| 引用本文: | 裴婷婷,张方正,陈 伟,等.融合退化特征与随机效应逆高斯过程的光伏组件寿命预测方法[J].电力系统保护与控制,2026,54(08):116-128. |
| PEI Tingting,ZHANG Fangzheng,CHEN Wei,et al.Lifetime prediction method for photovoltaic modules integrating degradation characteristics and random effects inverse Gaussian process[J].Power System Protection and Control,2026,54(08):116-128 |
|
| 摘要: |
| 针对光伏组件因环境因素、工艺或安装细节不同导致光伏组件退化过程的随机性、非线性和个体差异性，提出融合退化特征与随机效应逆高斯过程的光伏组件寿命预测方法。首先，构建基于随机效应逆高斯过程的光伏组件功率退化模型，刻画组件退化过程中的非线性和随机性特征。其次，根据光伏组件的退化特征，基于贝叶斯的马尔科夫链蒙特卡洛方法对退化模型参数进行估计。然后，通过蒙特卡洛方法从后验分布中提取参数组合，模拟光伏组件随机退化路径，记录每一条退化路径首次达到失效阈值的时间，获取光伏组件寿命分布与可靠性函数。最后，仿真结果表明，与随机非线性伽马过程和指数扩散过程的寿命预测方法相比，所提模型寿命预测结果的平均相对误差低至 2.8%，最大相对误差仅为 6.7%，验证了所提方法的有效性。 |
| 关键词: 光伏组件 寿命预测 退化特征 随机效应 逆高斯过程 |
| DOI：10.19783/j.cnki.pspc.251281 |
| 分类号: |
| 基金项目:国家自然科学基金项目 (51767017)；甘肃省联合科研基金重大项目 (25JRRA1143)；甘肃省兰州理工大学青年教师学科交叉研究培育项目 (LUTXKJC-25001) |
|
| Lifetime prediction method for photovoltaic modules integrating degradation characteristics and random effects inverse Gaussian process |
|
PEI Tingting, ZHANG Fangzheng, CHEN Wei, WU Yang
|
|
College of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou 730050, China
|
| Abstract: |
| To address the randomness, nonlinearity, and individual variability in the degradation process of photovoltaic (PV) modules caused by environmental factors, manufacturing processes, and installation conditions, a lifetime prediction method for PV modules integrating degradation characteristics and random effects inverse Gaussian process is proposed. First, a power degradation model for PV modules based on the random effects inverse Gaussian process is established to characterize the nonlinear and stochastic features of the module degradation process. Second, according to the degradation characteristics of PV modules, parameter estimation of the degradation model is performed using the Bayesian Markov Chain Monte Carlo method. Then, parameter sets are extracted from the posterior distribution via the Monte Carlo method to simulate the stochastic degradation paths of PV modules. The time when each degradation path first reaches the failure threshold is recorded to obtain the lifetime distribution and reliability function of the PV modules. Finally, simulation results demonstrate that, compared with lifetime prediction methods based on random nonlinear Gamma processes and exponential diffusion processes, the proposed model achieves a mean relative error as low as 2.8% and a maximum relative error of only 6.7%, demonstrating its effectiveness. |
| Key words: photovoltaic modules lifetime prediction degradation characteristics random effects inverse Gaussian process |
