Research And Application Of Multi-objective Optimal Combination Model In Power Load And Generation Forecasting

In recent years,with the rapid development of computer technology,a variety of forecasting models based on neural network have been established and rapidly applied in time series forecasting.With the outstanding advantages of neural network model in solving complex nonlinear regression,we can get high precision forecasting results in nonlinear time series forecasting.However,the accuracy of neural network model applied in different data sets is quite different,so it is not universal.Therefore,at present,many scholars use the combination forecasting model theory combined with intelligent optimization algorithm to integrate the forecasting information provided by different models through the combination weight,so as to improve the generality and accuracy of the forecasting model as much as possible.In view of this,this thesis establishes a combination forecasting model based on swarm intelligence optimization algorithm to improve the generality of forecasting,combines multiple neural network forecasting information by using combination weight,and makes full use of the advantages of each model,so as to improve the forecasting accuracy of time series.First of all,this thesis preprocesses the predicted power load demand and the original time series of each energy generation to reduce the useless information in the data and the noise interference in the data.Secondly,a combined prediction model is designed based on the neural network forecasting model,which can be used in power load demand and power generation forecast by using the advantages of strong function approximation ability,self-learning ability,nonlinear adaptability ability and strong robustness and fault tolerance brought by fast optimization calculation ability.Finally,the modified multi-objective dragonfly algorithm is used to optimize the combination weight of the combination model.DM test and forecasting validity were used to verify the model.The results show that the forecasting error of the combined forecasting model established in this thesis is far lower than that of the single model in the forecast of power load and energy consumption in 2016 to 2019.The verification results show that the combined model can not only provide high-precision forecasting results for power load forecasting,but also predict the power generation of different energy sources more accurately.