Design And Application Of Function Weight Single Input Module Connected To Fuzzy System

With the in-depth study of the theory and method of fuzzy inference system,it has gradually involved in all aspects of people's production and life,and has become one of the indispensable high-tech means.However,with the number of input variables increases,the traditional fuzzy inference system is easy to face the problem of "rule explosion" when modeling complex systems and forecasting,which brings great difficulties to the construction and design of the system.Some researchers have proposed many solutions to this problem in recent years,among which the proposed functionally weighted single-input-rule-module connected fuzzy inference system(FWSIRM-FIS)with simple structure and few rules,and has shown superior performance in modeling and control.But there is little research on it,in order to further improve the performance of FWSIRM-FIS and obtain reasonable and effective output result,this paper combines the intelligent algorithm with the fuzzy system to further improve its the structure determination and parameter optimization.The main research contents are as follows:Firstly,the single-input-rule-module connected fuzzy inference system(FWSIRM-FIS)is introduced in detail,then the FWSIRM-FIS and its parameter training process is introduced in detail.To verify the performance of FWSIRM-FIS,this paper applies it to the field of building load forecasting,and compares it with the commonly used the Back-Propagation Neural Network(BPNN),the Adaptive-Network-Based Fuzzy Inference System(ANFIS),and the Multivariable Linear Regression model(MLR).The experimental results prove that this method can accurately predict the building load,and its performance is better than the other three methods.Secondly,in order to improve the rationality of prediction results and provide more effective information to decision-makers,an interval functionally weighted SIRM-FIS(IFWSIRM-FIS)is proposed in this paper.It replaces the original functional weight with the interval functional weight to extend the traditional FWSIRM-FIS,so as to achieve the interval output and provide reasonable prediction for the target value.At the same time,in order to make the output interval width more reasonable and obtain the optimal parameters and minimum error of the system,in this paper,the parameters of IFWSIRM-FIS are optimized by using the iterative least square algorithm and multi-objective optimization algorithm.Finally,in order to verify the performance of the proposed IFWSIRM-FIS,it is applied to the building load prediction.The experimental results have proven that the output interval of the system can effectively cover the real observations.Finally,considering the periodic characteristics in many applications(energy consumption,traffic flow,etc.),this paper proposed a construction method of hybrid forecasting model based on FWSIRM-FIS,which is driven by periodic pattern and data.The prediction model is composed of a mixture of periodic knowledge module and data-driven FWSIRM-FIS module.Firstly,the wavelet transform is used to dete rmine the original data before the periodicity extraction,and then the periodicity knowledge is extracted by means of the mean method.In addition,in order to improve the performance of the system,this paper uses the subtraction clustering algorithm and the least squares algorithm to optimize and design the structure and parameters of the system,respectively.Finally,the FWSIRM-FIS constructed by this method is applied to the problem of building load prediction,and compared with the traditional four methods which are FWSIRM-FIS,BPNN,ANFIS,and MLR.The experimental results show that the hybrid prediction model constructed in this paper can effectively improve the prediction accuracy of the system and has the best prediction performance among all models.