| The main steam temperature and reheated steam temperature of a thermal power plant are important indicators reflecting the unit’s safe,stable and economic operation.In the thermal control system,the steam temperature control system is very difficult due to the problems of large overshoot and slow response speed.At present,the control method of the steam temperature control system is based on simple conventional PID partial local control.This control method places high requirements on the operating personnel.With the development of intelligent algorithms,neural networks and optimization algorithms have also begun to be applied to the production and operation of thermal power plants.Based on a 1000 MW power plant unit steam temperature control system as the object,determine the steam temperature,main steam flow,main steam pressure,water flow reduction and reheat superheater steam temperature and reheat steam pressure,reheater warm water flow reduction and swinging angle of burner,superheater gas baffle opening,reheater flue gas damper opening,condenser circulating water inlet temperature,blower inlet air temperature 12 parameters as input,to consider side of boiler and turbine side heat consumption of equivalent heat consumption as output.The actual historical data were used to determine the optimal parameters of the neural network,that is,the 4-layer BP neural network structure was adopted,the number of nodes of the hidden layer in the middle two layers was 20,and the learning rate between each layer was 0.02,0.06 and0.02,respectively.The bat algorithm,gravity search algorithm,particle swarm algorithm and genetic algorithm were used to optimize the adjustable parameters of the steam temperature control system of the thermal power plant.Correct each optimization algorithm separately to determine the optimal parameters of each algorithm when the steam temperature control system of the thermal power plant is optimized.The comparison of the four algorithms under the same operating conditions shows that the genetic algorithm has the fastest convergence speed,and the gravity search algorithm and the bat algorithm have better global search capabilities.The gravitational search algorithm,genetic algorithm,particle swarm algorithm,and bat algorithm reduce the average equivalent heat consumption by70.21 k J/(k W·h),67.27 k J/(k W·h),60.89 k J/(k W·h),and 71.58 k J/(k W·h).The maximum equivalent heat consumption can be reduced to 91.82 k J/(k W·h),89.07 k J/(k W·h),86.93 k J/(k W·h),and 95.62 k J/(k W·h).The average reduced equivalent heat consumption value and the maximum reduced equivalent heat consumption value of the bat algorithm are higher than the other two algorithms.In order to further improve the optimization effect,this paper proposes an improved bat algorithm based on the compass operator by referring to the pigeon homing behavior in the pigeon optimization algorithm.The simulation experiments of 6 kinds of classic complex test functions and two-sided t-test verify that the algorithm has the characteristics of fast convergence,avoidance of local optimality,and strong robustness.Experiments on the above conditions also show that the bat algorithm based on the improved compass operator reduces the average equivalent heat consumption value and the maximum equivalent heat consumption value reduction are 72.73.89 k J/(k W·h)and 96.21 k J/(k W·h).Its average reduced equivalent heat consumption value and maximum reduced equivalent heat consumption value are higher than other intelligent algorithms. |
