Research on reducing pollutant, improving efficiency and enhancing running safety for 1000 MW coal-fired boiler based on data-driven evolutionary optimization and online retrieval method

Abstract

This article adopts data-driven evolutionary optimization and online retrieval method to generate the boiler online combustion decisions and improve the boiler working performance. Improved sparrow search algorithm-based least squares support vector machine (ISSA-LSSVM) is utilized to develop the boiler's static mathematical model with self-adaptive capability under steady-load operating condition at first. And then improved sparrow search algorithm and long short-term memory neural networks (ISSA-LSTM) are combined to construct the dynamical combustion model for the boiler with self-adaptive capability under variable-load running condition. Whereafter, improved strength pareto evolutionary algorithm-II (ISPEA-II), future dynamic time-steps prediction models (FDTSP) and Bollinger Band-based safety assessment technique (BBSAT) are applied to obtain a number of combustion decisions owning better working state, higher economy and lower pollutant discharge offline. At last, safety assessment, mutation operation and the determination principle of the unique similarity case are introduced to the online retrieval method to generate the boiler combustion decisions in time. To illustrate the usability of proposed online optimization approach, several different on-line combustion optimization methods are applied in a practical online optimization process. The results indicated that based on proposed optimization method, the boiler thermal efficiency was improved by 0.210% and the NOx emission was reduced by 32.132 mg/m³ and the Bollinger Band, reflecting the fluctuation characteristic of wall temperature, was reduced from 32.685 to 10.249, simultaneously. Consequently, proposed on-line combustion optimization method of boiler is effective and it can realize the on-line combustion optimization of boiler.