Intelligent Control for Maximum Power Point Tracking of Photovoltaic Power Generation Based on Local Reinforcement Learning Strategy Improving Newton-Raphson-Based Optimizer
Purposes Aiming at the maximum power point tracking (MPPT) problem of photovoltaic power generation, a strong control method based on intelligent optimization algorithm is proposed. Methods The Newton-Raphson-based optimizer discussed in this paper was mainly divided into Newton-Raphson search rules and trap avoidance operators, which were applied to the maximum power tracking problem of photovoltaic power generation after being improved by the partial reinforcement learning strategy and variable spiral search strategy. The simulation was carried out in three experimental environments. Three parameter changes were set in each simulation environment to change the maximum power generation of photovoltaic modules. The improved Newton-Raphson-based optimizer (INRBO) was compared with the traditional and common maximum power control technologies in the simulation results. Results The results show that there are excellent global convergence ability, convergence speed performance and robustness in INRBO for MPPT.
光伏发电作为国家大力提倡的可再生能源发电方式,具有低成本和高可普及性的特点。影响光伏发电能力的因素有很多,如光照强度(辐照度)、温度、环境湿度、组件质量等。其中最主要的因素就是辐照度和温度。由于地球自转等因素的影响,光伏发电组件接收的辐照度会在一天内发生变化,其工作温度也会随之改变,光伏发出的光生电流也相应变化[1]。此时光伏组件输出的电压-功率特性曲线呈现多峰值的现象,其中只有一个是全局最大功率点(Global Maximum Power Point, GMPP)。在大规模分布式发电中,为了在当时当地的条件下产生尽可能多的发电功率以创造最大的经济效益,最大功率点跟踪技术(Maximum Power Point Tracking, MPPT)应运而生[2]。光伏组件的最大功率点(MPP)位于电流-电压(I-V)特性曲线某一峰值的拐点处,该最大功率拐点对于每个组件是唯一的,并且随着周围或环境干扰而变化。因此通常运用各种方法跟踪实时的最大功率点,如扰动观察法(P&O)[3]、增量电导法(IncCond)[4]以及智能优化算法[5]。扰动观察法和增量电导法作为传统的MPPT控制方法,因其简单性和较低的计算需求而被广泛使用。然而,这些方法在部分阴影和快速变化的光照条件下,易于陷入局部最大功率点,从而导致能量损失。近年来,智能控制方法在MPPT中的应用得到了广泛研究,如模糊逻辑控制[6]、神经网络控制[7]和遗传算法[8]等。这些方法具有自适应性强、响应速度快等优势,但也存在计算复杂度高、实现难度大等问题。特别是在复杂环境下,智能控制方法能够更好地适应多变的光照条件,提高系统的整体性能。
a和b分别代表0到1之间的随机数,r1和r2表示1到Np 之间不相等的两个随机整数。公式(16)表示根据牛顿-拉夫逊方法构造x1 nt 的方法,公式(17)表示根据当前最优位置xb 构造x2 it 的方法。x1 it 和x2 it 分别是根据已知量构造的探索和开发参数,各自具有较强的全局探索性能和局部开发性能,在牛顿-拉夫逊搜索规则中,利用这两者的特性构建了公式(18-19)。其中,r2 表示0到1之间的随机数。
PandeyA K, SinghV, JainS.Maximum power point tracking algorithm based on fuzzy logic control using P-V and I-V characteristics for PV array[J].IEEE Transactions on Industry Applications,2023,59(4):4572-4583.
[7]
ZhangY, WangY J, ZhangY,et al.Photovoltaic fuzzy logical control mppt based on adaptive genetic simulated annealing algorithm optimized bp neural network[J].Processes,2022,10(7):1411.
[8]
EdaraD C, VanukuriLP, SistlaV,et al.Sentiment analysis and text categorization of cancer medical records with lstm[J].Journal of Ambient Intelligence and Humanized Computing,2023,14(5):5309-5325.
[9]
RazaA, AnsM, KheraA R,et al.Designing efficient materials for high-performance of non-fullerene organic solar cells through side-chain engineering on DBT-4F derivatives by non-fused-ring electron acceptors[J].Journal of Molecular Modeling,2024,30(6):190-198.
[10]
MudgalS, DubeyK P, GlasserZ,et al.Large optoelectronic chromatic dispersion in PN-type silicon photodiodes and photovoltaic cells[J].Optics Letters,2024,49(8):2185-2188.
RavichandranS, ManoharanP, PradeepJ.Newton-Raphson-based optimizer:A new population-based metaheuristic algorithm for continuous optimization problems[J].Engineering Applications of Artificial Intelligence,2024,134:108976.
[14]
IdaL M.Learning analytics as modes of anticipation:enacting time in actor-networks[J].Scandinavian Journal of Educational Research,2024,68(2):218-232.
[15]
AhmadT, KeyvanR, AminB,et al.Partial reinforcement optimizer:An evolutionary optimization algorithm[J].Expert Systems with Applications,2024,238:121740.
[16]
AJ H.The sequencing of trials during partial reinforcement affects subsequent extinction[J].Journal of Experimental Psychology:Animal Learning and Cognition,2023,50(1):1-10.
WuX, LiS, JiangX,et al.Information acquisition optimizer:a new efficient algorithm for solving numerical and constrained engineering optimization problems[J].Journal of Supercompoting,2024,80(18):25736-25791.