Research on the Operational Strategies and Benefit Allocation Mechanisms of Virtual Power Plants for Multi-Timescale Peak Regulation Services Within the Electricity-Carbon Synergy Market Environment

This paper employs a quadratic function to articulate the relationship between incentives and user output, utilizes Long Short-Term Memory (LSTM) networks for predicting deterministic parameters, and establishes the probability density function of uncertain parameters through Mixture Density Networks (MDN). This approach forms the basis for multi-timescale peak shaving optimization scheduling in virtual power plants. The model encompasses wind turbines, energy storage systems, and adjustable loads, while also addressing carbon emission costs. These costs include nationally mandated emissions, emissions trading between virtual power plants, and taxes on excess emissions resulting from prediction inaccuracies. The implementation of multi-timescale optimization scheduling enhances prediction accuracy, minimizes penalties and energy waste, and boosts the profitability of virtual power plants. Additionally, this paper introduces an internal benefit allocation mechanism tailored for virtual power plants, which assesses rewards and penalties based on output, risk, forecast deviations, and carbon emissions. This mechanism promotes fairness and ensures the delivery of high-quality peak shaving services.