Modeling And Advanced Control Strategy Research Of Supercritical Unit

The supercritical thermal power unit is the main development direction of China's thermal power industry.Compared with subcritical unit,supercritical unit has higher efficiency and lower emissions,but its dynamic characteristics are more complex.The coordinated control system(CCS)of supercritical unit has the characteristics of large lag,strong coupling and nonlinearity,etc.So the existing control strategy is difficult to meet the higher variable load requirements.Now the research about the modeling and controller design is carried out in the following three aspects:1.An improved T-S fuzzy model identification approach is proposed.First of all,the k-means++ algorithm is employed to identify the premise parameters so as to guarantee the number of fuzzy rules.Then,the local linearized models are determined by using the incremental historical data around the cluster centers,which are obtained via the stochastic gradient descent algorithm with momentum and variable learning rate.The effectiveness of the proposed approach is validated by the given extensive simulation results,and it can be further employed to design the overall advanced controllers for the CCS in an USC unit.In addition,the subspace-based state space method is also applied in modeling the CCS of an USC unit,then the accuracy and correctness of the model are verified.Finally,this model provides the model foundation for the design of the state space controllers.2.The new T-S fuzzy model is transformed into a time-varying CARIMA model in this part.Combining with the improved multi-model weighted predictive control algorithm,the variable weight performance index is added to the original algorithm to improve and optimize.The simulation results show that the new multi-model weighted predictive control has a fast tracking speed and good anti-interference ability.As a result,the proposed method's overall control effect is better than the traditional GPC.3.In order to reduce the amount of online calculation in control algorithms,a fast predictive control algorithm is put forward in this paper.This algorithm can optimize the structure of the quadratic programming in the ordinary predictive control algorithm.Since it combines the improved primal barrier method with the warm start method.Simulation results show that the fast predictive control algorithm can greatly reduce the system adjustment time and rolling optimization time.In addition,fast predictive control has fast tracking speed and strong anti-interference ability.