Application Of Neural Networks PID Controller To Heating Networks Flow Regulation

On the basis of Laboratory-scale heating system control, this paper aims at the study of running adjusting and control methods for District Heating Network. It systematically discusses the application of Neural Networks technology to District Heating System to adjust the flux. includes:The first part of this paper summarizes the running management, adjusting and control technology at home and abroad. By analyzing the characteristics and possible problems of District Heating System in China, the flow control strategy based on temperature is presented, which adapts to the situation of our country.The second part of this paper discusses the methods of multivariable decoupling control by the numbers. In accordance with the District Heating System having strong coupling characteristic, this paper combines the conventional PID controller with a simple structure and the BP Neural Networks with the ability of solving nonlinear system control problem. Then, a Neural Networks PID multivariable controller with the ability of solving coupling problem is proposed. And then the characteristic and design idea of the controller are thoroughly discussed. Furthermore, on the basis of the improvement on BP algorithm and the controller's structure, this paper adds a nonlinear prediction model to solve the time-delay problem.The last part of this paper introduces the composing of the Laboratory-scale heating system, the design of the real time monitoring platform and the implementation of the Neural Networks PID multivariable controller with a nonlinear prediction model. The macro control method is applied to the actual management and control of the Laboratory-scale heating system. By adjusting the flux of the user's primary pipe network the user's backwater temperature of secondary pipe network can be controlled. Large numbers of running results prove the Neural Networks PID multivariable controller with a nonlinear prediction model has a perfect effect of solving nonlinear, strong coupling, huge time-delay problem comparing with normal BP Neural Networks, non-coupling, non-prediction controller and they also prove that the controller can rationally distributes the heat throughout the heating network, and the heating quality is improved. Moreover, the goal of well-proportioned heating is also achieved.In the end. the work done is summarized and future work is prospected.