The Optimization Of Grey Multi-variable Forecasting Model And Its Application

After more than 30 years of development,the grey system theory has achieved a wealth of research results.While the grey multivariable forecasting model has been a hotspot of scholars' research because of its better response to the characteristics of "part of the information is known,part of the information is unknown".However,with the rapid development of economy and society,the limitations of the existing grey multivariable model in dealing with the practical application of complex problems began to emerge.Therefore,it is necessary to improve the grey multivariable model according to the types of the actual problems,so as to enhance the timeliness and application scope of the grey multivariable model.In this paper,TMGMC(1,N)model and DDGMD(1,N)model were constructed to solve the problem that the mechanism of driving variables is not clear and the dynamic time-delay characteristic of driving variables is not taken into account,and explored the mechanism of driving variables of grey multivariable model and the new method to identify driving variables and their time-delay,then the identification theorem of model parameters and the time response formula are verified.Finally,the optimization models were applied to the forecast of energy consumption in Jiangsu province.In this paper,the research is mainly from the following three aspects:(1)In view of that multi-variable grey model has rarely considered the mechanism between the driving variables and the changeable characteristics of the action intensity of the driving variables on the behavior sequence of the system during the period of action,through the trends of multiple driving variables model to reflect the mechanism between the driving variables.Based on that,a new grey prediction model called the trends of multiple driving variables grey model with convolution integral,abbreviated as TMGMC(1,N),was constructed.And the convolution integral was used to solve the whitening equation in order to reflect the changeable characteristics of the action intensity of the driving variables on the behavior sequence of the system.Finally,the forecast of China's grain yield is adopted for demonstration.The result shows that the accuracy of the output is higher by the TMGMC(1,N)model than by the existing GM(1,N)model.(2)Aiming at the questions that traditional grey multi-variable discrete model are not considered the time-delay dynamic characteristics and the utilization of previous data of the driving variables,a new grey multi-variable discrete model,abbreviated as DDGMD(1,N),was constructed.And model parameter calculation method is discussed by introducing adjustment coefficient Ti and effect coefficient ?i of the driving information control.Then,identified the adjustment coefficient Ti,which clarified the driving variables and their time-delay parameters,by the method of grey extended dimension identification;determined effect coefficient ?i,which reflected the influence of the previous data of driving variables to system behavioral sequence,by the particle swarm optimization algorithm.Finally,a real application about the forecast of the GDP in Jiangsu Province is used to demonstrate the feasibility and practicability of the DDGMD(1,N)model.The validity of the model is proved to be suitable for the prediction of small sample data with time-delay characteristics.(3)The grey multivariable optimization models are applied to the energy consumption forecast in Jiangsu Province.Firstly,the TMGMC(1,N)model is used to analyze the driving factors and development trends of the total electricity consumption in Jiangsu Province from 2009 to 2014.Based on that,the total electricity consumption of Jiangsu Province was modeled and predicted.Then,the DDGMD(1,N)model is used to model and forecast the energy consumption in Jiangsu Province from 2003 to 2014,and the results are explained.