Research on Pollution Gas Position Based on Multilayer Perceptron Algorithm

Abstract

At present, the national treatment and prevention measures for industrial exhaust gas pollution have transitioned from industrial upgrading, capacity reduction, and energy structure regulation to the stage of precise positioning and precise treatment due to the phenomenon of stealthy and excessive discharge of exhaust gas in industrial parks, which has brought serious threats to the safety of the atmospheric environment. In addition, the exhaust gas emission of industrial parks is characterized by randomness, abruptness, complex gas composition and large gas emission. Therefore, traditional pollution location methods such as emission inventory method and grid method are not effective. Based on the above problems, this paper proposes an air pollution gas location strategy based on neural network multi-layer perceptron algorithm. Through pattern recognition, fuzzy linear discriminant function is applied, and the feature space is segmented by hyperplane discriminant boundary and the fuzzy area is retained. Determine the orientation of the fuzzy discriminant surface by measuring the weight vector of a specific neuron and formulate a method to initialize the initial weight of the network on the hypersphere. Determine the weight initialization hypersphere by measuring the distance from the origin to the discriminant surface and the offset and then to further determine the specific location information of the polluted gas. After proposing the control strategy, this paper performed a specific simulation verification on the MATLAB platform. The verification results show that the algorithm strategy can greatly reduce the learning time of the neural network, improve the convergence performance of the network, and significantly improve the accuracy of polluting gas positioning.