Optimal Configuration Planning of Multi-Energy Systems using Optimization-based Deep Learning Technique

Abstract

There has been discovered that the freestanding hybrid renewable resource, which combines energy sources like solar, wind, etc., is an efficient substitute for supplying electricity to isolated places that are not connected to utility grids. The appropriate supply of electricity is subject to several restrictions, including variations in peak load, power outage supply, and other factors like swings. This study suggests an effective strategy for managing and sizing hybrid renewable energy resources to get around the drawbacks. First, the prediction of weather and loading demand is carried out using Bi-Directional Gated Recurrent Unit (Bi-GRU) network, in which past data are utilized to predict the uncertainties in load. This is done to assist the right sizing of resources for that purpose, a Long Short-Term Memory (LSTM) is used. Furthermore, the Search and Rescue Optimization Algorithm (SRO) employed for optimizing the parameter of power management. Utilizing MATLAB R2020a, the proposed model is tested by simulating it and comparing its performance to that of other models using metrics like variation rate, battery energy, Loss of Power Supply Probability (LPSP), forecasting error, net present cost, and cost of energy.