基于博弈论赋权的光伏功率组合预测模型

The Combined Model for PV Power Forecasting based on Combination Weighting Approach of Game Theory

针对单一预测方法的局限性,本文采用物理方法和统计方法相结合的预测方法,建立光伏发电功率组合预测模型,并在预测模型中考虑电站的限电及检修计划。采用理论功率法、基于改进相似日的BP神经网络法、基于改进相似日的支持向量机对预测日的光伏功率分别进行预测,通过提出的博弈论组合赋权法来计算各模型的权重。通过某光伏电站的实际数据验证,计算分析了预测误差,结果表明,通过组合赋权法得到的权重克服了单一权重的片面性,使得综合评价更合理、科学,在限电情况下仍具有较高的预测精度,对光伏发电系统的功率预测具有一定的学术价值和工程实用价值。

Aiming at the limitations of single forecasting method, a combined method was proposed which considers the power rationing plan, based on physical methods and statistical methods. The theoretical power method, BP network method based on improved similar days and support vector machine method based on improved similar days are used to predict PV power for predicting daily, and uses the combination weighting approach of game theory method to calculate the weight of single method. The method is validated by photovoltaic system data and the forecast error is calculated and analyzed. The results show the combination weighting method, compared with a single weight that is unilateral, are more reasonable and scientific, and the method has high accuracy even in the station of the power rationing plan, which has good academic value and practical value to forecast power for PV generation system.