A Study Of Advanced Optimization And Tuning Methods Of Controllers And Their Applications To Power Plant Control

The thermal power systems in power plants are the main source of energysupply. It is also the main unit of coal consumption. So the key energy-saving andemission reduction units must be the thermal power systems in power plants. It isnecessary to develop power plants toward the direction of operation with highefficiency and low pollution. As the modern thermal power systems go toward tothe large scale and complexity, especially the thermal power units go toward to thedevelopment of high parameter, large capacity, and high level of automation, itscontrol level plays a decisive role in safety and economic operation. At present, thelarge DCS (Distributed Control System) is widely used in large-scale coal-firedpower plant to monitor and control its whole equipments. However, the DCS stillkeeps stay only in the controlling, not in the optimum state of full use. The researchgoal of new method for thermal power system control is to further improve thesafety and economy of equipments operation. On the one hand, the processparameters of thermal power systems restrained in a certain range are an importantguarantee for the safe operation. On the other hand, the control precision in makingimprovement of some important operating parameters can also promote andupgrade the operation economy of thermal power systems indirectly. For theautomatic control of thermal power systems, the biggest difficulty comes fromtheirselves of being strong nonlinearity, strong coupling, and large delay.Meanwhile, the load fluctuation and fuel property changes increase the uncertaintyof the systems. This may result in the function deterioration and invalidity of thetraditional control method. Therefore, according to the above-mentionedcharacteristics and requirements of thermal power systems, in the first place, it isabsolutely imperative to develop more advanced thermal process control theory andtechnology. In the second place, based on the existing equipment and technology, itis very necessary to further optimize and improve the performance of the controlsystems.According to the characteristics of control systems in power plants, theoptimization design of PID controller is studied from two points of views,respectively. One is from single input single output (SISO) system and multi-inputmulti-output (MIMO) system, and another is from considering model of plants andconsidering unique input and output test data set. The main research contents andinnovative points include:1) A method for tuning the parameters of PID controllers via modified binarydifferential evolution algorithm (MBDE) for general SISO systems,especially self-regulating systems, is proposed. Firstly, the parametersstability region is predicted presumedly according to the parametersstability region theory. Then the MBDE algorithms is employed to tune theparameters of PID controllers combined with a certain performance indexaccording to the practical applications. The range of parameters to beoptimized is confirmed by the parameters stability region. In the end the method is applied to control the bed temperature of a circular fluidized bedand a non-minimum phase system, respectively. The simulation resultsillustrate its validity and effectiveness.2) A method is proposed to retune the parameters of the structure-knowncontrollers via VRFT based on data-driven when the system is operatingwith noise in closed loop. Firstly, the sensitivity function is identifiedbased on I/O data set to construct a new virtual I/O data set in open loop.Then the standard VRFT method in open loop can be employed to retunethe parameters of the controllers. The effectiveness is verified after theparameters of the controllers in a Benchmark problem is retuned using thismethod. And a general procedure for tuning parameters of PID controllerswith a standard type in closed-loop systems with noise based on VRFT isproposed. A case shows its validity and practicability.3) A framework is proposed to design the PID controllers for generalindustrial plants including thermal power plants with pure delay, whetherthey are self-regulating or not. These plants can be considered as the plantswith non-minimum phase zeros. And the standard VRFT method combinedwith MBDE algorithms can be used to design the controllers. This can beapplied in the level control of a boiler condensate water makeup systemwith a pump and a superheating steam temperature cascade control,respectively. The former is a non-self-regulating system and the latter isself-regulating system. The simulation results show its validity and wideapplicability.4) A direct Nyquist array (DNA) method for the design of PID controllers formultivariable boiler-turbine units with specifications of gain and phasemargins is proposed. The essential objective is to propose a method for thedesign and auto-tuning of both straightforward and robust PID controllersthat can be more easily implemented for the boiler-turbine units incoal-fired power generation plants in the same framework. For this purpose,the model of the original multi-input multi-output (MIMO) system is firstconverted into a diagonal/diagonal dominance matrix after the system isappropriately compensated. Then the PID controller design methods basedon gain and phase margins (GPM) for SISO system can be easily extendedto decoupled or quasi-decoupled MIMO system by using the directNyquist array (DNA) method and the linearization of arcsine function inthe range (0,1). A coordinated control system (CCS) is designed for aboiler-turbine unit (BTU) by this method and compared with an existingsystem via another method. The results illustrate its validity. In particular,the proposed method allows the user to specify the robustness and otherkey performances of the system through the gain and phase marginsspecifications.5) In general, the standard VRFT method may not be applied in commonMIMO systems for its restrained structure such as the noncommutativity ofmatrix operation in mathematics. It can only be used in completedecoupling system. However, the research shows that it can also be used inthe diagonal dominance systems in industrial fields. A coordinated controlsystem for a known model—BTU，is designed based on VRFT for MIMO system, and good results are achieved.6) A scheme for self-tuning the parameters of PID controllers is proposed andapplied to coal-fired thermal power plants in DCS without introducing thethird party system. By using the configuration software of the existingDCS, the on-line identification of the parameters of FOPDT models andIPDT models and the self-tuning of the parameters of PID controllers isrealized. A strong robustness area method is employed in identificationalgorithms which is improved by reasonable simplification to avoid therestraint of the configuration software function. Several typical methodsare chosen to tune the parameters of PID controllers which are put intopractice and compared with each other. Experimental results show that themethods have good quality even in noise environments.