随机交叉-自学策略改进的教与学优化算法

针对非原点最优的复杂优化问题(最优解不在坐标原点)，提出了一种基于随机交叉-自学策略的教与学优化算法(teaching and learning optimization algorithm based on random crossover-self-study strategy, CSTLBO)。对标准教与学优化算法的“教阶段”和“学阶段”的空间扰动进行了几何解释，改进了原有的“教阶段”和“学阶段”，并引入随机交叉策略和“自学”策略来提高算法的全局寻优能力。通过使用20个Benchmark函数进行仿真，并与6种改进的教与学优化算法进行结果比较及Wilcoxon秩和检验分析，结果表明CSTLBO算法能有效避免陷入局部最优，具有良好的全局搜索能力，求解精度高，稳定性好。     

基于K-均值的“教”与“学”优化算法

在解决复杂多峰优化问题时,传统的"教"与"学"优化算法易于陷入局部搜索且优化效率较低.针对此问题,提出了一种基于K-均值的"教"与"学"优化改进算法,算法采用K-均值来降低种群规模,又针对"教"和"学"两个阶段进行相应改进,提高全局收敛速度;还加入了"变异"操作来避免算法陷入局部最优.实验对7个单峰值优化问题和2个有代表性的多峰值优化问题进行优化,并与手榴弹爆破算法和传统"教"与"学"优化算法进行比较,实验结果表明,该改进算法在单峰和多峰测试函数中,均能快速高效地寻得全局最优解,优于原始"教"与"学"优化算法.     

基于改进人工鱼群算法的支持向量机预测

由于参数的选择范围较大,在多个参数中进行盲目搜索最优参数的时间代价较大,且很难得到最优参数.为此,提出一种基于改进人工鱼群算法(AFSA)的支持向量机(SVM)预测算法.对AFSA进行改进,并使用改进算法优化SVM.实验结果表明,与遗传算法、粒子群优化算法和基本AFSA优化的支持向量机相比,该算法的均方误差降低为2.51×10-3,提高了预测精度.     

基于改进教学算法的无人机航路规划

针对传统教-学优化（TLBO）算法进行航路规划时收敛速度慢、容易陷入局部最优的问题,提出一种自适应交叉教-学优化（AC-TLBO）算法。首先,该算法令传统教-学优化（TLBO）算法的教学因子随着迭代次数而发生变化,提高算法的学习速度;其次,当算法可能要陷入局部最优时,加入一定的扰动,使算法尽可能地跳出局部最优;最后,为了进一步提升算法的收敛效果,在算法中引入遗传算法的交叉环节。利用传统教-学优化（TLBO）算法、自适应交叉教-学优化（AC-TLBO）算法和量子粒子群优化（QPSO）算法进行无人机航路规划,仿真结果表明,在10次规划中,自适应交叉教-学优化（AC-TLBO）算法有8次找到了全局最优路径,而传统教-学优化（TLBO）算法和量子粒子群优化（QPSO）算法分别只找到了2次和1次;而且自适应交叉教-学优化（AC-TLBO）算法的收敛速度高于另外两种算法。     

基于新型蛙跳算法的低碳柔性作业车间调度

针对低碳柔性作业车间调度问题(flexible job shop scheduling problem,FJSP),提出一种新型蛙跳算法(shuffled frog leaping algorithm,SFLA)以总碳排放最小化,该算法运用记忆保留搜索所得一定数量的最优解,并采取基于种群和记忆的种群划分方法,应用新的搜索策略如全局搜索与局部搜索的协调优化以实现模因组内的搜索,取消种群重组使算法得到简化.采用混合遗传算法和教–学优化算法作为对比算法,大量仿真对比实验验证了SFLA对于求解低碳FJSP具有较强的搜索能力和竞争力.     

嵌入局部一维搜索技术的混合粒子群优化算法*

通过将粒子群优化算法(PSO)与经典局部一维搜索技术相结合,提出一种嵌入局部一维搜索技术的混合粒子群优化算法(LLS-PSO)。该算法在基本粒子群优化算法中引入一维搜索技术,选取最优粒子进行局部一维搜索,增强了在最优点附近的局部搜索能力,以加快算法的收敛速度。对三个经典复杂优化问题进行数值实验,并与基本PSO算法进行比较。实验分析和结果表明,LLS-PSO具有更好的优化性能。     

改进蚁群算法在支持向量机中的应用

提出一种改进的蚁群算法(ACA)来优化支持向量机(SVM)训练参数.该改进算法建立于每只蚂蚁只根据参数β在其前次迭代的最优解附近搜索,可快速减少搜索范围.参数β的提出可以保证蚁群快速地达到最优解.仿真结果表明:使用该方法优化SVM参数可有效避免陷入局部极值,提高收敛速度.     

一种带有附加记忆策略的改进教与学优化算法

教与学优化算法是一种模拟班级教学现象的新型群体智能优化算法,算法参数简单,收敛速度快,已经在函数优化、工程计算等领域取得广泛应用。但是算法后期容易陷入局部收敛,为此提出了一种带有附加记忆策略的教与学优化(MTLBO)算法。该算法首先在教学阶段增加教师记忆策略,学生的历史记忆知识与教师历史教学能力对提高班级的整体教学水平具有重要的作用,在每次更新学习者的同时考虑教师上一代的最优值和当代的最优值,有效增强算法局部搜索能力;在学习阶段增加个体向最优个体和随机个体学习策略,多个学生互相学习,充分利用班级内的知识信息,从而增强了算法的全局搜索能力。采用具有不同特征的多个测试函数对算法进行仿真实验,并与基本TLBO算法和2种改进的TLBO算法进行对比分析,结果表明提出的MTLBO算法在获得较高的收敛精度和稳定性的同时还提高了收敛速度,有效避免算法局部收敛。     

基于EM算法和PSO算法的混合优化算法

针对粒子群优化算法容易陷入局部最优解并且存在过早收敛的问题,将类电磁机制算法中的吸引-排斥机制引入到粒子群优化算法中,提出一种类电磁机制算法和粒子群优化算法的混合优化算法(EMPSO).首先按照基本粒子群优化算法的寻优方式对各粒子进行更新,再利用类电磁机制中的吸引-排斥机制对个体最优粒子和群体最优粒子进行移动,最后通过几个标准测试函数进行了测试,并与标准粒子群算法(PSO)、免疫粒子群算法(IPSO)、混沌粒子群算法(CPSO)进行对比.测试结果表明,改进算法提高了全局搜索能力和熟练速度,改善了优化性能.     

多元优化算法的渐近性分析

本文提出了一种多元化智能个体分工明确、协同合作的超启发式智能优化算法—–多元优化算法.多元优化算法通过交替的全局、局部搜索迭代对解空间搜索以逐渐逼近全局最优解.搜索个体按照分工不同可以分为全局搜索个体(全局元)和局部搜索个体(局部元).全局元负责对整个解空间进行全局搜索以快速找到较优潜在解区域,局部搜索元负责对各个潜在解区域进行局部搜索以提高解的质量.该算法具有两个特点:分工明确的搜索策略不需要考虑均衡全局搜索和局部搜索,能够保证局部搜索能力的同时加强全局搜索以避免陷入局部最优解;全局、局部交替搜索保证了算法对全局最优解的渐近性.本文从理论上证明了算法的渐近性并且基于复杂多模态测试函数比较了几个优秀的进化算法.实验结果表明多元优化算法在渐近性方面优于其他几个比较的算法.     

Study of differential search algorithm based automatic generation control of an interconnected thermal-thermal system with governor dead-band

An attempt has been made to the effective application of a recently introduced, powerful optimization technique called differential search algorithm (DSA), for the first time to solve load frequency control (LFC) problem in power system. In this paper, initially, DSA optimized classical PI/PIDF controller is implemented to an identical two-area thermal-thermal power system and then the study is extended to two more realistic power systems which are widely used in the literature. To assess the usefulness of DSA, three enhanced competitive algorithms namely comprehensive learning particle swarm optimization (CLPSO), ensemble of mutation and crossover strategies and parameters in differential evolution (EPSDE), and success history based DE (SHADE) are studied in this paper. Moreover, the superiority of proposed DSA optimized PI/PID/PIDF controller is validated by an extensive comparative analysis with some recently published meta-heuristic algorithms such as firefly algorithm (FA), bacteria foraging optimization algorithm (BFOA), genetic algorithm (GA), craziness based particle swarm optimization (CRPSO), differential evolution (DE), teaching-learning based optimization (TLBO), particle swarm optimization (PSO), and quasi-oppositional harmony search algorithm (QOHSA). A case of robustness and sensitivity analysis has been performed for the concerned test system under parametric uncertainty and random load perturbation. Furthermore, to demonstrate the efficacy of proposed DSA, the system nonlinearities like reheater of the steam turbine and governor dead band are included in the system modeling. The extensive results presented in this article demonstrate that proposed DSA can effectively improve system dynamics and may be applied to real-time LFC problem.     

融合改进天牛须搜索的教与学优化算法

针对教与学优化算法易早熟,解精度低,甚至收敛于局部最优的问题,提出一种新的融合改进天牛须搜索的教与学优化算法.该算法利用Tent映射反向学习策略初始化种群,提升初始解质量.在"教"阶段,对教师个体执行天牛须搜索算法,增强教师教学水平,提高最优解的精确性.在"学"阶段,对学生个体进行混合变异,从而跳出局部最优,平衡算法的...     

阴影条件下基于迁移强化学习的光伏系统最大功率跟踪

在光伏系统中,光伏阵列往往会受到阴影条件(partial shading condition,PSC)的影响,造成光伏系统输出功率偏低以及功率-电压($P-V$)特性曲线出现多峰值的现象,从而导致常规最大功率跟踪(maximum power point tracking,MPPT)算法易陷入局部最优的问题.对此,设计一种基于迁移强化学习(transfer reinforcement learning,TRL)的MPPT算法.该算法将连续变量的动作空间分解为若干个小范围的子搜索空间,从而有效提高TRL的学习效率.同时,引入知识迁移,即将旧任务的最优知识矩阵应用到新任务中,进而大幅提高TRL的收敛速度.通过对3种算例的研究,即恒温变光照强度、变温变光照强度和香港实地测试,其仿真结果表明,与传统增量电导法(incremental conductance,INC)、遗传算法(genetic algorithm,GA)、粒子群优化(particle swarm optimization, PSO)算法、人工蜂群(artificial bee colony,ABC)算法、布谷鸟算法(cuckoo search algorithm,CSA)、教-学优化(teaching-learning based optimization,TLBO)算法以及Q学习算法相比,TRL能在PSC下实现最快速的全局最大功率跟踪,同时具有最小的功率波动.最后,基于dSpace的硬件在环实验验证了TRL的硬件可行性.     

C-V2X通信中资源分配与功率控制联合优化

在C-V2X通信中,Mode 4资源分配方式使用基于感知的半持续调度（SB-SPS）算法进行资源分配,但该算法以最大功率传输安全消息,在高密度交通流状态下会导致系统的可靠性下降。为对SB-SPS算法进行优化,提出一种基于深度强化学习的联合资源分配与功率控制算法。车辆在感知到信道后,为安全消息选择干扰最小的子信道,并根据信道状态自适应调整传输功率,通过与环境交互学习的方式求解最优的子信道选择方案和功率控制方案。仿真结果表明,与SB-SPS优化算法相比,该算法在高密度公路场景下分组接收率提高5%,有效提升了车间通信的可靠性。     

一种基于教与学的混合灰狼优化算法

针对灰狼优化算法(GWO)存在收敛精度不高、易陷入局部最优的不足,提出一种基于教与学的混合灰狼优化算法(HGWO).首先,采用佳点集理论进行种群初始化,提高初始种群的遍历性;其次,提出一种非线性控制参数策略,在迭代前期增加全局搜索能力,避免算法陷入局部最优,在迭代后期增加局部开发能力,提高收敛精度;最后,结合教与学算法(TLBO)和粒子群优化算法,修改原位置更新公式以优化算法搜索方式,从而提升算法的收敛性能.为验证HGWO算法的有效性,选取9种标准测试函数,将HGWO算法、GWO算法以及其他群体智能优化算法和其他改进GWO算法进行仿真实验.实验结果表明,所提出的HGWO算法性能优于GWO算法和其他群体智能优化算法,且在改进算法中具有一定优势.     

基于反馈的精英教学优化算法

精英教学优化算法（Elitist teaching-learning-based optimization,ETLBO）是一种基于实际班级教学过程的新型优化算法. 本文针对ETLBO算法寻优精度低、稳定性差的问题,提出了反馈精英教学优化算法（Feedback ETLBO）. 在ETLBO算法的基础上,通过在学生阶段之后加入反馈阶段,增加了学生的学习方式,保持学生的多样性特性,提高算法的全局搜索能力. 同时,反馈阶段是选举成绩较差的学生与教师交流,使成绩较差的学生快速向教师靠拢,使算法进行局部精细搜索,提高算法的寻优精度. 对6个无约束及5个约束标准函数的测试结果表明,FETLBO算法与其他算法相比在寻优精度和稳定性上更具优势. 最后将FETLBO算法应用于拉压弹簧优化设计问题及0-1背包问题,取得了满意结果.     

基于DQN的超密集网络能效资源管理

小基站的密集随机部署会产生严重干扰和较高能耗问题,为降低网络干扰、保证用户网络服务质量（QoS）并提高网络能效,构建一种基于深度强化学习（DRL）的资源分配和功率控制联合优化框架。综合考虑超密集异构网络中的同层干扰和跨层干扰,提出对频谱与功率资源联合控制能效以及用户QoS的联合优化问题。针对该联合优化问题的NP-Hard特性,提出基于DRL框架的资源分配和功率控制联合优化算法,并定义联合频谱和功率分配的状态、动作以及回报函数。利用强化学习、在线学习和深度神经网络线下训练对网络资源进行控制,从而找到最佳资源和功率控制策略。仿真结果表明,与枚举算法、Q-学习算法和两阶段算法相比,该算法可在保证用户QoS的同时有效提升网络能效。     

DG placement in radial distribution network by symbiotic organisms search algorithm for real power loss minimization

Incorporation of distributed generation (DG) in distribution network may reduce the network loss if DG of appropriate size is placed at proper strategic location. The current article presents determination of optimal size and location of DG in radial distribution network (RDN) for the reduction of network loss considering deterministic load demand and DG generation using symbiotic organisms search (SOS) algorithm. SOS algorithm is a meta-heuristic technique, inspired by the symbiotic relationship between different biological species. In this paper, optimal size and location of DG are obtained for two different RDNs (such as, 33-bus and 69-bus distribution networks). The obtained results, using the proposed SOS, are compared to the results offered by some other optimization algorithms like particle swarm optimization, teaching-learning based optimization, cuckoo search, artificial bee colony, gravitational search algorithm and stochastic fractal search. The comparison is done based on minimum loss of the distribution network as well as based on the convergence mobility of the fitness function offered by each of the comparative algorithms for both the networks under consideration. It is established that the proposed SOS algorithm offers better result as compared to other optimization algorithms under consideration. The results are also compared to the existing solution available in the literature.     

Energy management for multi-microgrid system based on model predictive control

To reduce the computation complexity of the optimization algorithm used in energy management of a multi-microgrid system, an energy optimization management method based on model predictive control is presented. The idea of decomposition and coordination is adopted to achieve the balance between power supply and user demand, and the power supply cost is minimized by coordinating surplus energy in the multi-microgrid system. The energy management model and energy optimization problem are established according to the power flow characteristics of microgrids. A dual decomposition approach is imposed to decompose the optimization problem into two parts, and a distributed predictive control algorithm based on global optimization is introduced to achieve the optimal solution by iteration and coordination. The proposed method has been verified by simulation, and simulation results show that the proposed method provides the demanded energy to consumers in real time, and improves renewable energy efficiency. In addition, the proposed algorithm has been compared with the particle swarm optimization (PSO) algorithm. The results show that compared with PSO, the proposed method has better performance, faster convergence, and significantly higher efficiency.     

A Case-Based Micro Interactive Genetic Algorithm (CBMIGA) for interactive learning and search: Methodology and application to groundwater monitoring design

Interactive optimization algorithms use real–time interaction to include decision maker preferences based on the subjective quality of evolving solutions. In water resources management problems where numerous qualitative criteria exist, use of such interactive optimization methods can facilitate in the search for comprehensive and meaningful solutions for the decision maker. The decision makers using such a system are, however, likely to go through their own learning process as they view new solutions and gain knowledge about the design space. This leads to temporal changes (nonstationarity) in their preferences that can impair the performance of interactive optimization algorithms. This paper proposes a new interactive optimization algorithm – Case-Based Micro Interactive Genetic Algorithm – that uses a case-based memory and case-based reasoning to manage the effects of nonstationarity in decision maker’s preferences within the search process without impairing the performance of the search algorithm. This paper focuses on exploring the advantages of such an approach within the domain of groundwater monitoring design, though it is applicable to many other problems. The methodology is tested under non-stationary preference conditions using simulated and real human decision makers, and it is also compared with a non-interactive genetic algorithm and a previous version of the interactive genetic algorithm.