discussed by Marano et al. [80], that ECMS algorithm gives the results on Blended Mode control strategy at known driving distance. Musardo et al. [175], proposed a real-time energy management control strategy by adding the fly algorithm on the ECMS framework, known as Adaptive-ECMS. It updates the controlling parameter periodically as per the situation of road traffic and situation. Velocity forecasting is also having a vital role in A-ECMS to optimize the equivalence factor [176], by which it can improve the 3% fuel economy [177]. Won et al. [178], converted the multi-objective nonlinear optimum torque delivery problem into a single objective linear optimization problem by describing an equivalent energy consumption rate for fuel flow frequency and battery charging. Onori et al. [179], design feedback corrected strategies in the A-ECMS controller, which is capable to generate a solution robust and quasi-optimal. It consumes 1-2% more fuel in comparison to another method. Li et al. [180], proposed ECMS based Markov chain model to predict the future driving condition. Sun et al. [181] forecast the velocity by neural network and combined with adaptive ECMS, by this strategies fuel consumption reduced up to 3%. The chaining neural network (CNN) based velocity forecast discussed by Zhang et al. [182] and reduced the fuel consumption up to 5%. Payri et al. [183] describe a unique control approach for optimum power management in HEV, in which the eminent ECMS method upgraded by a stochastic estimation based on past power demand in the vehicles for future driving patterns and applied the S-ECMS method to obtain S parameters for battery energy via log-likelihood ratio. Sciarretta [36] define the equivalent factor for battery charging and discharging on current energy depletion without knowing future driving condition.

2) MODEL PREDICTIVE CONTROL (MPC) BASED STRATEGIES
MPC is a famous technique used in industry to address multi-dimensional constrained control problems. The main purpose of its introduction is to address the DP algorithm issues. When all future information is known ahead of time, the DP's global optimal control can be obtained. For realtime applications, obtaining such conditions in advance is not practical. Therefore, MPC ordinarily consists of three core stages: (i) Measure the optimal control sequence in a predictive horizon which minimizes the cost function subject to constraints; (ii) Apply the first part of the derived optimal control sequence for the physical plant; and (iii) Moving the entire predictive horizon one step forward and repeat step 1. Cairano et al. [184], applied an automatic control via learning of driver behavior through stochastic model predictive control with learning (SMPCL) method in power management. Gomozov et al. [185], applied a computational rate in MPC to provide control of dynamical parameters and horizon prediction. Chaudhuri et al. [186], discussed a hierarchical control strategy in which a higher-level controller is considered to apply traffic signal information and the lowerlevel controller provides the optimum velocity in an MPC framework. Huang et al. [187], proposed a unique anti-idling system for a service vehicle, where coordination in between the different sources is compulsory for efficient operation. In this strategy prior information is unavailable of driving cycle [188], so it will work on the ordinary concept. Here, MPC applies to increase the efficiency of the regenerative auxiliary power system (RAPS). MPC algorithm applied by Johannesson et al. [189], which use the feedback of vehicle position and single nominal drive cycle. It improves the performance by 0.3% over minimum attainable fuel consumption at the studied route. Siampis et al. [190], design the three MPC strategies (Linear MPC, nonlinear MPC-Real Time Iteration (RTI), nonlinear MPC-Primal-Dual Interior-Point (PDIP) for handling the different levels of complexity and compare these strategies on different aspects then found it NMPC-PDIP is the best control strategies in comparison to other. Guo et al. [191], discussed a new formulation of MPC for a continuous-time non-linear system. In the realtime optimization, the GPM combined with MPC, and finite horizon non-linear optimum control problem converted into a linear problem, which is solved by SQP algorithm. The accuracy of the proposed method is higher than the Euler method. Borhan et al. [192], proposed MPC-based full minimization strategy, where the power management problem is divided into two parts. The first part is based on linear time-varying MPC with a quadratic cost function to calculate future control sequence, for minimizing a performance index then applied for the implementation in the computed control sequence. Trovao et al. [193], proposed an EMR approach model for an effective EMS in EV. The joint optimization technique for speed of vehicle and power management is a large-scale non-linear problem [131] discussed by Chen et al. [194] and improves the fuel economy by 6%-14%. Stroe et al. [195] proposed a generic parameterization under the certain assumption to control the energy flow for power management, it is controlled by MPC strategy. Optimal EMS is analyzed for plug-in HEV and reduces the average consumption of fuel [196].

3) OTHER ALGORITHMS a: EXTREMUM SEEKING (ES)
ES is a real-time adaptive optimization algorithm. ES method may be used in a stationary non-linear system to find the real-time extreme value efficiently. It is also called a derivative-free algorithm, which is used to find the optimal functional level in the output function. The objective function is mandatory, to express a sliding surface for following a time-accumulative function and the optimization parameter is chosen by a discontinuous switching function. Bizon [197], proposed the Global Extremum Seeking (GES) algorithm for real-time optimization. GES operation improves the energy efficiency up to 1-2.1% in comparison to Static Feed-Forward (SFF) approach. The application of ESS is studied by using a first-order high-pass & band-pass filter to control the charging/discharging policy of the battery. The band pass filter has the better ability to reduce the load dynamics and improve the durability of ESS [198]. D. Zhou et al. [199], proposed the fractional-order extremum seeking a method to increase the fuel efficiency and durability. It is a more rapid convergence speed and advanced robustness [198].

b: ROBUST CONTROL (RC)
Robust control (RC) aims at determining an output feedback controller which minimizes fuel consumption. In estimation, RC can handle parametric uncertainties, sensor noises, and defects, assuring stability and toughness. However, RC can only produce a sub-optimum result due to the translation of a nonlinear time-variant system into a linear time-invariant. P. Pisu et al. [200], discussed the comparative analysis of four different techniques (Finite State Machine, H-Infinity Control, Adaptive Equivalent Consumption Minimization, and Dynamic Programming) for Energy Management in a sports vehicle. Zaher et al. [201], design optimal robust control for real-time EMS. In this strategy, mechanical wastage is used to regenerate the energy for the energy storage device for later usage. This hybrid configuration reduces the fuel consumption on the machine by 20-30% at different drive cycles.

c: DECOUPLED CONTROL METHOD
Decoupled Control (DC) is a model-based control approach, which applied to resolve competing performance targets, for example, fuel efficiency, regulation, and drivability SOC. By developing the dynamic model powertrain arrangement, the battery regulator and drivability control decoupled by using the constraint on power demand and vice-versa. Chen et al. [202], investigated problems regarding fuel economy and vehicle speed in HEV. Besides this, it also analyses the energy management strategy for the electric power train. Here author separately analyses the optimization problem for power train losses and the speed characteristics. Two sliding mode controllers detached the DC strategy's control operations. In which, the first-order sliding mode precisely regulates the dc-bus voltage with the modest voltage drip caused by rapid load fluctuations, while the second-order sliding mode produces less chatter and recovers earlier from voltage losses.

d: PSEUDO SPECTRAL OPTIMAL METHOD
A pseudospectral control method is an additional modern optimization-based mathematical method stretched to an energy management system. A direct method resolves the optimum control issues. In which PSOC set down an optimum control problem into a nonlinear problem (NLP) by parameterizing the state and control variables in a series of collocation nodes using global polynomials. Li [203], studied a low-temperature characteristic of HESS and its structure for better utilization of power density of UC by applying a logic threshold control strategy to decrease the power loss. The pseudo spectral method has a better control effect than the LTCS and has good real-time, high reliability, and high-efficiency characteristics. Dosthosseini et al. [204], proposed direct method (Legendre, Chebyshev, and Haar Wavelets polynomials) to reduce the optimal control problem with inequality constraints by orthogonal function. This method does not require discretization of the control problem. The PSOC was engaged in seeking the best global results integrated into a logic threshold management approach.

D. LEARNING BASED ENERGY MANAGEMENT STRATEGY
Learning-based EMS works on sophisticated schemes of data mining for broad real-time and historical knowledge to extract the maximum control legislation. The exact model information is not required in the LB-EMS to decide on the control. It is conversely challenging and time taking to establish an exact database whose configuration and size directly influence the controller performance. Machine learning and data-driven approaches are versatile and capable handle massive data sets under varying driving environments and drivers outside.

1) MACHINE LEARNING BASED EMSS
ML-based EMS is used widespread in Intelligent Transport Systems. Boyah et al. [205], designed a neuro-dynamic programming-based real-time controller to get the optimum result in power distribution while computational complexity and the resulting burden are very critical. The Quality(Q)-learning-based vehicle learning system combined with neuro-dynamic programming (NDP) discussed by Liu et al. [206] to estimate expected energy costs in near future. Yanqing et al. [207], developed an instancebased machine learning algorithm to learn the rolling driving condition that can be predicted by the k-Nearest Neighbor (k-NN) algorithm. Venditti [208], addressed the performance of cluster optimization & rule extraction (CORE) and cluster extraction & rule optimization (CERO) and compare with DP, and provide the almost same result with the small discrepancy about 1.84% and 4.85 %. Langari et al. [209] implemented an intelligent energy management agent (IEMA) whose role is to assess the driving environment, by which learning vector quantization (LVQ) network can efficiently determine the driving condition within a limited time period of driving data.

2) REINFORCEMENT LEARNING (RL) METHOD
A RL framework contains two modules: a learning agent, and an environment in which the learning agent communicates with the environment continuously. At every point in time, the learning agent obtains an assessment of the status of the world. After that, the learning agent takes an action to perform, which is later implemented in the environment. The environment then shifts to a new state as a result of the action, and the reward connected with the shift is calculated and communicated back to the learner. The agent receives an immediate reward along with each state change, to establish a strategy of control that records the existing state to the appropriate control decision on a particular place. Deep Reinforcement Learning (DRL) based EMS associate as a deep NN with a conservative RL, called a deep Q-network. Recently, some RL-based EMSs have been recorded. Lin et al. [210], proposed RL technique for optimal power management in HEV, without prior knowledge of driving cycle. It improves the fuel economy up to 42%, while Lee et al. [211]obtain that the RL-based approach is more suitable for a time-variant controller with boundary value limitations. He and Cao [212], proposed a restructured algorithm framework based on Deep Q-learning (DQN) to obtain a better pedal's control strategy. The deficiencies in traditional approaches are discussed by Liessner et al. [213] and offered a deep reinforcement learning framework to overcome these deficiencies. Deep RL is capable to achieve optimal fuel consumption [214]. The prior route information is not required as DP, so it can apply to line vehicles [215]. Wu et al. [216], applied DP for split power optimization, FLC for dynamically adjusting coefficient α, and reinforcement learning applied as an online correction algorithm to resolve the optimal control problem and obtain 4% improvement in fuel economy. Liu et al. [217], proposed an energy management strategy based on a Dyna agent of RL approach for optimizing fuel efficiency and improving fuel economies. Here, the author made a relative analysis of the one-step Q-learning, Dyna, and Dyna-H algorithms. Xia et al. [218], proposed a real-world driving data-based energy management to recover the energy consumption in PHEV and reduce the computational complexity of the optimization method, while the K-means Clustering method is used to calculate the sensitivity of new factor to the energy consumption.

a: NEURAL NETWORK BASED LEARNING METHOD
The learning approach based on neural networks is modeled after human brain neurons. As with a real neuron, which contains a plethora of connections, A neural network's nodes are objects with numerous inputs and outputs. Various kinds of behaviors would be modeled by involving multiple neurons in layers that make a network. The three-layer neural network optimization controller was designed for energy management issues by Xie [219]. Raj et al. [220], proposed the loss model control (LMC) and search control (SC) method for optimal control and also discussed the different optimization techniques such as ANN, FL, GA, Nature Inspired Algorithm (NIA), an evolutionary algorithm.
The evolutionary algorithm discussed by Potvin [221] addresses difficult vehicle routing problems in many different ways in which branch cut and price algorithms are used to resolve the capacitated vehicle routing problem [222] Baldacci et al. [223], discussed an innovative approach to solve the vehicle routing problem and retract performance and comparison analysis of a different exact algorithm for the time window [224] with VRP and Capacitated VRP. Huang et al. [225], designed an intelligent vehicle control system, which is based on both membership functions and control rule base neural fuzzy network tuned by mixed genetic/gradient parameter algorithm to obtain an optimum control performance in the vehicle. This technique is also used in the load-leveling strategy, which consists of fuel economy and reduced emission for a different driving pattern [226]. Rubaai et al. [227], presented a training set for Fuzzy NN, which included different methods such as back-propagation (BP), extended Kalman filter (EKF), Genetic (GEN), PSO, and found the EKF is the best learning method in pattern matching. Lee et al. [228], define the ANN-based fuel economy as much better than others and Power Split Ratio (PSR) technique is very simple and robust. Wang et al. [229] applied an Elman neural network (NN) in an optimal EMS, and reduced the fuel consumption by 9.1% & 24.6% in comparison to logic threshold & conventional ICE bus. Park et al. [230] developed an ML algorithm to learn the efficiency of different road types and traffic jamming levels, as well as a neural learning algorithm for learning the NN to forecast the road type traffic jamming levels. These are all things processed by the University of Michigan-Dearborn intelligent power controller (UMD_IPC). Murphey et al. [231] developed an energy optimization technique with ML framework tuned all three online intellectual controllers, IEC_HEV_SISE, IEC_HEV_MIME & IEC_HEV_MISE integrate into ford escape hybrid vehicle for real-time performance calculation and found IEC_HEV_MISE give the best performance by saving the fuel consumption from 5% to 19% as compared to ford escape controller. Long [232], studied the EKF for estimation of the SOC [233] of the battery based on Stochastic FNN. This model is used for the non-linear dynamic model of battery and filter the effect of noise at the input. The online SOC estimation of the Li-ion battery [234] is operated by the adaptive Luenberger observer method and the function parameter are optimized by the least square algorithm [235] which are giving the better result in comparison others.

b: PATTERN RECOGNITION BASED EMSS
This method is based on driving cycle behavior which is part of the Intelligent Transportation System (ITS). GPS has a big role to identify the traffic condition and road map. Driving behavior-based EMS have been reported. Marano et al. [236], discuss the intelligent transport system, in which information is given by different vehicles to everything, enables the power management system for small & long distance power splitting for improved fuel efficiency. The prediction mechanism for a piece of prior knowledge about future driving cycles is discussed on the ML framework with the combination of DP [237]. Fan et al. [238], proposed a map-based approach for optimal energy management to decrease consumption and enhance the economies of parallel plug-in HEV. Wu et al. [239],parallel chaos optimization algorithms are used to optimize control strategy, instability of torsional [240] as well as to minimize the cost, while an intellectual multidimensional statistical method is used to discriminate automatically the driving condition of HEV [241].

3) NATURE INSPIRED ALGORITHM BASED EMSS
Some nature-inspired new algorithms applied to EMS in which the author discusses the results and application in HEV/FEV. Hmidi et al. [242] presented the meta-heuristicbased grey wolf optimization (GWO) algorithm for optimal energy management on fuel consumption, CO 2 emissionand optimal gain absorb in the urban cycle is 13.9 % in comparison to simple rule-based strategy. Improved binary GWO and give the better experimental result in comparison to the GA and PSO [243]. Multi-objective GWO is used to minimize the power loss and voltage abnormality in the supply system [244]. Ullah et al. [245] proposed the bioinspired Grasshopper-Optimization Algorithm and Cuckoo Search Algorithm to reduce the energy consumption budget, peak-to-normal power ratio, and quick time responses because of different loads. Preetha et al. [246] proposed ALO (Ant Lion Optimizer) to solve the energy management problem and also compare with GA, PSO, BAT algorithm, while the Salp Swarm Algorithm is used to optimize the energy consumption and cost [247]. Kayalvizhi et al. [248] applied the firefly algorithm to optimize the power consumption from the battery and devices automatically switching by dynamic EDF based on allocated priority. Liu et al. [249] optimize the real-time constraint to charge completion strategy by the grey wolf optimizer method. Mohseni et al. [250] proposed the Whale Optimization Algorithm (WOA) to reduce the computational burden on microgrids for energy management which is required lower iterations in comparison to PSO and GA. This is also coordinate to energy management in PV-BES units and electric vehicles [251]. Trovao et al. [252], presented a technological approach for simultaneous optimization by simulating an annealing metaheuristic for power and energy management, for a supercapacitor and battery, in the electric vehicle. Bagherzadeh et al. [253] proposed Salp Swarm Algorithm (SSA) to decrease the objective function of the problem to determine the optimal location and capacity of RES.SSA optimizer showed its preeminence with great attitude and correctness in problem-solving of renewable distributed generators (RDGs) and shunt capacitor banks (SCBs) [254]. Deb et al. [255] proposed the Chicken Swarm Optimization (CSO) with Ant Lion Optimization (ALO) for efectively resolve the and efciently coping with the charger placement problem.

4) OTHER ALGORITHM FOR EMSs a: NEURO-FUZZY METHOD
Neuro-fuzzy is a term used in artificial intelligence to describe a method that combines fuzzy logic and artificial neural networks. Mascioli et al. [256], proposed a control technique to optimize the energetic flows via a neuro-fuzzy approach with the vehicle state especially of their energy consumption in HEV.

b: ACTION-DEPENDENT HEURISTIC DYNAMIC PROGRAMMING (ADHDP)
An algorithm for approximative dynamic programming is action-dependent heuristic dynamic programming. It is not necessary to have a system model that is explicit. use the Action and Critic, two neural networks. Over a predetermined period of time, this plan can minimise a specific Utility Function. Hui et al. [257] examined issues regarding lowering average cost over a period of time for electric vehicles by proposing a strategy to optimally control vehicles in a heterogeneous vehicular network. An ecological adaptive cruise control for HEV in the car, following scenario to optimize the fuel consumption. As well as ADH-DP in adaptive cruise control (ACC) is used to optimize and maintain the velocity and inter-vehicle distance in normal driving conditions [258].

c: SHUFFLED COMPLEX EVOLUTION (SCE) ALGORITHM
This algorithm developed by Duan et al. is useful for calibrating hydrological models. This approach, a type of differential evolution (DE), is effective because it uses geometric operations to look for potential optimal solutions to space parameter problems. Chen et al. [259], studied an adaptive tracking control, for nonlinear stochastic systems. VOLUME 10, 2022 The shuffled Complex Evolution (SCE) algorithm resolves the waiting time problem in the optimization model of the charging station [260]. Memetic Algorithms are known as evolutionary algorithms that use local search technique rather than global search technique to refine individuals. Memetic Algorithm (MA) is used to reduce fuel consumption and emissions [261]. Dijkstra's technique use to determine the shortest path between any two graph vertices. In [262] Tribioli, presented Dijkstra algorithm (DA), which is much more computationally efficient than any other optimization technique.

f: COYOTE OPTIMIZATION ALGORITHM
Pierezan and Coelho debuted a brand-new meta-heuristic Coyote Optimization Algorithm (COA) in 2018. The algorithm is based on how the coyote adapts to its surroundings and exchanges experiences with other coyotes. Fathy et al. [263], discussed the Coyote Optimization Algorithm which moderates hydrogen consumption by 38.8% in comparison to the EEMS technique and got the first ranked in between GWO, SSA, GOA, MVO, GA, PSO, and EEMS based on the lowest hydrogen feeding.

V. DISCUSSION & OUTLOOK
In light of the previous section's analysis, it is clearly seen that researchers have made significant efforts in the field of EMSs for HEV and FEV, with promising outcomes.
However, the recent rapid developments in the application of smart transportation systems, developing innovations in powertrain components, and computational methodologies have created tremendous opportunities to improve EMS performance. With the current evaluation of renewable energy charging systems and new communicative techniques like a vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and Automated connected vehicle (ACV). Auspicious potential required to unleash for additional development in driving performance and fuel budget. Hence, this segment looks at perspectives that haven't been discussed before and have received little attention but could be expected as future research directions in this field.
To that purpose, an overview of vehicle configuration and their application of HEV has been shown in Table 1 and found that traditional ICE vehicles generate high greenhouse gas emissions and low-efficiency drivetrain. IC Engine also suffered with the inherent and pronounced delay in torque production [264]. Hence IC engine revved-up by torque filling and boosting to achieve maximum power and torque. Electric automobiles are much quicker than their combustion counterparts [265]. This is due to the fact that electric automobiles can create high torque right away, whereas combustion engines must acquire speed to achieve that torque. So that vehicle transmuting in electric vehicles, have an alternate solution [63]. Ever since Lohner Porshe, developed the first HEV in 1901. HEV technology has gotten a lot of attention in terms of research and development, but when people talk about the benefits of HEVs, they typically forget the drawbacks, like; limited range, low power, expensive costs, maintenance costs, batteries. The current state of the HEV power management approach is summarized from the standpoints of real-time implementation and optimum forecast capabilities.
RB-EMS seems to be the only method that has demonstrated successful capabilities in commercially implemented real-time systems, however, it falls short of providing the optimum solution. OB-EMS overcomes the inherent drawback of RB-EMS through an optimization control approach. In the same way that offline OB-EMS are being challenged in terms of their application in online use due to computational burden. Table 6 lists the benefits and drawbacks of the primary EMS under investigation.
It has been observed that none of them can address all of the control objectives' criteria at the same time. Therefore, numerous researchers have used different optimization algorithms to enhance EMS performance by combining their VOLUME 10, 2022 complementing qualities. In terms of optimization, it appears that the majority of the research has emphasized the usage of older algorithms (Eg. PSO, GA, and SA) for OB-EMS control. However, the literature has more than 40 different nature-inspired algorithms [266]. There are a plethora, black widow [267] of novel algorithms that have not been used in the EMS optimization sector among EVs.
In terms of optimization, incorporating more recently developed algorithms into EMS applications, particularly MPPT in solar PV-based FEV, would be a promising field of research. The introduction of a new algorithm will help in the field of computational cost, efficiently handling complex multiple objective cases in the direction of extremely necessary raw data that are received at the input of any intelligent EMS.
The offline EMS aims to reduce worldwide fuel usage. Even though they cannot be directly deployed in real vehicles, they serve as a benchmark for other EMS and receiving modified online EMS. As ITS technology has advanced, driving cycle prediction has become increasingly crucial in predictive EMS. They are more adaptable and perform better than other EMS. In addition, infrastructure that can recognize, save, and combine datasets of traffic paths, vehicles, weather, road signs, preceding cars, speed, and other factors at the same time and use them for forecast purposes should be explored. Finally, EMS can be expanded to include multitime scalar multi-vehicle interactions as well as several information layers.
The use of the OB-Algorithm in conjunction with machine learning techniques can help speed up the evaluation of larger space out EMS. In this aspect, thanks to the new smart devices, EMS now considers a fleet of vehicles rather than a single vehicle when interacting with the smart grid and optimizing charging rates. The primary goal of these initiatives is to boost road capacity and overall performance in all aspects. These methods are mostly used in heavy-duty applications. like: city buses. It is expected that groups of passenger vehicles would be thriving research topics in the future. Which will be the designing EMS framework for smart and sustainable city concepts.
The discussed item can be summarized in an integrated EMS (i-EMS) concept which included the level of information (like: Data from the server, ITS, V2V, V2G, GPS & Traffic Lights), Time Horizons, and the number of the vehicle (Transportation Level). Integrated EMS can be considered various integration possibilities for future research trends: Waste Heat Recovery (WHR) System. Some dynamic behaviors, including battery temperature, catalyst temperature, engine out temperature, and engine cold start circumstances, can have an impact on the WHR system.
The proposed methods for incorporating the investigated components into an optimal power control problem, previously utilized effective methodologies require the development of high-fidelity models that include the engine and battery's dynamic transient behavior. Self-learning and model-based control systems that can autonomously decide the best control settings on the road would be a solution to the shortcomings of a typical EMS based on quasi-static and map-based models, in the coming years.
The integration of several control layers into a concrete holistic EMS framework will be one of the future research trends. The inclusion of eco-driving into an EMS for the double vehicle level thru an Adaptive/Predictive cruise control approach as at multiple vehicles platooning warrants consideration as a potential field in the coming years with the support of cyber-physical systems.

VI. CONCLUSION
According to the review, many researchers are becoming more interested in the design features of powertrains and EMSs for hybrid and electric vehicles. To address control goals such as decreasing fuel consumption and emissions, preserving ESS charges, and increasing drivability and vehicle performance, many topologies for powertrains and associated EMSs have been suggested. In creating energy management techniques, there is a trade-off between optimality and execution. The advantages and disadvantages of different optimization techniques & algorithms are shown in Table 3, 4 & 5. All energy management techniques are influenced by the driving cycle, and the application of all the optimization methods is shown in Table 6. The first time, focus on intelligent transport systems for improving the vehicle performance and uses the recently originated metaheuristic algorithm inspired by nature. In light of current advancements in smart and information-based methods, it has been suggested that new frameworks/algorithms, communication ideas, technologies, and infrastructure be incorporated into the design of an EMS to overcome existing uncertainties and attain real-time robustness.