Research On Neural Network Based Superheated Steam Temperature Control Algorithms And Engineering Application

Large-capacity coal-fired power units above 600 MW have become the main units in the development of China’s power industry.As an important parameter to measure the operation level of the unit,the superheated steam temperature(SST)of the boiler will pose a threat to the safety and economy of the unit if it is too high or too low,so the stable control of SST is an important factor to realize efficient and safe operation of the boiler unit.Water spray cooling is a common means of SST control.In order to achieve good control effect,cascade PID control is often adopted because of the complex boiler structure,huge system,strong non-linearity,large delay,large inertia and so on.At present,large units generally participate in power grid primary frequency regulation and automatic generation control(AGC),which are often in fast,deep and frequent variable load conditions.In order to meet the assessment requirements of the "two detailed rules",the regulations of coal,feedwater,air and other subsystems of the unit have been accelerated,and the coal quality is variable,so that the existing cascade SST control is often not suitable for this change,resulting in large fluctuations of SST,often deviating greatly from its set value.In order to keep the steam temperature within the limits,the operators need to frequently interfere with the steam temperature setting value(or bias)of the water spray cooling system,which increases their working intensity.It is very urgent for the power plant to optimize the water-spray cooling SST control system.For this reason,this paper combines the traditional PID control technology with the artificial neural network(ANN)inverse control method,and proposes two kinds of superheater spray cooling intelligent control strategies: one is the neural network feedforward inverse control and PID compensation control strategy,and the other is the cascade control strategy in which the outer loop is neural network inverse control and the inner loop is PID.For the two algorithms,detailed control simulation experiments are carried out by using the full-scope simulator of 600-MW supercritical thermal power unit to verify the effectiveness of the two methods.On this basis,working for two actual engineering projects,the cascade intelligent control strategy,of which outer loop neural network adopts inverse control and inner loop adopts PID control,is applied to a 600-MW supercritical unit and a 600-MW subcritical unit respectively,in order to realize real-time optimal control of water spray desuperheating system.The practical application results show that after adopting the intelligent control strategy,the control quality of SST under dynamic variable working conditions is effectively improved,and the steam temperature setting tracking performance and anti-interference performance are greatly improved.it effectively reduces the frequent intervention to the steam temperature setting value and reduces the operators’ working intensity.The work of this paper has certain reference significance for attempting application of neural network intelligent control technology in actual process control of coal-fired power plants.