Data-driven based multi-objective combustion optimization covering static and states

In order to reduce pollutant emissions and improve energy efficiency under the working conditions of static and dynamic states in the coal-fired power plants, data-driven based multi-objective combustion optimization (MOCO), which is around the implementation of air flow adjustment, is proposed to consider NOx emissions and boiler efficiency simultaneously. The strategy of combining kernel extreme learning machine (KELM) model with multi-objective evolutionary algorithm based on decomposition (MOEA/D) is presented to regulate the air distribution mode under static state. The double optimization is proposed to control the air flow under dynamic state. The first optimization is to employ auto-regressive moving-average with exogenous inputs (ARMAX) model and particle swarm optimization algorithm to optimize the parameters of PI controller. The second optimization is to use ARMAX models and tendentious guidance MOEA/D (TG-MOEA/D) to optimize the setpoint of air flow. The experiments showed that the KELM + MOEA/D strategy could further improve the effect of MOCO under static state, and the double optimization had decreased NOx emissions by an average of 12.18 % with keeping the boiler efficiency at a desired level under dynamic state. Moreover, data-driven based MOCO owns a high real-time performance, so that it is suitable for online application.