小型水电站发电量预测模型研究

Study on Forecasting Model of Power Generation of Small Hydropower Stations

针对当前小水电发电量预测中资料短缺、发电不稳定等问题,采用LM-BP神经网络法和多元线性回归法进行了对比分析。神经网络法通过相关分析和自相关分析筛选出与日发电量有显著关系的因子作为模型输入,以小水电日发电量作为输出,构建LM-BP神经网络模型,通过试错法确定神经网络最优隐含层节点数并进行模型训练,该方法能够在不同预见期下取得较高的预测精度。多元线性回归法通过检验日发电量与待选因子的相关系数,筛选与日发电量有显著关系的因子作为预测模型方程自变量,利用最小二乘法计算模型参数。多元线性回归方法在预见期为1 d时预测精度与LM-BP神经网络模型接近,但在更长预见期的滚动预报中精度低于神经网络法。

Aiming at the problems of data shortage and unstable power generation in the forecasting of power generation of small hydropower stations,the forecasting methods based on LM-BP neural network and multiple linear regression respectively are compared.Based on correlation analysis and autocorrelation analysis,the factors significantly related to the daily power generation are selected as the model input,and the daily power generation of small hydropower station is used as the output to construct the LM-BP neural network model.The optimal number of hidden layer nodes of the neural network is determined by trial and error method and the model is trained.The LM-BP neural network model can achieve high prediction accuracy under different forecast periods.The multiple linear regression is used to test the correlation coefficient between the daily power generation and the factors to be selected and the factors with significant relationship with daily power generation are selected as independent variables of the prediction model equation,and then the least square method is used to calculate the model parameters.The prediction accuracy of multiple linear regression model is close to that of the LM-BP neural network model when the forecast horizon is 1 day,but the accuracy in the rolling forecast with longer forecast horizon is lower than that of the neural network model.