Analysis Of Dynamic Characteristics Of Boiler Thermal Parameters Based On Fractal And Time Series Theory And Its Application In Control

Energy has an important strategic position in the national economy. The boiler, one of the most important equipment in the thermal power plant, consumes large amounts of energy. Therefore, it is significant and necessary to conduct investigations in boiler control operation theory and optimization control research. The thermal parameters of boiler such as drum level main steam temperature, main steam pressure are characterized by non-uniform in space, non-equilibrium in time, and are controlled by a variety of non-linear, non-steady-state coupling effects. A more profound and comprehensive research on dynamic characteristics can be conducted from the respective of nonlinear dynamics (fractal, etc.). In this paper, based upon the practical operation data of boiler, the variation characteristics of thermal parameters such as drum water level, main steam temperature and main steam pressure etc. are analyzed by using fractal method. After that, the parameter trend model is developed according to time window partition and dynamic time series regression modeling. Then, this model is applied in multi-models switch predictive control of drum water level and other parameters. Aiming at the three aspects, this research is conducted as follow.Firstly, Aiming at the existence of boiler operation parameters in the nonlinear, time-varying and uncertainty problems, based on the time series R/S analysis method of the fractal theory, the variations of Hurst indexes and fractal dimension dynamic on main operating parameters, such as steam drum water level, main steam temperature, main steam pressure etc. under different load are studied. Hurst index and the time series fractal dimension represent the correlation and stability of the thermal parameters variation during boiler operation. The fractal characteristics can be found in the time series changes of drum water level, the main steam temperature and main steam pressure, the obvious volatility or stability can be reflected in fractal dimensions due to the difference among different boiler load, the fractal dimension is smaller in higher load. In addition, the time series R/S analysis method based on the theory of fractal are developed for distinguishing the false water level phenomenon which is common in boiler operation. It can be concluded that the variation of drum water level fractal dimension is obvious when the load changes suddenly, while it is stable before the changes. Hence, the changes of the load and the drum water level fractal dimension can be used to determine if the changes in water level are false, and a correct control strategy can be enact to guide the structure adjustment. The load fluctuation can be determined by the time series fractal dimension, if the fractal dimension Dτ suddenly changes, it means the load changes either, therefore, this phenomenon provides the theoretical foundation for models shift according to relative load.Secondly, in order to solve the problems of uncertainty and time-varying in thermal parameters, the fluctuation phenomenon that the thermal parameters deviate from the given reference value is analyzed, and the definitions of the deviation fractal dimension (Dj), the ascending fractal dimension (Djs) and the descending fractal dimension (DjX), which are used to determine thermal parameters are rising, falling or fluctuating range are presented. With comprehensive utilization of the characteristics of deviation fractal dimension and Hurst index, an innovative time series time windows partition algorithm based on D+H boiler dynamic data is introduced in this paper and proved by selecting the boiler operating data. That can be used to distinguish patterns:rising, falling or fluctuating according to the thermal parameters under different load condition in the process of boiler operation, and helps to divide suitable time window to complete dynamic time series regression modeling.Thirdly, focusing on inter-coupling among the thermal parameters in boiler operation, from the perspectives of time series of thermal parameters such as drum level, main steam temperature, main steam pressure in the process of boiler operation, three kinds of trend models are analyzed at different stage of load. Also, the time window modeling method based on fractal time series is proposed by using time series analysis dynamic regression modeling, the characteristics of the changes of thermal parameters such as drum level, main steam temperature, main steam pressure and other are analyzed and the time series trend models under different boiler load are established. The simulations verify the root mean square error of the model is small so that the requirements of the system dynamic characteristics identification is met, and lay the foundation for follow-up studies of boiler multi-model predictive control method.Finally, In accordance with the characteristics of nonlinear, variability and uncertainty in the operation of boiler, as well as the problems that the quality cannot be guaranteed by most of traditional PID system so far, this paper proposed multi-fractal model switching predictive control strategy based on a sliding time window modeling. Based on the five time window models concluded from historical data of boiler operation, the new data are captured by the method of slide time window and used to match models. And on that basis, the models will be switched timely in order to adapt to the dynamic variation and achieve predictive control. This paper designed the multi-model switching predictive control strategy for different characteristics of drum level system, the main steam temperature system and the main steam pressure system. Computer simulation and research of practical application show that the control strategy is superior to both of the traditional PID control and the predictive function control which based on a fixed model and greatly improves the robustness and stability of the system.