Adaptive DDL Algorithm to Elucidate the Protection Misoperation in Malaysian Rapid Rail DC Traction System

In modern railway traffic systems, direct current (DC) electrification is a prevalent choice, with numerous traction networks adopting a variety of voltage levels to accommodate varying load current dynamics. These dynamics are influenced by passenger density, aggregate demand for electrical power, headways, and frequency of locomotive operations. Load currents are prone to surges during periods of dense traffic and transient phases such as acceleration, deceleration, and the start–stop sequences of trains. Such surges hold the potential to precipitate fault currents within the traction system, which are similar to those engendered by external anomalies. Conventional protection systems, such as the Détection Défaut Ligne’—French for ’Line Fault Detection), may not always effectively identify remote faults or prolonged overcurrent situations. These scenarios necessitate an advancement beyond the traditional fault detection methodologies, which primarily rely on fixed thresholds and may not account for the dynamic nature of the railway system’s electrical load. This paper addresses the limitations inherent in the existing DDL protection mechanisms by focusing on the feeder attributes specific to the DC Traction System. In pursuit of this objective, we introduce an innovative adaptive current DDL algorithm to refine the rigid threshold paradigm inherent in the conventional approach. To facilitate a pragmatic assessment, the Rapid Rail network of Malaysia serves as a reference for emulating the railway’s electrical system. This comprehensive analysis yields insights that are potentially useful for safety protocols in DC electrified railroad traffic systems.


I. INTRODUCTION A. BACKGROUND
The accelerated development of Malaysia's rapid rail services, including Mass Rail Transit and Light Rail Transit (LRT) in support of the Greater Kuala Lumpur The associate editor coordinating the review of this manuscript and approving it for publication was Wei Quan.
Conurbation within the framework of the Klang Valley Integrated Transit System (KVITS) [1] has inadvertently increased the frequency of fault occurrences within traction systems.Such faults result in heightened security and reliability concerns for both humans and equipment involved in traction.The burgeoning demands on railway transportation infrastructure have illuminated critical vulnerabilities in traditional overcurrent protection systems.These systems, notably in DC traction operations, are challenged by the low line voltage and high impedance that characterize the network, leading to fault currents of magnitudes similar to the maximum inrush starting currents emanating from the traction power system substations.These circumstances precipitate frequent failures by modern current DDL (Détection Défaut Ligne) overcurrent protective relays, which often cannot distinguish between regular fluctuations and sustained overcurrent faults.Consequently, the reliability of the DDL system is called into question, particularly as it struggles to adjust to the evolving impedance landscape of the railway system.This evolution has been driven by extensive enhancements to the electrical railway infrastructure, such as the modernization of power rails, expansion of supply lines, and integration of additional traction motors.
These systemic changes and the resulting challenges underscore the pressing need for more sophisticated and responsive protection mechanisms [2].Such a mechanism must not only accommodate the complex dynamics of an expanding railway network but also ensure the unerring detection and isolation of faults, thereby safeguarding the continuous operation and integrity of the traction system.As the rail network continues to grow and adapt, so too must the protective measures that underpin its safe and reliable functioning, to assure the well-being of passengers and the preservation of critical transport equipment.

B. LITERATURE REVIEW
The global railway transportation infrastructure relies on either alternation current (AC) or DC electrical power systems for train mobility.DC systems have been widely adopted internationally as the preferred solution for many rail networks.There are several reasons why DC traction has become the standard choice.DC trains tend to be lighter in weight, have lower capital costs, and have more energy-efficient operation than AC rail vehicles.Additional advantages of DC electrification include a lack of reactive power concerns and frequency synchronization, simplifying the system design.
The DC power supply is typically provided through rectifier substations located along the rail line that convert incoming high-voltage AC into DC [3].This allows electric locomotives and multiple units to be powered through third-rail or overhead catenary lines using lower-voltage DC.While DC systems have achieved broad implementation, further advancement has been hindered by some challenges.The chief impediment to more widespread DC network development is the lack of a robust lightning and short-circuit protection scheme.Lightning strikes pose a risk of damage to rectifiers and other electrical equipment without sufficient safeguards [4].Short circuits, if not isolated and interrupted quickly, can also disrupt train operations on tracks protected by the affected substation [5].The simplicity and efficiency of DC traction have made it the preferred solution to energize rail traffic for many operators globally.Continued research aimed at enhancing DC system reliability and resilience, such as through advances in lightning and short circuit protection schemes, may allow the benefits of DC electrification to be leveraged even more extensively in the future.Reliable and uninterrupted train movements will also be better supported through improvements to critical infrastructure components.
If the protective systems fail to recognize the fault rapidly, the rail transit system's operation is disrupted, as is the supply system's solidity.Such conditions undermine the ability of traditional overcurrent relays to accurately discern faults from routine operational disturbances such as motor starting currents.Failure to promptly detect and isolate faults risks damage to equipment and threatens continuous service availability across the network.The most prevalent DC traction feeder protection systems are based on the amplitude of the feeder current.Because of the amplitude dependence, determining a good threshold that may offer sufficient sensitivity and reliability for the protective system under various operating situations is difficult.If the overcurrent relay's pickup current is adjusted such that it does not erroneously trip for the train's initial inrush current, it will not function for low current faults, and if the pickup current is set to detect low current faults, it will malfunction if the train begins to move [6].Malfunctions or protection failures could also endanger worker and passenger safety in some failure scenarios.To ensure adequate line protection within a particular fault current range around the start-up current, contemporary DC traction systems have already incorporated an overcurrent protection known as DDL [2].The protective relays installed in the Traction Power System Substation (TPSS) were instantaneous or time-delayed DDL protection, which were set to send the tripping signal to the circuit breaker when the current remained above the maximum current threshold for a certain amount of time [7].
Recognizing these challenges, developing an advanced protection scheme is necessary to ensure the long-term viability and reliability of railroad electrification.An ideal solution would reliably detect and differentiate faults from other events that produce high currents, such as motor starts.This would allow faults to be rapidly cleared while avoiding unnecessary tripping under normal operations, and the system should detect sustained overcurrent faults.By improving the situational awareness of system conditions, a suitable new protection technology could help railway operators safely optimize asset use over time.Maintaining continuous service is also crucial for meeting passenger transport demands and supporting supply chain dependability.With further innovation, fault protection can be strengthened in a way that enables rail infrastructure to adapt safely to increasing use.

C. RESEARCH GAP
Various studies have explored supplemental methods for safeguarding DC railway networks with differing levels of achievement.One approach discussed in [8] centered on computing the ratio of a feedline's peak current to its steady-state current.However, the velocity of this technique was not sufficiently rapid, and uncertainties persisted concerning the dependability of its triggering points.While this approach provided a framework for analysis, the speed at which faults could be detected and the reliability of the triggers left practical challenges for implementation in a railway safety system.
A structured neural network strategy for railway safety under a combination of AC and DC grids has been described previously [9].However, the extensive training data requirements of neural networks were considered disadvantageous for practical implementation.Neural networks show promise for complex pattern recognition tasks.However, the large volumes of detailed training data required to develop accurate network models present logistical obstacles to incorporating this approach into an operational railway protection system [10].However, this technique requires careful selection of the appropriate mother wavelet for analysis.It also mandated a high sampling rate and carried the risk of being susceptible to noise that could interfere with accurate detection.While wavelet analysis provides a powerful tool for signal processing, selecting the optimal wavelet basis function and sampling fast enough to capture transient events increases the complexity of implementation and heavily depends on the specific characteristics of the railway system.Environmental noise also poses a challenge to the reliability of detection.
An auxiliary approach based on the traveling wave theory is provided in reference [11].However, to identify the moving waves, this approach required an exceptionally high sample rate that faced pragmatic hurdles for operation.The traveling wave analysis shows the potential for locating faults.However, the very high data collection rates required to track wave propagation may exceed the technical capabilities or cost considerations of equipment deployed in operational railroad networks.
An amplitude-based low-current defect detection approach that uses publicly accessible data on energy consumption trends and internal load records from various train systems.The proposed method analyzes fluctuations in the overall load amplitudes to identify potential faults occurring on distribution lines with low currents [12].However, successfully implementing this technique in practice and determining the optimal threshold values for identifying faults proved challenging.Factors such as natural variations in public demand and periodic equipment start-ups or shutdowns on train networks complicate the setting of precise thresholds needed to distinguish routine load changes from actual underground faults.Although the amplitude-based approach showed promise as a low-cost solution for monitoring light-loaded distribution circuits, further refinement of the threshold methodology is necessary before widespread industrial adoption.
Various techniques have been devised to address the challenges inherent to differential current time step feeder protection systems.One of the more commonly employed alternative approaches for the overcurrent protection of DC traction feeders is the DDL method.The DDL approach bases fault detection on the analysis of the rate of change of current and current increment [13].

D. MOTIVATION CONTRIBUTION
The motivation for advancing adaptive DDL protection schemes is underscored by the necessity for enhanced reliability and responsiveness in DC traction power systems.The conventional DDL methodology, while robust and cost-effective, is not without its limitations, chiefly its reliance on static threshold values, which may not accurately reflect the dynamic conditions of the system.This becomes particularly problematic with sustained overcurrent faults, which may not trigger the existing protection mechanisms on time, leading to potential system hazards and inefficiencies.In addressing these challenges, our proposed adaptive DDL protection technique represents a significant step forward.By integrating an adaptive threshold concept and an improved DDL algorithm, the proposed method promises a more responsive and accurate fault detection capability.This adaptability ensures that the protection system is not only tuned to the system's present conditions but is also capable of accommodating the natural variability inherent in DC traction systems, such as fluctuating impedance parameters and shifting electrical configurations.Specifically, the DDL methodology monitors the feeder current and di/dt as well as the difference between successive current measurements, separated by T. Significant or abrupt changes to either di/dt or T values across that predefined period may indicate that a fault has occurred on the line by tracking fluctuations in both the rate of current rise and the step change in current magnitude.DDL protection methods require accurate information regarding the dynamic impedances of the DC traction system.Such impedances include those of DC power lines from the TPSS, DC power rails, return rails, trains, and so on, which are required for establishing precise relay settings of operation.Nonetheless, these impedance parameters are inherently changing and depend on the operating conditions.
In addition, the proper operation of the DDL is dependent on the train distance from the substation, the fault-occurring position, and the perpetually shifting electrical configuration of the traction power system [14], particularly while attaining the KVITS.This circumstance is quite akin to the problem of changeable impedance in typical power systems [15].The research delineated in the following sections aims to substantiate the viability of this advanced scheme.This study seeks to demonstrate how adaptive DDL protection can effectively mitigate the risks associated with sustained overcurrent faults, thus ensuring the safe and efficient operation of DC traction power systems.This not only enhances operational reliability but also paves the way for more sophisticated protection strategies in the industry, reaffirming the commitment to innovation and continuous improvement in power system protection.

II. MODELING THE LRT-RAILWAY ELECTRIFICATION SYSTEM
The electric railway simulation of a 750V DC power supply generated from a 33-kV substation replicates the modern rapid rail system utilized across LRT, Malaysia.This configuration is used to study the protection system's operation in a realistic manner that accurately models the real-world conditions of the Malaysian rail infrastructure.The schematic representation of the TPSS is shown in Fig. 1.
The key components include a grid connection, a 33-kilovolt (kV) feeder bus, and a 12-pulse rectifier transformer.A threephase 33-kV AC power supply at 50 Hertz (Hz) periodically distributes traction power across the railway network in a regulated fashion through this critical infrastructure.
A rectifier transformer is used to convert the voltage from AC to DC through a process known as rectification.A cable connects the rectifier circuits to the transformer's low-voltage secondary circuit, with the cable running to the rectifier's AC side.Traction substations house both rectifiers and step-down transformers to facilitate the efficient conversion of power for railway locomotives and carriages.Transformers lower voltages in a controlled manner by exploiting the magnetic coupling of coils.Rectifiers then convert the stepped-down alternating current to the direct current required to power the trains.
Within each substation, the rectifier converts the three-phase power source obtained from the grid and transformer to the direct current voltage used on the electrified rails to propel trains.To achieve high-fidelity modeling by replicating real-world railway networks, the simulation includes two equivalent substations (A and B) in Fig. 1.While the core components of transformers and rectifiers remain consistent, modern rapid transit systems commonly employ equivalent 24-pulse rectification systems comprising two coupled parallel 12-pulse rectifiers for each substation [16].This configuration, illustrated in Figure 2, has been adopted because of advances in power electronics that facilitate more efficient rectification and reliable delivery of traction power.The simulation models short-circuit faults using a fluctuating resistance value, allowing faults to originate either within the train or externally on the electrified track section.Internal failures occur inside the train or pass through the DC link capacitor.In contrast, exterior faults develop on the third rail, potentially causing track-related issues that could disrupt service.Within a DC-powered system, short-circuit current rises sharply, making quick detection critical to prevent overburdening of the power supply infrastructure and potential component failures in severe cases that could endanger passenger safety.
To replicate the 24-pulse equivalent rectification systems, each modeled traction substation contains a pair of coupled 12-pulse rectifier units.The windings of the two rectifier transformers have a 7-degree phase shift, achieving greater rectification efficiency.Comparable resistance and inductance parameters simulate the electrical characteristics of contacts on the running rails as well as the operating resistance and inductance exhibited by trains, accurately representing real-world railroad conditions.An accurate frequency-dependent model for the entire rail system network is proposed in [17].Within the DC-DC converter subsystem, a buck chopper uses two gate turn-off thyristors (GTO) switched 180 degrees out of phase at a frequency of 250 Hz each, resulting in a total chopper operating frequency of 500 Hz.The Voltage source inverter (VSI) with a maximum pulse frequency of 400 Hz, complies with the requirements of the real-time simulation system.
A short simulation period is necessary to analyze short circuits within the ultimate duration of 200 ms.The VSI is modeled as a GTO/diode bridge within the internal simulation parameters.GTO was developed with 0.01 Cl on resistance.The rectifier diode is modeled with an off-state resistance of 106 and an on-state resistance of 0.1 .The traction load model is calculated based on the tractive force (F) and the load torque (T M ) of the track as [18].
The train dynamics are given as The wheel dynamics are given as where N is the normal wheel-rail contact force, µ the adhesion coefficient, M is the train mass, r is wheel radius, T e is the motor torque, ω w is the wheel angular speed, J is the inertia, A & B are the static and rotational friction coefficients, R is the resistance to motion, GR is the gear ratio, θ is the gradient, R is the resistance to motion and ω m is the rotor angular speed.The rotor dynamics of the motor are given as The wheel's angular speed is represented as The wheel dynamics may be represented in terms of rotor angular speed by replacing (6) with ( 5) The simulation accurately models train dynamics by incorporating coefficients that reflect friction, inertia, and counteracting load torque, thereby emulating the actual performance of trains in operation.The torque T e in Equation (7) integrates the DC Motor in the train [18] as depicted in Fig 2 .The parameters of the DC railway traction system and Train load parameters selected based on the LRT train are provided in Table 1.For simulation purposes, each train car is equipped with four traction motors.These motors are collectively represented by a single equivalent induction motor with specified parameters: a voltage rating of 554 volts and a power capacity of 560 kilowatts.

III. CONVENTIONAL CURRENT DDL PROTECTION SCHEME
A protection mechanism known as DDL has already been implemented in modern DC traction systems to ensure effective line protection within specific fault current thresholds closely associated with start-up current levels [19].The notations used in the DDL algorithm are mentioned in Table 2  conditions on the line.When di dt rises above a predefined, configured threshold, the algorithm automatically records the exact time that this threshold is exceeded.Sudden, significant increases or decreases in the current level may indicate issues such as short circuits, ground faults, or other electrical anomalies and irregularities that require attention.This precise timing data is then used to determine whether the observed change in the current rate was transient, as seen during normal operations like train acceleration cycles, or potentially indicative of an underlying fault that warrants further examination [20].The DDL algorithm performs ongoing calculations to determine the precise di dt across the electrical line continuously.It then carefully compares the calculated di dt values against pre-configured threshold settings that are designed to effectively differentiate between standard operational variations and readings more suggestive of actual fault conditions requiring protective responses or corrective actions.If both di dt and the change in current ( I) exceed their respective thresholds continuously for longer than the maximum allowed time ( T), the algorithm immediately initiates the predefined protection response.This response aims to promptly isolate and clear the detected fault to prevent equipment damage, ensure safe and reliable system operations, and maintain compliance with industry best practices, standards, and guidelines.If no monitored parameter surpasses its threshold for the necessary duration, the algorithm will determine that no fault currently exists and protection actions are not required.
Once the necessary protection response is executed, the algorithm will automatically reset and smoothly return to its ongoing, continuous monitoring of the current level across the distribution line.This allows it to remain vigilant and quickly identify any subsequent events that could require attention, thereby ensuring comprehensive protection for the electrical infrastructure.The conventional current-based DDL protection algorithm likely represents one critical component of the overall protection system, incorporating overcurrent and undervoltage logic.This algorithm is precisely designed for high sensitivity to current changes while avoiding unnecessary trips that could disrupt operations or reduce system availability, reliability, and efficiency.

A. DRAWBACKS OF THE EXISTING ALGORITHM
The conventional DDL protection scheme used by rail transport systems relies on predetermined thresholds for current waveform parameters to detect faults on the line on time.These thresholds were established based on in-depth analyses of normal and start-up operational patterns observed throughout the system.The thresholds must be carefully calibrated to the unique electrical characteristics of the specific rail infrastructure.However, employing fixed thresholds may not enable the protection scheme to adapt to changes in operating conditions that naturally occur over time as train traffic patterns fluctuate.Without regular updates to the protection settings to reflect these inevitable changes in the system, the scheme risks failing to detect actual faults that breach the thresholds (if said thresholds are set too high) or triggering unnecessary false alarms (if the thresholds are established too low).
The conventional DDL protection scheme inherently lacks adaptability.It does not learn from new fault patterns or variations in system behavior over extended periods through automated means.Unless the protection settings are manually reconfigured by system operators, the conventional approach cannot independently calibrate its detection algorithms and thresholds to adjust to evolving conditions.This restriction limits the scheme's effectiveness in dynamic operating environments that naturally change over long periods as rail infrastructure usage fluctuates.In addition, the algorithm often struggled to accurately distinguish the temporary high inrush currents typically associated with train start-ups from legitimate fault conditions.Consequently, they incorrectly interpret these routine start-up currents as faults, needlessly tripping the rail system and unintentionally interrupting passenger and freight rail services.Without self-learning capabilities to teach the protection software to differentiate between normal operation and anomalies, the conventional approach was not optimally suited for complex rail networks with unpredictable traffic patterns that vary over time.
Figure 3 displays the different high starting currents of the trains in the DC traction system.Also, the following are the prerequisites for the trip protection: (1) As the step current di dt increases and I rise over the trip setting level.However, the duration ( T 1 ) is smaller than the setting duration ( T).As a result, no excursion is planned.
(2) If I is greater than the trip setting level and the time is greater than the trip setting duration ( T), tripping is commenced.
(3) In the course of the current rise, di / dt momentarily reduces to below the setting level.However, this duration is less than the time setting of protection return ( T re ), therefore, a trip decision is made.
(4) In the course of the current rises, di dt is reduced to below the setting level with a duration of more than T re , therefore protection will return.
(5) The waveform represents the existing DDL's restriction for constant di dt and ramp input with prolonged overcurrent.The technique may fail to respond appropriately to prolonged overcurrent faults that do not generate quick fluctuations in current waveform characteristics but instead create a constant higher-than-normal current draw below the short circuit threshold.Because the direct digital line protection 11000 VOLUME 12, 2024 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.technique is designed to detect sudden changes that indicate a fault, it runs the risk of not appropriately responding to a failure that causes a persistent but modest overcurrent.This scenario might result in the system remaining electrified despite a fault, perhaps causing damage over time or evolving into a more significant fault state.

IV. PROPOSED CURRENT DDL ALGORITHM
The detailed operation of the current Differential Drift Limiting (DDL) protection system as depicted in The first stage of the DDL protection system is the filtration of the incoming current signal.This step is essential to remove any noise or transient fluctuations that are not related to the actual current flow.Electrical noise can come from various sources such as electromagnetic interference, and can distort the true current signal.By filtering out this noise, the system ensures that the electrical current is accurate and reliable.Figure 5 shows a comparison of the source with noise and the filtered output before determining the threshold.

B. ESTABLISHMENT OF BASELINE THRESHOLDS
After the signal is filtered, the system sets baseline thresholds for current (I) and time (T).These thresholds are essential for identifying deviations from normal operational parameters, which may indicate potential electrical faults.They are determined by predefined settings specific to the electrical system under surveillance.

Phase 2: Dynamic Monitoring and Decision-Making:
The advanced protection scheme delineates the dynamic monitoring and decision-making processes integral to the system's operation, which is crucial for precise fault detection and system reliability.

C. MONITORING THE RATE OF CHANGE OF THE CURRENT
The dynamic monitoring phase is initiated by continuously calculating the derivative of the current with respect to time, which represents the rate of change of the current.By logging the instances of these calculations, the system can identify  rapid fluctuations in current, which are often precursors to fault conditions.This monitoring is vital because it provides a real-time assessment of the system's electrical behavior.

D. COMPARATIVE ANALYSIS
At this juncture, the system employs a critical decisionmaking checkpoint.This involves comparing the calculated di/dt against a pre-established maximum permissible value.If di/dt exceeds this threshold, it implies an abnormal change in the current, potentially indicating a fault scenario.This comparison triggers the system to engage in a more thorough diagnostic evaluation to confirm the nature of the discrepancy.

E. DUAL PARAMETER CHECK
To further enhance the decision accuracy, the system conducts dual parameter verification.This study examines two crucial factors: the magnitude of the current change ( I) and the time interval ( T) associated with this change.The system confirms a fault condition if I surpass the maximum current threshold (I max ) and if the corresponding T is within a predefined range ( T 1 ).This dual check is designed to filter out transient changes that are not indicative of a fault, thus preventing false alarms.Only substantial and swift changes in current, which are symptomatic of severe electrical faults such as short circuits or ground faults, are considered valid triggers for activating protection mechanisms.This phase underscores the system's sophistication in dynamically monitoring electrical parameters and making informed decisions to ensure accurate fault detection and reliable system protection.
Phase 3: Fault Assessment and Adaptive Thresholding: If the current drops below a minimum threshold (I min ) within a time greater than a reference time (T re ), it may indicate that a transient fault has occurred and cleared on its own, which does not require intervention from the system.

G. ADAPTIVE THRESHOLDING
In the absence of faults, the system employs adaptive thresholding, which adjusts the baseline thresholds based on the most recent operational data collected within a moving window.This enhances the system's sensitivity and reliability over time.
The algorithm conducts a multivariate analysis of the current parameters over time.By analyzing both the rate of change of current and absolute current deviations from normal operating ranges, a comprehensive understanding of the electrical system's behavior is gained.One of the key features of the algorithm is its adaptive thresholding mechanism.This DDL mechanism adjusts the thresholds for fault detection based on environmental variables and diverse data inputs.In the context of high-impedance faults, the algorithm dynamically adapts its thresholds to detect variations that may indicate such faults.To enhance decision accuracy, the system conducts dual parameter verification, which examines both the magnitude of the current change ( I) and the time interval ( T) associated with this change.This dual check is crucial for filtering out transient changes and ensuring that only substantial and swift changes in current, indicative of severe electrical faults such as highimpedance faults, are considered valid triggers for activating protection mechanisms.The algorithm includes a transient fault assessment phase.If the current drops below a minimum threshold (I min ) within a time greater than the reference time (T re ), it may indicate that a transient fault with high impedance has occurred and cleared on its own.The algorithm is designed to distinguish between transient and persistent fault conditions.The adaptive learning approach of the algorithm enables it to continuously learn from the system's normal operational patterns.This learning is used to dynamically adjust the threshold values in real time, allowing the algorithm to differentiate between normal inrush currents (such as those during start-up) and genuine fault conditions, including high-impedance faults.
Phase 4: Protection Trip Activation: Upon the detection and confirmation of a fault condition, the system's protection trip mechanism plays a pivotal role in safeguarding the electrical circuit.The Sequential checks include monitoring di/dt, conducting comparative analysis against set thresholds, and performing a dual parameter check for current magnitude and time interval to ensure that the system's response is both accurate and warranted.Once a fault is validated through these meticulous processes, the trip mechanism is actuated.This action is both decisive and targeted; it involves isolating the specific segment of the electrical circuit where the fault has occurred.The isolation is typically executed by circuit breakers or relays, which are instructed to open, thereby disconnecting the faulty section from the rest of the electrical network.The primary objective of the tripping mechanism is to maintain the overall integrity of the electrical system and prevent any further damage to the infrastructure.By swiftly removing the faulted segment from the network, the system minimizes the risk of the fault escalating into a more severe incident, such as an electrical fire or a widespread power outage.Moreover, isolation helps preserve the continuity of service in the unaffected parts of the network, ensuring that disruption is localized and does not impact the entire system.This is especially critical in complex electrical systems where reliability and uptime are of paramount importance.
Phase 5: System Reset or Maintenance Indication: Following a protective response, the system can reset itself to resume continuous monitoring and prepare for any new anomalies.Alternatively, if the nature of the detected event is severe or indicative of a persistent issue, the system may signal that maintenance actions are required.The DDL protection system is a comprehensive approach that combines real-time monitoring with intelligent decision-making to protect DC traction systems from electrical faults.It is designed to be both sensitive to actual fault conditions and to avoid unnecessary activation, ensuring the reliable operation of rail systems and the safety of its infrastructure.Adaptive thresholds are a cornerstone of modern electrical protection systems, offering a dynamic and tailored response mechanism to the unique and ever-changing demands of electrical infrastructure.These thresholds are not static; rather, they are fluid, evolving with the operational conditions of the system they protect.This adaptability is achieved using a moving window approach, which continuously samples the most recent electrical system data over a specified period.The analysis of these data allows for the dynamic adjustment 11002 VOLUME 12, 2024 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
of threshold values, ensuring that they remain in tune with the real-time conditions of the system.If the system detects an overcurrent condition that not only exceeds the adaptive threshold but also does so for a duration that extends beyond the moving window's time frame, it identifies this as a sustained overcurrent fault.It is at this juncture that the protection mechanism is brought into play.Typically, this takes the form of a circuit breaker, a device designed to intervene by isolating the affected section of the electrical network.Tripping the circuit breaker is a protective measure that effectively prevents the escalation of damage and ensures the continued safety of the system.
The intelligence of this system lies in its current differential detection and localization algorithm, which uses the moving window approach to continuously refine and update the adaptive threshold values.These values are recalculated in real time based on the live data flowing from the electrical system.The precision with which these thresholds are set allows the system to differentiate between a normal operational fluctuation and one indicative of a fault.This real-time recalibration is particularly beneficial for systems characterized by variable loads.In such environments, fixed thresholds can lead to two undesired outcomes: unnecessary tripping due to normal load variations, which can lead to unwarranted operational disruptions, or a failure to trip in the event of an actual fault, which can have serious safety implications.
Thus, adaptive thresholding strikes a delicate balance, ensuring responsiveness to faults without being overly sensitive to routine operational changes.Upon the occurrence of a protection trip, the system's response is determined by the nature of the fault.For transient faults that are temporary and clear without intervention, the system is designed to reset automatically, reinstating the protective monitoring without delay.However, for persistent faults, where the issue does not resolve independently, the system requires manual intervention.
This typically involves maintenance or repair actions to address and rectify the underlying cause of the overcurrent.
Only once these issues are resolved does the system reset, thereby ensuring that operations are restored only when it is safe to do so.
Therefore, the ingenuity of the adaptive thresholding mechanism lies in its capacity to provide a customized protection response, its agility in adjusting to the electrical system's operational profile, and its strategic differentiation between transient and persistent fault conditions.This ensures that the electrical system is not only protected against the current faults but is also robustly poised to pre-emptively safeguard against future incidents, thereby maintaining operational integrity and safety.

V. RESULTS AND DISCUSSION
The method suggested in section IV is simulated in the MATLAB/Simulink environment using the dynamic model of the TPSS to evaluate the performance of the proposed protection technique.The Malaysian rapid rail model is simulated with minor changes, and various scenarios are validated and explained.
Case 1: Addressing Short Circuits With Varied Impedances: The improved algorithm analyzes both the rate of current change and absolute current deviations from normal operating ranges.Through multivariate analysis of current parameters over time, it can reliably distinguish between normal operations and high-impedance fault scenarios.This comprehensive input evaluation feature ensures high impedance faults are accurately identified and isolated through timely tripping.
The simulation results were obtained for the model of what would happen during the starting duration of a train departing from the station on a track with both power rails providing power over a 2 km stretch.Fig. 6 displays the current levels measured at three points: the 0.1 km mark from substation A, the 1 km midpoint, and finally the full 2 km distance from the starting point.It also shows data collected from the substation when different fault scenarios were intentionally introduced into the system.Three impedance values were set to observe their impact on the output current 0.1 , 0.3 , and 0.5 , respectively.
At the 0.14 th second, a purely resistive short-circuit fault is introduced directly on the power rail irrespective of the current waveform.The goal is to closely monitor how current levels across the network respond to this type of issue.As the train first received power to begin moving, a short spike in current was evident as the electric motors rapidly increased their draw.This initial surge then quickly dropped back down to more steady levels.With a resistance of 0.1 in the red waveform, the maximum starting current peaked at approximately 1200 A. When the power rail fault was applied, the short-circuit current increased to approximately 1090 A. The blue waveform shows the results of a 0.3 resistance.Once again, the starting current hits a maximum of 900 A before stabilizing.When the same short-circuit fault is introduced at 0.14s, the current peaks at 950 A. Finally, the simulation ran with the impedance of 0.5 which is the green waveform, has the starting current peaked at 780 A, while the fault current climbed to around 1150 A.
The experiment confirmed that short-circuit and starting currents tended to be comparable in magnitude and how quickly they increased.Fig 6 .clearly illustrates Ohm's Law in action, with lower resistances consistently leading to higher current flows.As the measurement points got further away from the initial substation, both the current and oscillation rates exhibited a decreasing trend.
Case 2: Addressing Sustained Overcurrent Conditions: The improved algorithm incorporates continuous cumulative assessment capabilities.It monitors the duration and magnitude of any overcurrent condition about pre-defined thresholds.Even if the current rises slowly, the algorithm will trigger the protection.It will do so once the overcurrent persists over a given time.This process ensures that we catch  as a limitation of the conventional DDL algorithm.The fault scenario has been defined such that if the current value exceeds the value of 700 A for longer than 10 milliseconds, the overcurrent protection system must be activated.This threshold value is established by industry standards, and the safeguard should be triggered when the measured value exceeds 115% of the threshold.The maximum permissible current threshold in this scenario has been set at 805 amperes.Fig 7. illustrates that the conventional protection system does not activate under these conditions, which highlights an inherent disadvantage of the existing DDL technique.
The proposed DDL algorithm includes a provision ensure that the input current does not surpass either the maximum limit or the delay threshold.If this scenario occurs, the protection will be triggered.This improvement addresses the shortcomings of the DDL system by providing 11004 VOLUME 12, 2024 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.a more robust protection method that overcomes issues with sustained overcurrent relay protection, which can be seen in Fig 7. The improved protection system does not only focus on the spike in di dt values but also monitors the slowly developing faults over time which is under the threshold.By depicting the input, fault scenario definition, and industry standards, a more comprehensive understanding of the protection response is achieved.
Case 3: Addressing High Inrush Current During Start-Up: The improved DDL algorithm utilizes an adaptive learning approach to more accurately discern between normal and fault current profiles.This phenomenon can often be mistaken for a fault condition by conventional protection algorithms, leading to unnecessary tripping and system downtime.The improved DDL algorithm also distinguishes itself with an adaptive learning approach.This method enables the algorithm to differentiate between normal inrush currents, which are a natural part of start-up operations, and actual fault currents.As indicated in the hypothetical Fig. 8, the conventional DDL algorithm's threshold is represented by a fixed pink waveform.This rigidity means it cannot account for the high inrush currents seen as the blue waveform, potentially resulting in false tripping.
In contrast, the improved algorithm implements a dynamic thresholding mechanism.This advanced approach, depicted by the green waveform in Fig. 7, adjusts the threshold values in real-time in response to the observed current patterns, learning from the normal operational peaks and troughs due to varying demands.Consequently, the improved DDL algorithm can discern between customary start-up inrush currents and genuine fault conditions.The main advantage of adaptive thresholding in the improved DDL algorithm is its ability to maintain system stability during regular operations without compromising the system's reliability in actual fault situations.Doing so, greatly reduces the chances of unwarranted disruptions, thereby enhancing the overall efficiency and safety of the electrical system.
Case 4: Addressing Fluctuating Load Conditions: The rate of change is a critical parameter in fault detection algorithms as it can reveal rapid transitions indicative of fault conditions.The graph plots di dt against time is shown in Fig 8 .The initial spike observed corresponds to the inrush current typical of power system components energization or starting current of large motors.Following this transient, the di dt stabilizes but demonstrates regular perturbations, likely due to operational fluctuations within the system.These perturbations are within the operational threshold and did not trigger the fault detection mechanism.
Fig 9. also depicts the system's response to a fault which is induced at 0.14 seconds.The fault is simulated by setting a maximum current I max threshold at 780 A with the time the system must react to the fault T of 0.02 seconds.The adaptive DDL algorithm's efficacy is demonstrated by its rapid response to the imposed fault.Upon detecting a rise in current that exceeds the adaptive threshold, in other words, when the spike is detected, the algorithm instantaneously triggers the protective mechanism, thereby illustrating the system's fault detection capability.The red plot indicates the DDL's current output, which maintains normal operational levels until the fault occurrence, where it promptly drops to zero, signifying the activation of the protective protocol.
Case 5: Addressing Component Aging Effects: The advanced algorithm confronts the inevitable issue of component aging within power systems, which can significantly affect performance and reliability.To mitigate the impact of aging, the algorithm is equipped with a continuous self-calibration feature, which is pivotal for maintaining system integrity over time.This self-calibration process is not static; it's dynamic and data-driven, utilizing the influx of real-time operational data to refine and update the algorithm's internal models.As a result, the algorithm doesn't just respond to faults; it learns from the system's performance trends, adapting its protection settings proactively.One of the algorithm's standout features is its ongoing trend analysis capability.By consistently monitoring the system's performance, the algorithm can discern patterns indicating the progressive wear and tear on components.This capability allows for an anticipatory adjustment of protection settings to compensate for the gradual changes in the system's electrical characteristics due to aging.The adaptability of the algorithm ensures that decision-making remains precise, even as the physical properties of the system components evolve.By factoring in the degradation effects, the algorithm ensures consistent protection and minimizes the risk of fault misclassification that could lead to system failures or unnecessary outages.This adaptive capability of the algorithm is essential for the longevity and reliability of power systems, ensuring that protection mechanisms remain robust and accurate throughout the operational life of the infrastructure, despite the natural aging of its components.
The proposed algorithm includes a continuous selfcalibration feature to mitigate the impact of component aging within power systems.This dynamic and data-driven selfcalibration process allows the algorithm to adapt to changes in the system's electrical characteristics over time, ensuring sustainability and reliability.As a result, the proposed current DDL overcomes the disadvantage of sustained overcurrent protection, adds adaptive DDL, which is considerably superior to the fixed threshold idea in existing DDL and overcomes the other eventualities stated in this section.

VI. CONCLUSION A. OVERVIEW
The research elaborates on a comprehensive SIMULINK model that meticulously simulates a bilateral DC traction power system commonly used in railway networks.This model meticulously captures the nuances of an entire track segment, thus providing a robust framework for in-depth analysis.It integrates MATLAB programming to implement the current DDL algorithm, which is a sophisticated fault detection and classification technique tailored to the complexities of traction systems.In assessing the efficacy of the DDL algorithm, the model produces detailed feeder line current profiles, which are fundamental in detecting variations indicative of faults.These profiles are obtained at incremental distances from the substations, enabling a granular analysis of the current fluctuations due to train movements.The DDL algorithm utilizes these profiles to pinpoint and classify potential faults.
11006 VOLUME 12, 2024 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
This paper thoroughly examines conventional protection schemes, identifying their limitations, particularly during fault conditions.By juxtaposing these with the DDL protection waveforms, the research delineates the advanced capabilities of the DDL system in managing fault scenarios more adeptly.A critical aspect of the DDL algorithm's superiority lies in its adaptive thresholding mechanism.This mechanism intelligently adjusts the thresholds for fault detection based on environmental variables and diverse data inputs, thus curtailing the incidence of false alarms-a common shortfall in traditional systems.The research meticulously analyzes the performance of the DDL algorithm under stringent conditions, such as shortcircuit events.It assesses the algorithm's responsiveness and accuracy in adjusting to abrupt changes in current, ensuring that the protection mechanism is both dynamic and precise.An investigation into the DDL protection scheme's behavior during sustained overcurrent situations is also presented.This includes scenarios where the current rate of change remains consistent and within the algorithm's adaptive threshold limits, a test to verify the reliability of the DDL system against prolonged anomalies.
Unlike conventional protection algorithms that rely on fixed thresholds, the Adaptive DDL Algorithm excels in its adaptability.Through dynamic monitoring, continuous learning, and real-time adjustment of thresholds, it effectively distinguishes between normal operational variations and potential fault conditions.This adaptability mitigates the shortcomings of static algorithms that may lead to either missed fault detections or unwarranted system interruptions.The justification for its superiority lies in its ability to address specific challenges, such as high-impedance faults and sustained overcurrent conditions, with precision.The algorithm's comprehensive approach, combining dynamic thresholding, dual-parameter checks, fault assessment, and adaptive learning, positions it as a robust solution for the complexities of DC traction systems, outperforming traditional algorithms in terms of accuracy, reliability, and adaptability to evolving operational conditions.

B. THE NOVELTY OF THE PROPOSED ALGORITHM 1) ADAPTIVE THRESHOLDING
The introduction of an adaptive thresholding mechanism is a key feature of the proposed algorithm.This mechanism allows the algorithm to dynamically adjust thresholds based on environmental variables and diverse data inputs.This adaptability addresses the limitations of fixed thresholds in traditional systems.

2) MULTIFACETED ANALYSIS UNDER STRINGENT CONDITIONS
The algorithm is subjected to detailed analysis under stringent conditions, such as short-circuit events, sustained overcurrent situations, high inrush current during startup, fluctuating load conditions, and component aging effects.The multifaceted analysis demonstrates the algorithm's efficiency in fault detection and classification.

3) CONTINUOUS SELF-CALIBRATION FOR SUSTAINABILITY
The algorithm includes a continuous self-calibration feature to mitigate the impact of component aging within power systems.This dynamic and data-driven self-calibration process allows the algorithm to adapt to changes in the system's electrical characteristics over time, ensuring sustainability and reliability.

4) SIMULATION AND VALIDATION
The proposed algorithm is simulated in the MAT-LAB/Simulink environment using a dynamic model of the traction power supply system.Various scenarios are validated to assess the performance of the algorithm under different conditions, adding a practical dimension to the research.

5) ADDRESSING SPECIFIC CHALLENGES IN DC TRACTION SYSTEMS
The focus on addressing challenges specific to DC traction systems, such as high inrush currents during startup and sustained overcurrent conditions, demonstrates a targeted approach tailored to the needs of the Malaysian Rapid Rail DC Traction System.
The adaptive DDL algorithm, through these multifaceted analyses, demonstrates a high degree of efficiency in fault detection and classification, outperforming conventional methods significantly.The paper's outcomes reinforce the necessity for such advanced protection systems in the rapidly evolving railway infrastructure.Consequently, while the current findings are encouraging, the study acknowledges the need for further exploration and experimentation to tackle additional scenarios and refine the DDL algorithm, anticipating future expansions in rail network complexity and demand.

FIGURE 1 .
FIGURE 1.The schematic diagram of the entire DC traction electrification system.

FIGURE 2 .
FIGURE 2. The schematic diagram of the traction load and train motor.

FIGURE 3 .
FIGURE 3. Various waveforms of high starting current.
Fig 4 described in a multi-phase process: Phase 1: Filtration and Threshold Establishment: A. FILTRATION OF THE CURRENT SIGNAL

FIGURE 5 .
FIGURE 5.The simulation model of the input signal with noise and filtered signal.

FIGURE 6 .
FIGURE 6. Various output current characteristics of the DC traction system.

FIGURE 7 .
FIGURE 7. The response of conventional DDL and proposed DDL to the high starting current input of the DC traction system.

FIGURE 8 .
FIGURE 8.The conventional threshold vs adaptive threshold model for the input current.

FIGURE 9 .
FIGURE 9.The output response of the improved DDL algorithm when the fault is detected.

TABLE 1 .
DC railway traction system for study.
. The conventional current-based DDL algorithm continuously and precisely monitors the rate of change of current di dt across the electrical distribution line.Analyzing di dt provides valuable insight for identifying potential fault

TABLE 2 .
DDL parameters used in the algorithm.