摘要
针对化工数据多为高维数据,而粒子群算法对求解高维优化问题易陷局部极值,提出将共轭方向法与粒子群算法相结合处理高维数据。当粒子群算法迭代了一定步数而陷入局部极值并得局部最优解时,以为初值,用共轭方向法对其求解,利用粒子群算法对低维优化问题的有效性,将得新的更优的当前最优解,从而使算法跳出局部极值;在新极值的条件下,又用粒子群算法对原问题求解,如此反复直至结束。通过经典的测试函数对其测试,结果表明这一尝试是有效的。最后将算法用于SO2催化氧化反应动力学模型的非线性参数估计,获得满意效果。
Aimed at the data that we get in chemical industry most being high dimensional, and particle swarm optimization （PSO） being easily trapped into local minima value for high dimensional function, a method conjugate direction particle swarm optimization （CDPSO）, which combined conjugate direction method with PSO, is proposed to process high-dimensional data. To one optimization problem, after PSO had run some iterations and trapped into local minima and got local optimal solution x＊, conjugate direction method with x＊ as an initial guess is applied to optimize the problem. By the effectiveness of PSO for low-dimensional function optimization, it would get a better new local optimal solution x＊＊, so this tactics helped PSO to overcome local minima. Under x＊＊ is present optimal solution; PSO is used for the high-dimensional function optimization again, if it trapped into local minima value again, then conjugate direction method is applied again, running in this way until termination. Experimental results on benchmark functions show that the proposed tactics is efficient. At last the algorithm is applied to nonlinear parameter estimation of burning anteiso-dynamics model of sulfur dioxide acted on by caesium-rubidium-vanadium low temperature sulfur acid catalyst and got satisfying results.
出处
《计算机工程与设计》
CSCD
北大核心
2007年第6期1241-1243,1254,共4页
Computer Engineering and Design
基金
国家自然科学基金项目(20276063)
关键词
粒子群优化算法
共轭方向法
高维函数优化
非线性参数估计
数据处理
particle swarm optimization
conjugate direction method
high-dimensional function optimization
nonlinear parameter estimation
data processing
