Research And Application On Energy-saving Process Control In Power Plants

With the extensive attention on energy consumption, it is necessary to develop energy-saving control for the existed equipments in power plants. Improving the utilization efficiency of fossil fuels is one of the most important ways to solve the energy problem. Energy-saving project in power plants needs design a lot of control systems with necessary equipments. Nevertheless, only the novel control strategies for superheaters in power plants will be investigated for purpose of saving more energy in this thesis.. Stochastic distribution control algorithm will be applied to the primary controller of superheated steam temperature cascade control system in power plants. Since the disturbances existed in practical processes are probably non-Gaussian, the performance index will be constructed by the entropy and mean value of tracking error besides the constraints on control energy. We can have obtained the control laws by minimizing the performance index using stochastic gradient descent optimization algorithm. The shape of tracking errors’probability density function becomes as sharp and narrow as possible. In addition, the thesis also introduces back-propagation neural networks to the stochastic distribution control framework and design the primary controller of cascade control system for regulating superheated steam temperature based on minimum entropy of tracking error in outer closed loop control system. The stochastic gradient descent algorithm is also adopted to minimize the entropy of tracking error when parameters in primary regulator are obtained adaptively via updating the weights of BP neural networks. Consequently, the uncertainty of system is minimized and the shape of probability density function is as sharp and narrow as possible. The simulation results verify the effectiveness of both direct and neural networks assisted stochastic distribution controllers...