PV Prediction Model Based On Covariance Matrix Adaptive Evolution Strategyand Eroor Correction

Photovoltaic power generation has significant energy,environmental and economic benefits,and is one of the highest quality green energy sources.Photovoltaic power generation can help the country achieve its dual carbon goals from various aspects.In the PV power prediction needs to take into account the impact of changes in sunshine hours,weather changes,temperature,wind speed and other factors on the output power of PV systems.In-depth investigation of the multiple factors affecting the output power of PV systems and timely and accurate power prediction is made,for which accurate prediction is beneficial to optimise the operation of PV power systems and reduce the pressure on the power grid.In this paper,the PV data are pre-processed to identify anomalous data using the Lajda criterion(3-sigma),and then the anomalous data are repaired based on the K Nearest Neighbors(KNN)algorithm.The Pearson correlation coefficient was used to qualitatively analyse the features,and the irradiance was found to be the most important factor affecting the PV power,and then Principal Component Analysis(PCA)was used to analyse the factors affecting the PV power based on the principle of feature transformation.The analysis was then based on the principle of feature transformation,and PCA was used to reduce the dimensionality of the factors affecting PV power,to extract the most complete sample information and normalise the data.To address the unsteady fluctuation of PV power output,an Improved Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN)was used to decompose the PV power into A hybrid prediction model based on a Convolutional Neural Network(CNN)and a Gated Recurrent Unit(GRU)neural network optimized by an adaptive evolutionary strategy of the covariance matrix is proposed.To further improve the prediction accuracy,a prediction model combining Improved Particle Swarm Optimization(IPSO)and temporal convolutional networks(TCN)is introduced to predict IPSO-TCN with error correction for the predicted values.By combining the initial prediction model and the error correction model,a PV prediction model based on the adaptive evolutionary strategy of covariance matrix and error correction is constructed to perform prediction together with other prediction models.Simulation data show that the proposed method in this paper has significant accuracy improvement for PV power prediction.