Short-term PV Output Power Prediction Based On PLSR-DBN Neural Network

In the environment to promote the development of solar cells in the country,the solar cell is becoming more and more normal in the grid system,making the electricity grid electricity quality has been improved and the operating economy has been improved.However,there is some uncertainty about the production of photovoltaic(PV)electricity generation,and the influence of external meteorological factors on the prediction of photovoltaic(PV)effect is very deep,which has a high risk of deterioration in order to affect the operating level of the electricity grid.Given this problem,it is of great importance to make a reasonable power forecast for solar cells to be networked.The main work and innovation of this paper are as follows:(1)The characteristics of photovoltaic(PV)power were quantitatively analyzed,and the main influencing factors of photovoltaic(PV)power were determined by curve fitting of meteorological influencing factors and historical data of output power.In order to further explore the characteristics of the influencing factors,principal component analysis was used to reduce the dimensions of the multi-dimensional meteorological influencing factors affecting photovoltaic(PV)power,and it was verified that there were multiple correlations among the single influencing factors.(2)The neural network algorithm applied to photovoltaic(PV)power prediction is studied,among which DBN(Deep Belief Networks)is favored by many researchers in photovoltaic(PV)prediction due to its multi-layer feature extraction function and short training time.However,in the process of DBN supervised tuning,BP algorithm(Back Propagation)is easy to get caught in case of local minimum and has the problem of slow convergence speed,leading to low tuning accuracy.To solve this problem,a partial least square regression method(PLSR)was proposed to replace BP for tuning,and a short-term prediction model of PLSR-DBN was established.(3)With artificial classification of weather will affect the input of the model and prediction precision,to solve this problem,used a correction based on membership degree of fuzzy c-means clustering algorithm(FCM)cluster analysis was performed on the photovoltaic(PV)power plant history data,the application of the clustering results on three types of weather,and based on FCM-PLSR-DBN model to estimate the three kinds of weather types respectively.At the same time,the simulation results are analyzed,and the results show that the relative error of FCM-PLSR-DBN is lower than that of FCM-BP and FCM-DBN,which improves the accuracy of photovoltaic(PV)power prediction.