Simulation Study On Neural Network Inverse Control For The Coordinated System Of Supercritical Unit

A supercritical once-through boiler unit is a typical strong-coupled system withlarge inertia and non-linear, slow time-variant, time-delay characteristics, which oftenmakes the control quality of the coordinated system with traditional PID controllersdeteriorate under wide-range load-changing conditions, and thus influences the unit loadresponse speed and leads to heavy fluctuations for main steam pressure. For the purposeof improving the coordinated control quality of a supercritical power unit, a neuralnetwork inverse control method for the load and main steam pressure in the coordinatedsystem is deeply studied.Firstly, this paper analyzes the once-through unit’s operation characteristicsand the composition of the coordinated control system. Several different kinds ofcoordinated control modes are compared. Then, the principle, structure and thetraining method of a BP neural network, and the neural network inverse control’sconception, design principle and method are introduced. On this basis, the neuralnetwork inverse system models for the load and main steam pressure of a600MWsupercritical unit are built, trained and validated with MATLAB neural networktoolbox. The control schemes are determined and NN inverse controllers aredesigned and programmed with MATLAB language. Control tests are made bycommunicating real-time with the full-scope power plant simulators. The resultsshow that, with the neural network inverse compensation control, the load responserapidity and the stability of the main steam pressure are both greatly improvedcompared to the original PID control, thus significantly improve the coordinatedcontrol quality of the unit.