Research On Photovoltaic Power Generation And Load Forecasting Based On Improved Pso-lssvm

The continuous access of distributed photovoltaic power generation and the continuous increase of electricity load have brought new challenges to power grid construction planning and power supply planning.In order to reasonably plan photovoltaic power stations and provide data support for distribution network regulation,it is necessary to make accurate short-term power prediction for the current power grid.The randomness and intermittence of distributed photovoltaic power generation and electricity load put forward higher requirements on the prediction accuracy of photovoltaic power generation and electricity load.In this paper,the short-term power forecast is studied from the perspectives of photovoltaic power generation and power load.Least Squares Support Vector Machine（LSSVM）was used for modeling.A dynamic self-adaptive Particle Swarm Optimization（PSO）algorithm was used to search the key parameters of LSSVM,and a prediction model was established based on the historical data of power grid.Finally,the feasibility of the model was verified by an example.The specific work is as follows:Aiming at the problems of errors and anomalies in the collection of historical data of photovoltaic power generation and short-term load,this paper adopts data identification and correction method and introduces k-means clustering method to conduct quantitative screening of historical data.Since regularization parameter（C）and kernel parameter（σ）are the key to determining the accuracy of LSSVM model based on radial basis function（RBF）,a dynamic adaptive PSO optimization algorithm is proposed in this paper,which dynamically adjusts the inertia factorω,acceleration factor c₁and c₂,and introduces crossover mutation operation to expand the population,and performs iterative optimization on parameter（C,σ）.Compared with the traditional model（LSSVM and PSO-LSSVM）,the effectiveness and superiority of the algorithm are verified.In terms of pv power prediction,the characteristics of PV power and the related factors affecting pv power are firstly studied.Secondly,in order to improve the prediction accuracy,this paper uses k-means algorithm and data pre-processing method to process and classify the historical data and obtain the corresponding input sample set.Finally,the pv power of a certain area in Danyang is simulated and predicted under different weather types.The experimental results show that compared with the traditional model（LSSVM and PSO-LSSVM）,the improved PSO-LSSVM prediction model has better prediction results and higher accuracy.The results show that MAE,MAPE,RMSE and r RMSE are 6.22k W,6.83%,7.11 k W and 6.56%respectively in sunny weather.In cloudy weather,MAE,MAPE,RMSE and r RMSE are 7.25 k W,9.45%,8.33 k W and 9.32%,respectively.In rainy weather,MAE,MAPE,RMSE and r RMSE are 8.74 k W,15.42%,9.82 k W and 13.32%,respectively.In the short-term forecasting of power load,firstly,the factors affecting power load are deeply analyzed and the variation characteristics of power load are studied.Secondly,the historical data were divided into working days and non-working days by the similar day screening process.Finally,based on the improved PSO-LSSVM prediction model,short-term power load prediction of working days and non-working days in a certain area of Danyang is carried out.Experimental results show that the accuracy of short-term power load prediction of the proposed model is higher than that of the other two models（LSSVM and PSO-LSSVM）.The results are as follows:Based on the improved PSO-LSSVM model,the MAE,MAPE,RMSE and r RMSE of power load prediction on working days are 9.75k W,11.23%,10.35k W and 8.75%,respectively.MAE,MAPE,RMSE and r RMSE are 10.58 k W,9.87%,9.88 k W and 11.27%,respectively.In conclusion,the LSSVM model constructed based on the dynamic optimization PSO algorithm is effective and feasible for short-term prediction of photovoltaic power generation and electricity load.This study can not only provide effective support for the power distribution plan and grid grid work of Danyang City,Jiangsu Province,but also provide reference for short-term power forecast in other areas.