Photovoltaic Power Prediction Based on Nonlinear Adaptive Weight Combination Model Considering Error Correction
To improve accuracy of photovoltaic power prediction in photovoltaic power plants and solve shortcomings of two traditional single models of extreme gradient boosting model and long short term memory model as well as traditional combined model of extreme gradient boosting-long short term memory model,such as delayed results of photovoltaic power prediction,easy to change the prediction effect,large prediction error,and poor linear fitting,on the basis of extreme gradient boosting algorithm,long short term memory algorithm,and linear adaptive weight,a nonlinear adaptive weight extreme gradient boosting-long short term memory model considering error correction was proposed for photovoltaic power prediction.Preliminary predicted values and real values of the two single models which were trained by using extreme gradient boosting algorithm and long short term memory algorithm respectively were composed into a new training data set.Initial predicted values of the two single models were assigned adaptive weight coefficients through the proposed model trained by using neural network algorithm.Moreover,the distribution of prediction errors was calculated piecewise according to the predicted value of the proposed model in the training,and the mean of errors was accumulated according to the predicted value of the proposed model on the basis of the predicted result for error correction,so as to further improve prediction accuracy of the proposed model.Photovoltaic power prediction performances of the proposed model,traditional combined model and two traditional single models on sunny,cloudy,and rainy days were simulated by using Python language.The results show that compared with extreme gradient boosting-long short term memory model,extreme gradient boosting model,and long short term memory model,the square root mean error of the proposed model is reduced by 28.57% ,39.39% and 49.79% ,the mean absolute error is reduced by 44.25% ,53.33% and 64.8% ,and the coef-ficient of determination is increased by 1.43% ,2.38% and 3.34% ,respectively.The proposed model is more effective in reducing the photovoltaic power prediction error of the traditional single models and optimizing the weight coefficients of the traditional combined model.Under the three weather conditions,the photovoltaic power prediction error of the proposed model is relatively minimum and the robustness is the strongest,which verifies the effectiveness of the proposed model.
photovoltaicpower predictionadaptive weighterror correctionextreme gradient boosting algorithmlong short term memory algorithm
NETL
NSTL
万方数据
