Research And Application Of Improved Bayesian Combination Priority Model In Mid- And Long-term Power Load Forecasting

Load forecasting,in order to achieve reasonable planning of power grid,balance of regional power supply,coordination of power grid construction and safety and reliability of power grid,establish a strong grid that can adapt to the trend of national electricity consumption under the rapid development of economy.The realization of this goal is a guiding significance for the medium and long-term electricity load prediction data,which plays a decisive role in the system's economy and reliability.Mid long term load forecasting is affected by many factors such as time interval,economic structure,government measures,agricultural structure,population structure and so on.It is the key point to combine these factors with the existing forecasting technology.At present,the prediction models can be divided into single factor prediction model,multi factor prediction model and combination forecasting model.It is the important research content of this paper to discuss the application scope of these prediction models and determine the models that meet the long-term load forecasting in Yunnan province.The main contents of this paper are as follows:1.Learning the theory of power load forecasting,function,characteristic,principle,mainly from the basic procedure of power load forecasting,analysis of errors in error,long power load forecasting factors in collecting relevant data of power load forecasting of Yunnan Province,preparing for the following prediction.2.For a single prediction model,study the gray forecast theory,grey prediction and boundary value problem of selecting specific value model in the background,and the boundary values revise the ideal background obtained by genetic algorithm training,so as to establish the GM based genetic algorithm(1,1)model and GM(1,1)model,discrete GM(1,1)model,Verhulst model,residual GM(1,1)comparing the model prediction results,the results showed that the prediction results of the model are compared with their ideal.3.For the multi factor prediction model,first we need to identify the main influencing factors.We use principal component analysis to establish the main factors that affect the load forecast in Yunnan province.We use BP neural network training to establish PCA-BP neural network prediction and get the final result.Compared with the single factor prediction model,the prediction accuracy is higher than the single factor prediction model,but the efficiency of the prediction is low.4.According to the combination forecasting model,learning mechanism of the least squares support vector machine model,modeling principle of Bayesian forecasting,set the core parameters of LS-SVM,determine the criteria to determine the parameters of Bayesian theory parameters,establish Bayes-LS-SVM prediction model,the predicted results show that the combination forecasting model is better than single combination forecasting model and multi factor forecasting effect is better model prediction.