PMV Prediction Model Based Oil PSO And BP Neural Network Using In The Intelligent Office Buildings

Nowadays the air conditioning system has accounted50%-60%of the entire officebuilding energy according to survey.In this condition,new air conditioning energy savingcontrol technology become a top priority.Most traditional air conditioning’s energy savingcontrol system take indoor temperature as controlled parameter.It is difficult to achieve acomfortable indoor environment.What’s more,it also may increase the unnecessary energyconsumption.Professor Denmark Fanger has put forward the PMV index.Now most scholarsstudy Denmark Fanger’s scheme and develop application base on the PMV index. Due to thecomplex nonlinear relations between the PMV index and it’s several factors, It cannot adapt tothe real-time controlof the air conditioning energy saving control system because of tediousiterative operation.Therefore, there are some scholars using the BP neural network to predictPMV, but as the BP neural network itself exists defects, it also affects theprecision,convergence and stability of the PMV prediction model.At the same time, the BPneural network topology become different with the application scene.How to determine theoptimal network structure is also a problem.Based on the above problems, this paper proposesa new PMV prediction model based on particle swarm optimization algorithm and BP neuralnetwork,and applies this model to large office building energy system,making Energyconservation,Comfortable Intellectualization as the three development direction of a largeoffice building.The main work of ths paper are in following aspects:First, systematically discussing the situation in the large office building airconditioning energy saving control field, the human body comfort PMV index, artificialneural network theory, particle swarm optimization algorithm, then puts forward the existingproblems.Second, this paper studies the BP neural network algorithm,the determination of BPneural network model of prediction of the PMV index, the determination includes: the BPneural network topological structure (implicit layers, node number, etc.), learning rate,excitation function, sample size, etc. And it also study the basic advantages and the existingdefects of BP algorithm.Third, this paper studies the relevant principles of particle swarm optimization (p so)algorithm, the algorithm flow, and the feasibility analysis of its combination with BP neuralnetwork.At the same time,this paper proposes the concrete way to combined particle swarm algorithm with BP neural network, and compares its performance with the basic BPalgorithm’s.Fourth,this paper put forward and set up a PMV model based on improved BP neuralnetwork upon particle swarm,which as an intelligent analysis module joins into the intelligentcontrol system,through the cooperation with other modules (data acquisition module,command analysis module, communication interface module, etc.),building the intelligenceoffice building control system.This paper also applies the system to a real office buildingoperation,and verifies the model’s double purpose of energy conservation, comfortable changeby gathering the actual operation effect and the actual operation data.