Two-Level Optimal Bidding Strategy for Load Aggregator Based on a Data-Driven Approach Combined with LSTM-Based Forecasting and Agent-Based Models

Demand response (DR) is an economical way of addressing the challenges faced by massive penetration distributed energy resources, such as renewable energy. Residential consumers account for a significant proportion electricity consumption. However, their behavior highly random and uncertain, meaning it difficult to quantify impact DR programs in which they participate. This paper presents two-level optimal bidding strategy framework load aggregators that combines data-driven forecasting model agent-based (D-ABM) provide realistic estimate DR. First, aggregated all market prices are predicted via long short-term memory (LSTM) autoencoder model. Then, proposed D-ABM estimates quantifies difference rate terms total due Since bottom-up approach, each consumer can be treated heterogeneous agent changes individual usage patterns estimated. Changes collective consumption also quantified considering estimated interactions defined basic rules. In addition, assumptions about biases preferences explain irrationality decision-making given agents, uncertainty participation considered more realistically. Finally, based on these uncertainties addressed at level, various strategies obtained. The numerical simulation results indicate our provides estimation under DR, minimizes any deviations from strategies, ensures maximum profits aggregators.