Control Performance Assessment Of Thermal Power Plant Based On Closed-loop Operational Data

As an important support for energy,thermal power plays an important role in industrial production and people’s daily life,thermal power unit control performance of the pros and cons of thermal power plant power generation and production safety has a direct impact.The control system’s parameters in the power plant are generally based on full load operation debugging,as the working conditions change,the lack of quantitative performance evaluation criteria.The existing performance assessment methods require some prior knowledge of the system model,or the use of input and output data for system modeling,the calculation process is more cumbersome and there is a model mismatch and other issues,so it is difficult to apply to large systems.Therefore,in order to improve the safety and economy of production,it is necessary to find a new assessment index for control performance of control system and apply it to online assessment of control performance,and find out the deterioration of control performance in time.Aiming at the above problems,a data-driven control performance assessment index based on the existing steady-state detection algorithm was presented in this paper.Firstly,the improved performance assessment index was given by two different representations of variance under steady-state data to measure the quality of control performance.And the rationality of the indicator was verified through simulation.Then the index was applied to the online assessment of the real-time performance of the control system by exponential weighted moving average method.The threshold for online assessment was determined using a kernel density estimation method.Through the introduction of sliding window,the online data performance assessment index was compared with the threshold,and then the assessment results were given.On the basis of this,the method was applied to the performance assessment of a 1000MW ultra supercritical unit.The results showed that the method was more accurate and had wider applicability than the existing methods.Finally,the load conditions of the unit were divided by the clustering method,and the alarm rates under different load conditions were counted.Combining the command complexity of each interval,tracking situation of active power to load command under different load conditions was evaluated.The results showed that at 850MW there was a point of deterioration of the condition which needed to be further optimized.