| The thermal power generating units in China mainly evaluate the control performance of the analog quantity control system through periodic disturbance tests.In the context of the current thermal power unit participating in flexible peak shaving,the thermal power unit control system is required to have better control performance.The periodic disturbance test method has been unable to meet the requirements of the performance evaluation of the control system.It is of great significance to carry out the performance evaluation of the full control loop based on the online operation data.The research object is the analog control system of a 660 MW coalfired unit.The purpose is to study control performance evaluation technology.The main contents are as follows:First,for the deterministic index,an online algorithm for the basic evaluation index of the step test and a dimensionless error integration index index suitable for the change of the set value curve are proposed.Four step tests of #2 high plus water level control were selected to calculate the basic indicators,The maximum overshoot is in the range of 8-12 mm,The smaller the steady state judgment criterion,the longer the stability time.The air supply control data is selected to calculate the dimensionless IAE index.The calculation results of the three experiments are between 0.57-0.82,which avoids the rapid increase of the IAE calculation results with the increase of the sample size.The calculation results accurately capture the relevant change characteristics of the data.Secondly,for the randomness index,the minimum variance index calculation method that preferably uses period and sample size under non-stationary data is proposed.An index of outof-tolerance performance factors based on out-of-tolerance is also proposed.Taking the secondary superheated steam temperature of a thermal power unit as an object.The two calculation results of the randomness index effectively distinguish the different control effects.When the allowable error thresholds are taken as ±3℃,±4℃ and ±5℃,the out-oftolerance performance factors are 0.57,0.80 and 0.93,respectively.The results show that the larger the allowable error,the better the outlier performance factor index.Furthermore,a parameter optimization method and feedforward strategy optimization method are proposed.Among them,the parameter optimization method is based on recursive least squares model identification and particle swarm optimization algorithm,and the previous performance index is the optimization goal.Feedforward strategy optimization mainly compresses the nearest neighbors to make the data distribution more uniform,and uses genetic algorithms to optimize the parameters.Taking the unit #2 high plus water level control as the object,the simulation experiment results show that the error band and variance are reduced by more than 50%,and the performance index has been significantly improved.In the improvement of the feedforward strategy,the overshoot and decay rate decreased by 100%,and the variance,stability time and other indicators decreased by about 50%.The results show that the overall deviation becomes significantly smaller after optimization,effectively solving the problem of large water level fluctuations.Finally,the software platform was developed based on the data interface and computing resources provided by a power company.The online operation of the thermal power unit control system performance evaluation and optimization is realized. |
