Enhancing grid versatility: Role of thermostatically controlled loads in virtual energy storage systems

Abstract

Increasing electricity demand leads to a rise in power generation, particularly during peak hours, making it challenging to balance supply and demand. This often results in a mismatch between electricity production and consumption. Renewable energy sources (RES) generate electricity, but their intermittent nature limits effectiveness during peak demand periods. Excess energy from RES is stored in energy storage systems (ESS) that can be used during high-demand times, yet their capacity is limited. Large-scale storage is necessitated because of growing demand, which results in high capital costs and lowered lifespan, increasing the complications in bringing equilibrium between power generation and distribution. Demand response management (DRM) is a viable alternative method, which shifts the non-essential electrical loads to off-peak periods to reduce demand. Air Conditioners (ACs) and refrigerators are under thermostatically controlled loads (TCLs) to form a virtual energy storage system (VESS) and serve the purpose of DRM. VESS, along with DRM, optimizes energy usage instead of increasing supply. Although methods like linear regression, support vector machines, and neural networks have been explored for energy prediction, they often struggle with the complexities of TCLs behavior, especially under dynamic grid conditions. This study proposes a novel gradient-boosted neural networks (GBNN) model designed to more accurately predict TCLs self-discharge capacity, offering a more reliable solution for energy optimization and grid management. The proposed GBNN model achieves higher accuracy, as evidenced by a lower root mean square error (RMSE) of $4.34\times 10^{-5}$ and a higher coefficient of performance ($R^2$) of 0.9966.