In order to solve the problem of uneven ammonia injection in selective catalytic reduction (SCR) de-NOx system of coal-fired power plant, a precise ammonia control optimization method was proposed by combining the intelligent feed-forward cascade control technology with the partition control technology. In the intelligent feed-forward process, the improved one-dimensional convolution neural network model with higher prediction accuracy was used to predict the NOx mass concentration at the inlet of the SCR reactor, which was used as the intelligent feed-forward control signal to realize the fast and accurate control of the NOx mass concentration under the whole working condition of the unit. In the partition control process, the single sub area and multi-point synchronous sampling measurement technology and equalization control strategy were combined to realize the precise control of ammonia injection in each partition. The proposed optimization method has been applied to several coal-fired power plants. The operation results show that the precise control of ammonia injection amount is realized under the whole working conditions of the unit. The NOx mass concentration at the outlet section of SCR reactor is uniformly distributed. The ammonia consumption is obviously reduced and the economic benefit is significant.
