Prediction Of Environmental Parameters Of Metro Stations Based On Stochastic Network

With the improvement of China’s subway operation capacity,people in big cities prefer to travel by subway.Some subway lines have a large operating load during the morning and evening peak hours,causing the environmental quality of some early-built subway stations to decline.The environmental parameters in the metro station are affected by factors such as air flow and crowd flow,and have special changes.The environmental parameters at different locations in the station are both connected and different.Therefore,it is difficult to monitor and predict the environmental parameters of important nodes in the station for a long time.Environmental parameters for this characteristic,this paper studies and uses the Random Vector Functional Link(RVFL)neural network method in the stochastic network to explore the changes of environmental parameters.The neural network can find out the changing rules of environmental parameters in a period of time by training historical data.Compared with the traditional BP neural network,the RVFL neural network has the advantages of simple modeling,fast training speed and accurate prediction results.In the multi-step prediction of environmental parameters,the traditional single-step rolling prediction will gradually reduce the prediction accuracy as the number of steps increases.This paper proposes an error compensation strategy based on multi-output prediction based on RVFL neural network,which can make multi-step prediction more accurate.At the same time,only need to adjust the network parameters and the number of output layer nodes,it can predict the future data.The network structure is simple,no loop input is required,and the model can be updated online.Through experimental verification,both prediction accuracy and speed have great advantages.On this basis,this paper analyzes the environmental parameter data of multiple spatial nodes in the station,performs network modeling on the transmission relationship of environmental parameters between different nodes,and performs parameter regression on the environmental parameter data of a node.The combination of Stochastic configuration network(SCN)and RVFL model enhances the generalization performance and stability of regression model.This method use a small number of sensors to monitor and predict the environmental parameters of different nodes in the station space for a long time,reducing the cost of environmental monitoring.Through the experimental research in this paper,it provides a fast and effective method for environmental parameter monitoring and prediction in metro stations,and provides accurate data support for environmental control measures.