A Comparative Analysis for Short and Long-Term Time Series for Energy Consumption Predictions

A substantial portion of the Sustainable Development Goals rely on energy. Ensuring the availability of reliable, contemporary, and inexpensive energy to humanity is the specified goal of SDG Goal 7. The rapid expansion of high-density cities as accompanied by commercial and industrial development leads to an increase in consumption of energy, especially in the form of electricity. So, concerning reducing the energy footprint energy prediction is an important concern to enhance energy management strategies for buildings. With the goal of fulfilling future energy requirements, this paper provides an extensive literature evaluation of forecasting techniques utilized by researchers for usage of energy in smart buildings. Research continues to be done on various forecasting techniques for both residential and non-residential buildings. Based on the dataset, various load categories, accuracy of prediction, and evaluation criteria used, the research literature is further examined. Some of the challenges like load fluctuation, missing data, occupant electricity usage behaviour are discussed. To reduce the negative impact of high energy consumption, there is a need to implement energy-efficient technologies and practices in buildings and industries. While the implementation of smart meters has set up opportunities with the objective to anticipate power through sensors and energy monitoring, machine learning approaches are currently unable to analyze data with respect to their chronological dimensionality.As a consequence, time constraints are considered and recurrent neural networks, sequence-to-sequence deep learning algorithms, long short-term memory, convolutional neural networks, and hybrid machine learning approaches are implemented to develop forecasting load proximity. The techniques will be used to handle missing data like mean value imputation, Last observation carried forward, etc. to handle missing data as otherwise, it would lead to biased estimation. This paper provides an overview of energy consumption forecasting within smart buildings aimed at enhancing energy efficiency. It examines current research directions employing various forecasting algorithms and introduces a novel methodology for predicting energy consumption.