Research On Neural Network Predictive Control Method For Anti-condensation Of Electric Cabinet

In the operation of power system,the requirement of power supply security and reliability is getting higher and higher,and the condensation problem in cabinet will threaten the power supply security and reliability of power system.The existing anti-condensation methods for electric cabinets have some problems,such as poor effect,high cost,large power consumption,potential safety hazards,etc.Moreover,the method of using on-air conditioning to prevent electric cabinet condensation needs to keep the air conditioning open continuously,thus causing a lot of power loss.Therefore,in view of these problems,a neural network predictive control method for dew prevention of electric cabinet is designed to control the start-up and shutdown of air conditioning in electric room.Firstly,this paper analyses the mechanism of condensation in the cabinet,describes the process of condensation from two aspects of relative humidity and dew point temperature,introduces the calculation method of dew point temperature,and summarizes three condensation criteria of condensation in the cabinet.Taking GGD type AC low voltage distribution cabinet as an example,a simplified physical model and mathematical model of the cabinet are constructed,and the environmental simulation of the wet air in the cabinet is carried out by using computational fluid dynamics(CFD)simulation.The temperature distribution and relative humidity distribution of the wet air are analyzed and studied.Combining with condensation criterion,the location of condensation in the cabinet is determined.Then,taking GGD AC low-voltage distribution cabinet as an example,sensors are installed in the dew-prone position and other related positions in the cabinet.Two sets of dewrelated data acquisition experiments are designed in the air-conditioning closed state and the air-conditioning open state respectively.The original data of temperature,humidity and voltage and current at the specific location of the acquisition point in the specific environment are obtained,and the original data of the experiment are obtained.The data are pretreated as training data and test data for subsequent establishment of neural network prediction model.Secondly,based on BP neural network,this paper designs the predictive model of humidity and humidity of humid air in electric cabinet and the predictive model of condensable surface temperature in electric cabinet.Based on the pretreated experimental data,four models of open and closed state of air conditioning are trained,and the output of the model is used as the basis of start and stop of air conditioning.Finally,based on the previous research results,this paper designs the basic idea and concrete structure of the neural network predictive control for the condensation of the electric cabinet,gives the concrete form of the condensation criterion in the controller according to the actual situation,and simulates and verifies the effect of the neural network predictor for predicting the condensation time and the effect of the anti-condensation of the electric cabinet after opening the air conditioner.