Research On Performance Prediction Of Solar PV/T Collector System Based On Artificial Neural Network Model

Since the solar PV/T system is highly affected by environmental factors,it has obvious intermittent,random and fluctuating characteristics when generating thermal energy and electric energy,so that the electric energy output of the PV/T system may be in the process of grid connection.It will have an impact on the power grid,and the heat energy generated may not be used in time,resulting in waste of heat energy.In order to ensure the safe operation of the power grid and improve the utilization rate of thermal energy,it is necessary to accurately predict the thermal and power output of solar PV/T in advance,so as to coordinate with the peak regulation and distribution of energy and improve the operating performance of the solar PV/T system.The thermoelectric output of a solar PV/T system involves a variety of environmental factors,and thermal energy and electrical energy affect each other.The traditional numerical simulation is a complex process for modeling a solar PV/T system,and the simulation prediction accuracy is low.Artificial neural network has excellent performance in dealing with multi-dimensional nonlinear function relationship under the premise of a large amount of data.Starting from improving the prediction accuracy of thermal power output of solar PV/T system,this paper compares the working principles and application fields of various artificial neural networks.After that,two neural networks,BP and RBF,were selected to predict the electrical power,electrical efficiency,water storage tank temperature and thermal efficiency of the solar PV/T system,and the data acquisition equipment of the cloud platform was used to obtain the data of the heat pipe solar PV/T system experimental station,including various parameters.A total of more than 20,000 sets of data were used to train and predict the two models under similar weather conditions.The prediction results show that the R values,which show the fitting degree of predicted value and the real value,are both above 0.98,which verify the high precision of the artificial neural network in the prediction of solar PV/T system thermoelectric output.The prediction accuracy of the RBF neural network model is 25 % – 30 % higher than that of the BP neural network model in the above four thermoelectric output indicators.Although the prediction performance of BP and RBF neural network models is outstanding,they all have their own shortcomings: BP neural network is easy to fall into a local minimum,RBF depends on the selection of the center point of the hidden layer function,etc.These shortcomings are easy to cause network fluctuations and affect prediction result.In order to further improve the prediction accuracy and network stability of the two models and make up for the structural shortcomings of the models,this paper uses genetic algorithm and particle swarm algorithm to optimize the two models.The prediction accuracy is improved by 46% and 38%,the stability of the model is improved by 64% and 30%.Finally,by comparing the prediction results of the above four models,the BP neural network model optimized by genetic algorithm has the best prediction effect.They are 0.74%,0.97%,0.47% and 0.32%,respectively,and the average fitting degree of the model is as high as 0.998,which accurately predicts the thermal and electrical energy output of solar PV/T system.