Research On The Control System Of Reheating Steam Temperature Of Supercritical Unit Based On Internal Model Principle

Control system, like nerve center, has played a very essential role in thermal power plant in terms of security, stability and economy, especially for supercritical unit. For supercritical unit, reheat steam temperature control is a substantial procedure to ensure the safe operation and economic efficiency. Whereas much high and low reheat steam temperatures have quite adverse effects on safety and performance, it is an essential requirement to control a reheat steam temperature reasonably. Currently, the traditional PID control systems are still adopted to control the temperature of reheat steam in many supercritical power units. However, in fact, it is a typical non-linear process that the system governs the reheat steam temperature, where the traditional PID control may perform poorly. To solve these problems, this paper investigates in some aspects as below:1?Aimed at the typical characteristics of non-linear systems, the method of RBF neural network is employed on the model of reheat steam temperature system of supercritical unit. During the modeling process, referring to the actual case of a 600MW supercritical power unit, the openings of flue gas baffle and desuperheating water valve as well as the power instruction are considered as the input variables for control system, and the data of the unit which model bases on, cover the range of full power operations, so that the model has a decent adaptability for full power operations.2?To solve problem that the control of reheat steam temperature perform ineffectively in actual operation, in this paper, the system applies the method of internal model to control reheat steam temperature and use RBF neural network to establish reheat steam temperature system inverse model,where the internal model controller (IMC) is NNC, in order to establish the neural network model controller. In addition, a low-pass filter employed in the controller, reduces the impact of the noise signal on the control system.3? For traditional PID control of reheat steam temperature, due that the reheatd steam temperature system has large lag and inertia problems, the system has quite difficulty in achieving automatic control. In reality, the control operations are usually manual, as a result, one may likely adopt the resuperheating water as main method to control temperature, which may decrease the economic performance. This paper provides a method of Neural Network Internal Model Control for reheat steam temperature within an integrated economic index. During training the RBF neural network, the internal model controller NNC, traditional target just is to achieve a minimal deviation e between the temperature of set and output. Yet in order to improve the economic performance of the unit, not only to reduce the deviation, meanwhile the control system in this paper, but also considers to minimize the opening of desuperheating water valve. After the establishment of the controller, the first test is repectively at the high, medium and low power condition, giving a negative step instruction to examine ability of the reheat steam temperature neural network internal model controller to track changes of control instruction; then maintaining steam temperature setpoint control command unchanged, a fixed step disturbance is added to test the anti-jamming capability of reheat steam temperature neural network internal model controller; meanwhile, in these two tests, observing the in opening of desuperheating water valve and flue gas damper to assess and verify the economic performance.4?For the traditional RBF neural network model, it may be difficult to accurately determine network structure and learning is slow, this paper, using improved genetic algorithm, hierarchical genetic algorithm, is to optimize reheat steam temperature RBF neural network model and the structure and parameters of internal model controller NNC at the same time, and repeat the procedure as step 3 for verification.