摘要
对使用非线性微分-代数不等式方程的电力系统模型,采用免疫搜索算法非线性模型滚动预测控制.通过分级目标分解方法,根据每个预测时段上的控制性能要求,将全局多个控制目标分解为预测时段内的优化子目标,运用Pareto意义的子目标加权,集成为一个总目标函数.在搜索最优解中运用免疫算法,将具有多基因链结构的抗体来表达复杂优化问题的候选解,利用免疫算法的学习和记忆能力识别各预测时段内已求解的优化问题类型,用模式识别技术提取优良抗体的基因,预测未来时段内的最优解搜索过程估计出较好的初始解,以加快最优解搜索速度.将此方法和基于树搜索算法的非线性预测控制方法比较,通过一个6母线电力系统实例进行了仿真研究,结果表明:文中提出的算法改进具有更强的优化搜索能力和更好的实时性.
A novel immune search algorithm is proposed for a nonlinear model predictive scroll control scheme on the nonlinear differential-algebraic-inequality power system. A global control target function is integrated with gradational targeting decomposed global horizion control targets into sub-objectives optimized in receding prediction intervals via Pareto-type weighting functions according to controlling performance in each period of time. Using a multiple gene chain structure of antibodies to represent the solution candidates of the complicated optimization problem and employing pattern recognition techniques to extract gene patterns of better antibodies, similar antigen patterns are identified via learning and memorizing to create a better initial guess of solutions in order to accelerate the convergence of the optima searching procedure. The results indicate that the method proposed is optimal in search ability and real time effect.
出处
《江苏大学学报（自然科学版）》
EI
CAS
北大核心
2008年第1期56-60,共5页
Journal of Jiangsu University：Natural Science Edition
基金
江苏省高新技术项目(GB2004024)
江苏省教育厅博士基金创新技术项目(06KJB470015)
关键词
电力系统
电网
人工免疫算法
模型预测滚动控制
非线性系统
power system
power net
artificial immune algorithm
model predictive scroll control
nonlinear system
