Optimized BP Neural Network And Its Using In The Heat Exchange Station

Heat exchange station that is widely used in organizational form on concentrated heating system, for energy-saving, environmental protected, reducing operating costs, improving comfort considerations, must be studied in control, thus ensuring the heating efficiency maximization. Heat exchange station is a typical complex control system, its characteristics with large inertia, large delay, not easy to establish a mathematical model, the station quality channel and quantity channel exist coupled in the regulating process. The traditional control strategy difficult to obtain satisfying control effect, BP neural network is adopted to control the heat exchange station become effective choice.BP neural network is a kind of intelligent algorithms. the theory has the ability to fit any nonlinear, but BP neural network also exists shortcomings including network convergence speed slow, not easy to determine the learning rate, easy to fall into local minimum solution, difficult to determine the number of nodes in the hidden layer. Around these problems, the academics proposed a number of optimization strategies, Genetic Algorithm and Particle Swarm Optimization are the common methods optimized BP neural network.Genetic Algorithm(GA) and Particle Swarm Optimization(PSO) have many similarities, also the common methods to optimizing neural network. This paper introduces the basic ideas, the characteristics of the Genetic Algorithm and the Particle Swarm Optimization, and the research status of the academia. The improved algorithm of Genetic Algorithm is proposed. The improved PSO algorithm based on PID is proposed, and the validity of the improved algorithms is verified by the test functions.In this paper, in accordance with the problem of BP neural network, Genetic Algorithm and Particle Swarm Optimization are used to optimizing the BP neural network.Comparing the results of optimization reveal the effectiveness of BP neural network influenced by data. The precision of the BP neural networks influenced by the number of nodes in the hidden layer network; with the same structure of neural network, the initial weight value determine the neural network accuracy and can even lead to network can not meet the performance requirements. On fast the research of optimization algorithm show that the intelligent algorithm cannot simultaneously satisfy two properties: fast and accurate, expecting to meet the need of fast, may lose the optimal solution; expecting to meet the optimal solution, it is necessary to spend a long times。Finally, established a steady state model of heat exchange station using optimized BP neural network. In order to meet the heat exchang station energy-saving control, puts forward the energy-saving renovation scheme, using intelligent algorithm to optimize the BP neural network do as the controller, decouple the heat exchange station of quanlity channels and quantity channel. The water temperature and flow regulation of the secondary side pipe are respectively automatic controlled with dual controller.