Research On Power Prediction For Photovoltaic Micro-grid Based On Data

Solar energy is abundant and readily available,so more and more photovoltaic power stations are established and applied to power systems in the form of micro-grid.However,PV power is affected by environmental factors,and the output has great instability and discontinuity,which brings many problems to grid planning,system scheduling,and reliable and stable operation of the power grid.Therefore,it is of great significance to predict the power of PV micro-grid power.This paper research on power prediction for photovoltaic Micro-grid based on data.The main work is as follows:(1)PV data may be lost because the fault of sensor and process of transmission.In order to further obtain its use value from a large amount of data,this paper combines the characteristics of PV data,and proposes a method for filling the missing value of the micro-grid PV system,namely the Markov Monte Carlo imputation algorithm.The simulation results show the effectiveness of this method.(2)As PV power is very different under different weather conditions,this paper analyzes the influencing factors affecting the PV system power generation,studies the influence of various meteorological factors on PV power,the mutual influence between them,and the different regularities of PV generation under different weather types.After broadly classifying weather types,a grey correlation analysis data mining method was proposed to select similar days.And then,synthesize numerous indexes through principal component analysis,in order to eliminate information overlapping of the sample and reduce the input dimension of dataset.Data feature extraction of this part lays the foundation for prediction of photovoltaic Micri-grid.(3)This paper introduces a least squares support vector machine to build a predictive model,and uses EPSO algorithm to optimize parameters of LS-SVM.As can be seen from the prediction results,the EPSO-LSSVM prediction model shows higher prediction accuracy and adaptability in different weather conditions than the traditional prediction model.It has high application value to PV micro-grid system.At last,this paper research on the proposed method based on dSPACE real-time simulation platform.And the results verify the validity of this method.