A Data-Driven Trading Strategy for Wind–Solar–Battery Microgrids in the Spot Electricity Market
针对风光储微电网出力间歇性和现货市场波动，提出基于机器学习的优化框架，利用GMM聚类和LSTM预测，通过MILP模型最大化微电网利润，并用广东现货市场数据验证。
The integration of renewable energy sources, such as wind and solar power, into microgrids presents challenges owing to their intermittent output and the volatility of spot markets. This paper proposes a multi-energy complementary operation mode for wind-solar-battery microgrids, enabling engagement in day-ahead and ancillary service markets to optimize trading amid output uncertainties. We develop a machine learning-based optimization framework: Gaussian Mixture Models (GMM) with K-means clustering produce representative seasonal scenarios for wind and solar generation to capture stochastic variations; Long Short-Term Memory (LSTM) networks forecast electricity prices to model temporal dynamics. These inputs inform a Mixed Integer Linear Programming (MILP) model that maximizes microgrid profitability while ensuring a series of constraints on microgrid operations and electricity market transactions. An empirical evaluation is conducted using real operational data, meteorological records, and market prices from the Guangdong spot electricity market in China. We employ a comprehensive validation framework including out-of-sample evaluation of the LSTM price forecaster, clustering-based scenario generation preserving real-world renewable variability, and systematic sensitivity analyses across prediction errors among 5% to 10% and imbalance price coefficients among 1.05 to 1.35. The main findings reveal: (1) The data-driven operation mode provides feasible and reliable supply by coordinating energy storage and ancillary market procurement, supported by the LSTM model that achieves high prediction accuracy with a mean absolute error of 0.0093 on a 20% holdout test set; (2) The microgrid exhibits pronounced seasonal characteristics, requiring differ entiated trading strategies across seasons; (3) The imbalance price in the ancillary service market critically shapes microgrid profitability, with profits peaking around an imbalance price coefficient of 1.2 and displaying an inverted U-shaped response.