基于改进灰狼算法的并行极限学习机发动机排气温度预测策略

Engine Exhaust Temperature Prediction based on Improved Gray Wolf Algorithm and Parallel Extreme Learning Machine

摘要:

为精准预测在不同工况下发动机的排气温度,通过引入一种新型的并行极限学习机对发动机排气温度进行预测。同时提出一种改进的灰狼优化算法,对并行极限学习机中的隐层阈值和输入权值进行在线整定,提高并行学习机的预测精度和泛化能力。其次,对传统灰狼优化算法全局收敛精度低和在迭代后期容易过早收敛陷入局部最优的问题,通过立方混沌映射对全部灰狼个体的位置进行初始化进行改进,同时在灰狼算法进行位置更新的过程中,加入历史信息学习因子的搜索策略,使的算法在更新过程中不因过多依赖局部最优解的位置信息而早熟收敛,提高算法的收敛精度和收敛速度。实验数据显示,优化后的灰狼优化算法在求解过程中所得到的平均值、标准偏差以及最小值都是最低的,证明了改进后的算法具有较高的寻优精度,可以有效地预测发动机的排气温度。

Abstract:

In order to accurately predict the engine exhaust temperature under different operating conditions, a new paral-lel Extreme learning machine was introduced to predict the engine exhaust temperature. At the same time, an improved grey wolf optimization algorithm is proposed to adjust the hidden layer threshold and input weight value in the parallel Extreme learning machine online, so as to improve the prediction accuracy and generaliza-tion ability of the parallel learning machine. Secondly, to address the issues of low global convergence accura-cy and the tendency to converge prematurely and fall into local optima in traditional grey wolf optimization algorithms, cubic chaos mapping is used to initialize the positions of all grey wolf individuals for improvement. At the same time, a search strategy of historical information learning factors is added to the position update process of the grey wolf algorithm, The algorithm does not prematurely converge due to too much dependence on the location information of the local optimal solution in the update process, and improves the convergence accuracy and Rate of convergence of the algorithm. The experimental data shows that the optimized Grey Wolf optimization algorithm obtains the lowest average, standard deviation, and minimum values during the solving process, proving that the improved algorithm has high optimization accuracy and can effectively predict the exhaust temperature of the engine.