Decoupling Control Of Thermal Power Plant Boiler Combustion System Based On IsPSO-PIDNN

With the rapid economic development, the energy demand of Chinais growing increasingly. The energy conservation draws more and more attention ofthe many nations and enterprises. The energy consumption of thermal power plantoccupies a high proportion in our country, and the efficiency of control on it has asignificant meaning. As the power production process is mainly impacted by thecoupling of system parameter, the decoupling of boiler system is also benefit energysaving and environment protecting. There are limitations for traditional decouplingmethod on multi-variable, non-linear, strong coupling complex objects. Following thedevelopment of intelligent control, the neural networks and intelligent optimizationalgorithm draws a great deal of attention in the field of applied research.PID neural network (PIDNN) which can communicate with neural network to PIDcontrol law nature fusion, combines the advantages of both. For the unique advantageof PID neural network’s structures and algorithms, it can be applied to a variety ofsystems as a controller, and the design process is very simple. However, consideringthe traditional back-propagation algorithm (BP algorithm) of PIDNN, theshortcomings make it difficult to get the desired network weights valves, and even toachieve satisfactory control performance. In order to take full advantage of theexcellent performance of the PIDNN controller and expand its range of applications,we propose an improved simplified particle swarm optimization algorithm (isPSO),so that the improved initial weight valves can be obtained. The proposedisPSO-PIDNN algorithm is used in the decoupling problem of the thermal powerplant boiler combustion system and the effectiveness and efficiency are verified. Thefollowing issues are mainly discussed:(1) In-depth understanding of the working process of the construction of the thermalpower plant boiler combustion control system based on the multivariate couplingproblems, analysis of the principle and the advantages and disadvantages of thetraditional decoupling methods and modern intelligent decoupling method.(2) The structure and control algorithm of PID neural network are introduced andanalyzed. System of signal variable and multiple variables based on BP algorithmwere simulated by PIDNN to test its effectiveness. After the analysis of thetraditional BP algorithm training PIDNN shortcomings, propose an improvedmethod.(3) Clear PSO principle and the advantages and disadvantages of the particle swarmalgorithm, combined with some other scholars improve ideological and algorithm research carried out, the final article isPSO proposed strategy. The improvedalgorithm for the simulation study to verify the feasibility and efficiency of thestrategies.(4) Conjunction with the presentation of isPSO algorithm, using it to optimizePIDNN initial training weights, and the training process and the steps aresummarized.(5) In the MATLAB environment, the application isPSO-PIDNN decouplingalgorithm simulation of the combustion system. By comparison with the separatePIDNN training to prove the proposed algorithm is a good solution to optimizenetwork weights optimization problem. PIDNN decoupling performance has beengreatly improved, and expanded its range of applications.