Model Predictive Control For Air-conditioning Terminal Of Subway Platform Based On Load Prediction

With the continuous progress of science and technology,the subway has become the main public transportation mode for people to travel in the city.As an important part of the subway design system,the subway ventilation and air conditioning system has attracted widespread attention on energy saving and intelligent control.According to statistics,the energy consumption of the subway air-conditioning system accounts for about 30% to 50% of the entire subway system.At present,subway air conditioning systems generally carry out system design and equipment selection according to the upper load limit.Since general equipment cannot use effective frequency conversion control methods to improve the environment,it causes energy waste.At present,PID control is commonly used in subway air-conditioning systems for system control.This control method is difficult to achieve good results for nonlinear air-conditioning systems with strong coupling,large time delay,and multiple interference.In recent years,with the continuous development of intelligent control technology,predictive control strategies have been applied in all walks of life,which can meet the environmental needs of actual projects,and solve the problems of poor dynamic performance control by PID control methods and high investment requirements for system energy-saving transformation.In order to solve the problem of traditional predictive control optimization algorithms that are computationally intensive and not easy to implement in engineering,this paper studies the existing predictive control methods and proposes predictive control methods and strategies based on neural network controller.Combined with load forecasting research,a control simulation experiment is carried out for the air conditioning system of the subway platform.The results show that the neural network predictive control method is more superior to the traditional PID control method in the control of the ventilation and air conditioning system of the subway platform.First,in order to meet the optimal control requirements of predictive control methods,the paper predicts the air conditioning load of subway platforms.In order to obtain a reliable load forecast,the paper uses BP artificial neural network to construct a load forecasting model.This model can continuously optimize the model through self-correction of neural network weights without clarifying the internal relationship of the system.Through the analysis of the influencing factors of the subway platform load and the actual data,the input parameters of the neural network model are determined as the outdoor temperature,outdoor humidity,the current load value,the load value at the same time of the previous day and the load value at the same time of the previous week,and the output parameters are determined It is the load value at the next moment.In this paper,the experimental method and trial and error method are used to select the number of hidden layer neurons with the smallest system error.Finally,a three-layer neural network model with 5 input neurons,11 hidden layer neurons and 1 output neuron is established.At the same time,the paper uses the median average filtering method to filter the noise and interference of the actual data of the air conditioning operation for a whole year to ensure the accuracy of the data.On this basis,the paper divides the actual data into a training set and a test set,and performs a training test on the neural network load prediction model.The results show that the relative error between the predicted load and the actual load output by the neural network load forecasting model established in this paper is between-0.015 and +0.02.Satisfying the requirements of the actual engineering error range has laid the foundation for the predictive control method research of the subway station air-conditioning system described later.Secondly,a predictive model of the controlled object is established,and a predictive control method using a neural network controller is proposed to predictively control the subway station air-conditioning system.Through theoretical analysis of the terminal air conditioning processing equipment(AHU)of the subway station air conditioning system and the platform environment,the paper determines the coupling relationship between the control parameters of the air conditioning system.According to the law of energy conservation and the law of conservation of mass,combined with the mechanism modeling method,the establishment the predicted model of the controlled object.The thesis uses the first-order backward difference method to convert differential equations into difference equations,so that it can be used for predictive control.After establishing the predicted model of the controlled object,the thesis studies the predictive control strategy based on the neural network,uses the neural network as the controller to implement the predictive control rolling optimization algorithm,and determines the input of the neural network controller as the current temperature value of the station temperature and the station temperature target Value,current value of platform moisture content,current value of platform moisture content,current value of supply air temperature,target value of supply air temperature,and disturbance variables at the current time(outdoor temperature,outdoor moisture content,platform load and platform environment content Humidity,etc.),the output of the neural network controller is the system control quantity,including the chilled water flow and the air supply air volume.In order to obtain the best optimization performance index and the best control quantity in different time domains,the thesis calculates the Lagrange multiplier in the performance index and continuously revises the weights of the neural network controller.Finally,the thesis takes control quantity and state quantity as parameters,and establishes optimized performance indexes aiming at passenger comfort of subway platform and system energy saving.By establishing the Hamiltonian function,the constrained functional extremum problem is transformed into an unconstrained extremum problem.Taking the optimized performance index as the objective function,the control simulation experiment of the neural network predictive control method in the air conditioning system of the subway platform is carried out.Through the comparative study of predictive control simulation experiment and PID control simulation,it is concluded that the predictive control method based on neural network controller can quickly make the subway platform environment reach the set value from the initial value,and in the case of interference from external factors,it can still Maintain a good stable state.