Study On Regulation Strategy To Achieve Source-network Balance In District Heating System Based On Load Prediction

With the deepening of energy conservation and emission reduction,there are higher requirements for the rational and refined level of energy transmission and distribution and operation regulation in district heating systems.Focusing on the problem of supply-demand matching in district heating systems,this thesis proposes a strategy of balancing and regulating source-network based on load prediction.This thesis studied the key factors such as selection of parameters affecting heat load prediction,data pre-processing,prediction methods,regulation period and regulation strategies.Then,the proposed operation strategy was applied in a typical heating system to verify its effectiveness.The thesis is arranged asfollows:Firstly,simulation and data analysis were conducted to study the thermal delay characteristics of the primary network in order to determine the regulation interval of the heat source.The thermal substations were classified considering the influence of building types and heating terminals.The regulation intervals of thermal substations were determined based on the comprehensive delay time of the indoor temperature of the heating buildings.The regulation interval of heat source studied in this thesis is 12 h.Thermal stations are clustered into 3categories,i.e.supply heating service for(1)non-energy efficient buildings with radiators(2)energy-efficient buildings with radiators(3)energy-efficient buildings with floor heating.The regulation intervals of thermal substations are 6 h,8 h and 12 h respectively.Secondly,the factors affecting the heating load were theoretically analyzed.And the input variables of the prediction model were preliminarily selected based on the available data of practical engineering.Correlation analysis and significance test were carried out on the preliminary screening variables.The influence periods of historical data were determined according to the relative error.The prediction models of thermal substations with outdoor temperature,indoor temperature and historical water supply temperature as independent variables were established to predict water supply temperature.The prediction model of heat source with outdoor temperature,pressure difference between supply and return water,flow rate,and historical water supply temperature as independent variables was established to predict water supply temperature.Thirdly,data preprocessing method was studied.Outlier data elimination effects of 2σcriterion,3σ criterion and Box-plot on were compared.The results show that the 2σ criterion is highly sensitive to outlier data without distortion.Then,the vacant data was filled by cubic spline interpolation method.An integration method of Exponential Smoothing and Gaussian Window was proposed to eliminate data noises.Then,the superiority of proposed method over exponential method and gaussian window single processing is verified by using root mean square error as smoothing index.Fourthly,the principles of four prediction methods,i.e.multiple linear regression,BP neural network,GRNN neural network and ELMAN neural network,were analyzed.Then,the prediction models of thermal substations and heat source are determined as multiple linear regression and GRNN neural network respectively by comparing the prediction accuracy with actual data.Finally,the predicted parameters of heat source and thermal substations were modified based on energy conservation of the whole network and dynamic feedback of indoor temperature to achieve the balance between supply and demand of source and network.The time nodes of regulation period were determined with the minimum mean absolute percentage error of load prediction as the goal.Through verifying the regulation effect of a typical heating system,it is shown that the indoor temperature of all typical users is stable.In addition,the fluctuation range of pipe network pressure is reduced by 69.6 % and the energy saving rate of system is around 13.60 %.With the application of Internet,artificial intelligence and online monitoring,the dissertation proposes a strategy of balancing and regulating source-network based on load prediction.This strategy helps maintain indoor temperature stable and reduce energy consumption,which may provide theoretical guidance for practical operation.