Research And Implementation Of Photovoltaic Power Generation Forecast Algorithm Based On BP Neural Network

With the rapid development of all walks of life,human demand for electricity continues to grow,and the problem of energy shortage is becoming increasingly apparent.In order to solve the problem of energy shortage,people have discovered many clean and renewable new energy sources,such as solar energy,geothermal energy,nuclear energy,wind energy,ocean energy and biomass energy,and applied these energy sources to various industrial fields.Solar energy is an inexhaustible energy source that will not have a negative impact on the earth’s ecological environment.At present,solar energy has attracted much attention among all new energy sources and has become more and more widely used.However,photovoltaic power generation has problems such as intermittency and volatility,low photovoltaic conversion rate and poor controllability.For local users of large photovoltaic power plants,their consumption capacity is limited,and excess electric energy cannot be stored in a large amount,which will cause energy loss.Therefore,it is of great significance to accurately predict the photovoltaic power generation.This thesis uses the back-propagation（BP）neural network commonly used in photovoltaic power generation forecasting as the base.The BP neural network has the advantages of parallelism,nonlinearity,fault tolerance and autonomous learning,but the BP neural network also has some shortcomings.Improper values of initial weights and thresholds lead to the local optimal solution.In order to solve this problem,this thesis proposes improvements to the artificial fish swarm algorithm（AF）and the whale algorithm（WOA）.In order to select the best optimization algorithm to combine with BP neural network,this thesis combines particle swarm optimization BP neural network（PSO-BPNN）,artificial fish swarm optimization BP neural network（AF-BPNN）,surrounded artificial fish swarm optimization BP neural network（SAF-BPNN）,whale-optimized BP neural network（WOA-BPNN）,and selection of whale-optimized BP neural network（SWOA-BPNN）five algorithms are compared,by comparing the mean square error and fitness between the simulation results and the actual values Convergence curve and prediction results are used to judge the pros and cons of the algorithm.Finally,it is concluded that the effect of SWOA-BPNN is better,so this thesis chooses to use SWOA-BPNN to predict photovoltaic power generation.In order to make the algorithm applicable in practice,this thesis uses field programmable gate array（FPGA）to realize the first forward propagation of BP neural network.The weights and thresholds of the trained BP neural network are exported,and the FPGA is used to construct a forward propagation of the BP neural network according to the derived weights and thresholds.error requirements.