Research On Short-term Prediction Of Photovoltaic Power Generation Based On Improved BP Neural Network Model

Under the background of rapid economic and social development in China,environmental crisis and energy shortage have become increasingly prominent.The concept of vigorously developing and using clean and renewable energy is unstoppable.It also complies with the social theme of environmental protection and sustainable development.The grid connection of ne'w energy generation represented by photovoltaic grid connection has become an important part of the stable operation and optimal dispatch of China's power system.However,in the actual operation process,due to the environment and climate and other unstable factors,the output of photovoltaic power generation will produce huge randomness and volatility,so it is necessary to study the short-term prediction of photovoltaic power generation for photovoltaic grid connection and grid operation scheduling.Firstly,this paper studies the basic working principle of photovoltaic power generation and the volt ampere characteristics of photovoltaic modules,then introduces the structure of photovoltaic power generation system and the advantages and disadvantages of centralized,master-slave,distributed and series connected photovoltaic power generation modes,and analyzes various environmental factors affecting photovoltaic power generation in detail.On this basis,the historical missing and abnormal data of photovoltaic power generation are identified and corrected.The main factor analysis method is used in SPSS data analysis software to reduce the dimension and simplify the processing of historical complete and effective meteorological data,and the main factor simplified prediction model is selected to input.Considering the randomness of selecting initial weight threshold and the shortage of easily falling into the optimal solution of actual BP neural network prediction algorithm,A beetle antennae search algorithm majorization(MBAS algorithm)optimizes the weight and threshold of BP neural network is adopted,and the principle and operation process of the new improved algorithm are introduced in detail.The optimization results and running time of three algorithms(BAS,PSO and MBAS)are analyzed and compared by the function example,and the results show that the accuracy and convergence speed of the algorithm are improved;the initial parameters of BP neural network are optimized by the MBAS algorithm,and the main factor selected by the main factor analysis method(PFA)is used to simplify the model input,so a new algorithm based on PFA-MBAS-BP neural network is built Short term prediction model of photovoltaic power generation.Based on the complete and effective historical photovoltaic data,through comparative analysis of the prediction accuracy and convergence speed of the four combined prediction simulation models BP?PFA-BP?MBAS-BP and PFA-MBAS-BP,the experimental results show that:The PFA-MBAS-BP neural network improved prediction model is less affected by weather changes,especially under the sudden weather types such as rain and overcast,the prediction accuracy of the improved model is also more than 92.5%,and the overall prediction data The fitting degree is also very high,which achieves the desired effect and has a good application prospect.