基于自适应动态规划的系统边际电价预测

System Marginal Price Forecasting Based on Adaptive Dynamic Programming

针对目前电价预测算法的局限性,提出一种基于自适应动态规划方法的自学习、自适应智能算法。按照Bellman最优化基本原理,使用Agent逐步与环境的交互作用来寻求预测电价和实际电价的误差最小值,得到系统边际电价的最优解。采用美国加州电力市场的数据进行电价预测仿真。与常规方法相比,该方法的拟合精度和平均绝对百分误差均有很大提高。

Aimed at disadvantages of the normal electricity price forecasting algorithm, this paper proposes which a method based on Adaptive Dynamic Programming（ADP） that has self study and self adaptive ability. Based on the Bellman optimization theory, that is gradually interaction with the environment to seek error minimum of the actual price and the forecast price with the Agent. It can obtain the optimal solution of System Marginal Price（SMP）. It uses the California electricity market price data as forecasting simulation. Compared with the conventional method, results show that fitting accuracy and MAPE are greatly improved.