基于惯性环节的光伏组件实时热模型
REAL-TIME THERMAL MODEL FOR PHOTOVOLTAIC MODULE BASED ON INERTIAL ELEMENT
汪航宇 1徐梓彭 1田怡博 1张衡 1高丹 1陈海平1
作者信息
- 1. 华北电力大学能源动力与机械工程学院,北京 102206
- 折叠
摘要
提出一种基于一阶惯性环节的光伏组件温度的实时计算方法,首先,对光伏组件进行传热特性分析,基于一维非稳态导热分析解,推导基于一阶惯性环节的光伏组件温度简化计算模型;然后,使用遗传算法与拟牛顿法串行优化方法,通过数据驱动方式快速确定模型中的参数;最后,使用该文提出的模型,基于BP、LSTM的温度预测模型和传统经验公式对某光伏场站的组件温度进行分析和预测.对比结果表明:该方法表现出良好的预测精度,均方根误差＜2℃,且部署模型所需的计算规模更小,运算速度可达神经网络的10倍以上,方便应用于实际控制系统中,且与神经网络方法相比更具可解释性,可作为一种实时计算光伏组件温度的有效方法.
Abstract
A real-time computing approach for photovoltaic module temperature via first-order inertial elementsis proposed.First,the heat transmission characteristics of photovoltaic modules are comprehensively analyzed,leading to the derivation of a simplified computation model for photovoltaic module temperature based on first-order inertial elements,utilizing solutions from one-dimensional unsteady heat conduction analysis.Next,employing a genetic algorithm and the Quasi-Newton method in serial optimization,the model parameters are expeditiously identified using a data-centric methodology.Ultimately,adopting the proposed model,a temperature prediction model founded on BP and LSTM and orthodox empirical formulas,the component temperatures of a photovoltaic station are studied and predicted.Comparative results demonstrate that this approach yields excellent predictive precision,boasting an root mean square error less than 2℃,and necessitates a more modest computational scale for model deployment,with an operational pace surpassing neural networks by tenfold,thereby facilitating practical control system application.When juxtaposed with neural network methodologies,it is superiorly interpretable and serves as an efficacious method for real-time assessment of photovoltaic module temperature.
关键词
光伏组件/温度/预测/实时热模型/一阶惯性环节Key words
photovoltaic modules/temperature/prediction/real-time thermal model/first-order inertial element引用本文复制引用
基金项目
保定市科技"揭榜挂帅"项目(2021创004)
华北电力大学中央高校基本科研业务费专项(2023MS028)
出版年
2024
NETL
NSTL
万方数据