Research On Power Consumption Prediction Of Domestic Solar Water Heater Based On Improved BP Neural Algorithm

Solar water heaters with energy saving and environmental protection,easy operation,etc.,favored by the majority of users.With the increasing demand for solar water heaters,it has become China's new energy field in the development of more mature products.As the solar water heater operating system hardware equipment can not be used to accurately predict the heating time and the lack of heating or over burning phenomenon,resulting in shortened service life,there are security risks,and wasted a lot of power.The BP neural network in the field of artificial intelligence is more mature,it has a strong non-linear function mapping ability,can be more complex for the sample data to predict,therefore,is widely used in various industries.The shortcomings of BP algorithm are fixed in the form of algorithm function,and can not adjust the function flexibly for the specific sample data,which leads to the large error of the prediction result and the slow convergence rate.Aiming at the problems existing in BP algorithm,two improved schemes are proposed: an improved Sigmoid type function algorithm based on adjustable displacement factor and an improved energy function algorithm based on adjustable parameters.Based on the adjustable displacement factor Sigmoid type function to improve the basic idea: to improve the BP algorithm in the neuron node at the transfer function,mainly for the S-type function to improve the S-type function in the introduction of adjustable displacement factor,thereby improving the standard BP Algorithm function form is immutable problem,and the improved function is integrated into the MATLAB neural network toolbox.The objective function(error function)in the improved standard BP algorithm is improved by adding the sample relative error function on the basis of the original error function,and the scale factor is added.By adjusting the scale factor,Absolute error,the relative error of the weight,so as to determine the best sample objective function.Finally,this paper will improve the algorithm for home solar water heater electric heating time prediction,the forecast results are more accurate,the forecast results will help users,manufacturers in a timely manner to make technical treatment,to avoid the loss caused by burning.This paper is divided into six chapters,the first chapter is mainly part of the introduction,mainly elaborated the background,content and development process of the paper,and gives the main work content of the paper;the second chapter mainly for the BP neural theory part of the This paper introduces the theory of artificial neural network,gives the application fieldand characteristics of BP algorithm,and deduces the detailed formula.In chapter 3,we introduce the theoretical part of the two improved algorithms and give the derivation process.Five chapters for the improved algorithm simulation experiments,the fourth chapter in the two improved algorithm and standard BP algorithm comparison study;the fifth chapter in the two improved algorithm for comparison study.The sixth chapter is the introduction,which mainly summarizes the paper and prospects the future work.