Cooling Load Forecasting Of Public Buildings And Control Strategy For Refrigeration Station System

In the last decade,the average energy consumption per unit area of public buildings has remained high,and the country has clearly focused on energy efficiency in public buildings.At present,50% of the total energy consumption of public buildings is the energy consumption of central air conditioning systems.The research results of the research team show that most central air-conditioning systems are currently operating in a state of high energy consumption and low energy efficiency.This is because the central air-conditioning system has the characteristics of inertia,multi-loop and strong coupling,and the PID control method cannot achieve good control effects.In addition,PID control can not achieve energy-saving optimization control by using equipment energy efficiency as an optimization performance indicator;and,the existing engineering method is difficult to accurately calculate the energy behavior model of the whole building by using physical principles,so that the modern control theory system cannot cope with the air conditioning system feedback.Difficulties such as hysteresis and mutual coupling,the control results are not satisfactory.In view of the above problems,this thesis studies the predictive control strategy of the refrigeration station system,and carries out the control system experiment simulation for an office building refrigeration station in Guangzhou.The control system adopts two layers of control structure,and the upper layer is optimized layer.The core of the optimization algorithm is to combine the variational method with the neural network,and make full use of the variational method to realize the optimal control of the linear system and the neural network to optimize the nonlinear system.The advantage is that the ratio of the ratio of the cooling capacity and the consumed power generated by the refrigeration system system to the expected value is taken as the optimization performance index,and the optimal setting of the chilled water outlet temperature and the freezing pump frequency setting value of the refrigeration station system is solved by rolling optimization.solution.The underlying controller uses the PID algorithm to quickly track the controlled parameters to the optimal setpoints obtained by the optimization layer.The control objective is to make the equipment of the refrigeration station work in energyefficient conditions on the premise of meeting the cooling demand of the building,thus achieving the control goal of meeting the cooling demand of the building and saving energy.In order to realize predictive control under the premise that the cooling capacity provided by the refrigeration station meets the dynamic load requirements of the building,the load forecasting becomes the premise of optimal control,which requires identifying a reliable air conditioning load forecasting model.After comparing the advantages and limitations of common prediction models,this paper selects BP neural network to construct load forecasting model,and determines the three-layer structure of neural network,9 parameters of input layer,and 12 hidden layer neurons.The digital filtering method of training sample data is studied.The median value average filtering algorithm is used to filter out the random error and pulse error in the collected data.On this basis,the normalization and denormalization methods are used to process the neural network input and output data.The Bayesian normalization algorithm is used to improve the generalization ability of the neural network.The neural network model is trained by using the cold station system data collected on the engineering energy consumption monitoring platform of the research object.The trained load forecasting model is used to predict the future.After 3 days of air conditioning load,the experimental results show that the prediction error results are between-0.03 and 0.03,and the model fitting effect is good.The built-in load forecasting neural network model creates a prerequisite for building air conditioning predictive control.Predictive model,rolling optimization and feedback correction are the three elements of predictive control.The precondition for predictive control is to establish a predictive model.In this thesis,the neural network method is used to identify the air conditioning and refrigeration station model.The cold station system data collected on the engineering energy consumption monitoring platform of the research object is used to train the neural network model.The training error of the completed cold station model is-0.03?0.03.The model is highly accurate and can be used for predictive control.The output of the cold station system prediction model is the next-time cold station system EER,and the parameters of the nine input layers are determined,and the number of hidden layer neurons is 9.The variational algorithm based on variational method and neural network researched in this thesis is to construct the Hamiltonian function together with the equation of state of the cold station system model and the equation of the neural network prediction controller,and calculate the most of the field controller with the rolling optimization strategy.Excellent set value.The neural network controller structure is determined.In order to verify the control effect of the neural network based predictive control algorithm,the simulation is carried out by using MATLAB software.The experimental results show that the algorithm makes the cold station system meet the end cooling load requirements and makes the air conditioning cold station system The energy efficiency ratio EER reaches the set value,and the system energy-saving operation is realized.The air-conditioning system as a whole achieves an effect of energy saving of 14.6% compared with the original system using PID control.