Cold Load Prediction Based On Particle Swarm Optimization And Control Error BP Neural Network And Dynamic Goal Control

Building energy consumption accounts for about30%of the total energy consumption ofthe community, building energy efficiency has become an important part of energy saving.Air-conditioning system is the big part of building energy consumption, according to statistics,the energy consumption of central air conditioning accounts for about50%of the totalbuilding energy consumption. Whether the Building air conditioning system is energy savingand whether the energy consumption is reduced plays a key role for the success or failure ofbuilding energy efficiency. How to make air-conditioning energy-saving and how to minimizeair conditioning energy consumption and operating costs and improve the economic benefitsof air conditioning systems is a difficult and must be addressed. Predictive control for airconditioning provides an effective way to make Air conditioning energy to reduce. It meetsthe building cooling load minimizing the unnecessary waste, according to the actual buildingcooling load regulation and control. Accurate cooling load prediction is the effective basis ofair conditioning system operation and management. It is also the basis of predictive control toimprove the control and regulation of predictability and system stability. This paper takesbuilding air conditioning system energy-saving for goals, based on the ideas of ‘forecastguides control, control revises forecast’, researches and analyzes the cooling load forecastingtechniques, combined with control and feedback, and proposes the building cold loadforecasting techniques to improved BP neural network based on PSO and control errorfeedback, then builds the building cooling load prediction and dynamic target adaptive controlmodel, and applies to actual building.In this paper, the research work and achievements are as follows:First of all, this paper discusses the building cooling load prediction development andcurrent research, describes the significance of cooling load forecasting, and analyzes thecharacteristics and principles of building cooling load forecasting. It also introduces thecooling load forecasting techniques and describes the problems.Second, the artificial neural networks, particle swarm optimization and intelligent control theory are described. For the artificial neural network, the main research focuses on theprinciples and improvement method of BP neural network algorithm, including themathematical form, network structure, input parameters, learning algorithms and theapplication of BP neural network. For particle swarm algorithm, the relevant knowledge,optimization theory, algorithm processes are introduced. The advantages of the particle swarmalgorithm are pointed out. The feasibility analysis of the complementary effect with BPcombination with PSO is discussed. For the Intelligent control, this paper focuses on thedescription of the adaptive control theory, including details of the principle of adaptive controland content, control flow and application scenarios.Third, BP neural network is used to predict the building cooling load. Since theshortcomings of the BP neural network prediction method, such as slow convergence, easy tofall into local optimum and relatively low accuracy, the particle swarm optimization algorithmwhich has faster convergence and stronger global search capability is introduced to improvethe prediction. By adding control error feedback parameters to the input parameters of the BPneural network structure, the improved BP neural network forecasting techniques based onparticle swarm optimization and control error feedback is formed (PSO-CEF-BPNN),combines the particle swarm with BP neural network and enhance the prediction accuracy.Fourth, with the combination of the predictive value and actual operation of the controlparameters of building air-conditioning system, the adaptive control method based ondynamic target(DTAC) of building cooling load is proposed. The method puts airconditioning energy-saving as goals, based on forecasting techniques, and makes predictivevalue as control target value, then makes building real-time adaptive regulation and control.control process is subdivided according to the time period, combined with building the actualenergy consumption, and calculates the control error, then dynamically adjust the target toimprove the follow-up control, making the control approach the target. This achievesenergy-saving targets under the premise of meeting the building cooling load, minimizingunnecessary waste, with the regulation and control according to the actual building coolingload. In addition, the control error of different time period anti-spread to prediction model, as forecast input parameters, and then revised forecast model.Finally, large amounts of data are collected for a building, providing sample data for thecooling load prediction model. Combined with the cold load forecast value (PSO-CEF-BPNN)and the dynamic target adaptive control (DTAC) method, the model is built and a program iswritten to control building cooling supply. Based on the experiments determination of sampledata, this paper analyzes the forecast error and the actual effect of the control, and throughcomparison of the building cooling energy data, validates of the model which can achieve thedynamic adjustment according to actual building cold load demand and achieve energy savingeffect.