基于最小最大概率回归方法的中长期电价预测模型

The mid-term price forecasting model in electricity market using MPMR

中长期电价的预测无论是对于市场监管政策的制定,还是对于大用户和发电商的投资规划,都具有极其重要的意义。影响中长期电价的因素比较复杂,历史电价数据分布混乱增加了一般回归电价预测建模的难度。提出了一种基于最小最大概率回归方法的电力市场中期电价预测的新模型。在分析最小最大概率机(MPM)及其用于回归原理的基础上,使用最小最大概率回归(MPMR)方法对不同的训练样本集进行训练,并计算出预测期的预测值,取得了比较好的预测结果。训练样本的分割使中期电价预测模型更加准确。美国加州现货电能量市场的实例数据验证了所建模型及方法的有效性。

The mid-term price forecasting is important not only for ISO to make policies but also for generators to make investment programming. The complicated influence factors and the anomaly distributing of the price make the forecasting more difficult, and based on the minimax probability machine and the regression, a MPMR model for forecasting market clearing prince in spot marker was presented. After the training of regress MPM with the obtained sample set, MPMR forecast model is built whose forecasting results are more effective. The parameter of Kernel and the value of ~ which would influence the forecasting model＇s performance are decided by across testing. The forecasting model would be more effective if the obtained sample set is correctly divided up. Finally, real-word data of spot market in California is employed to demonstrate the validity of the proposed approach.