基于风格-支持向量回归的分布式光伏系统功率优化方法研究

A photovoltaic array exhibit various “styles” of output characteristics under different environmental conditions. How to improve the energy conversion efficiency in complicated environment is a research hotspot and difficulty in the field of photovoltaic applications. Most of the existing power optimization methods perform maximum power point tracking based on the sampled data. However, these methods usually ignore the “style”differences on the output characteristics and mix the data with various information, which causes a loss of precision. This project focuses on the optimization problem of a distributed photovoltaic system (DPS) under complicated environmental conditions. Its main work is divided into three parts: (1) quantitatively analyzing the output characteristics of a DPS under partial shading conditions; studying transformation method of nonlinear stylistic normalization which converts the “multi-style” characteristic data to independent identically distributed “single style” data (2) constructing a field-support vector regression based power optimization model, in which “new style” data are learned by a transfer learning algorithm; (3) developing an environment adaptation photovoltaic controller by using model algorithmic control. Research Achievements are of great significance to introduce the field - support vector regression theory and reveal the internal relationship between the environmental factors and the locus of maximum power points. The objective of this project is to further improve the energy conversion efficiency of photovoltaic systems, and provide design ideas for DPSs.

不同环境下光伏阵列呈不同“风格”输出特性，如何提高复杂环境下系统能源转换效率是光伏应用领域的研究热点与难点。现有功率优化方法多依据采样数据进行最大功率点跟踪控制，但常因忽略输出特性“风格”差异而混合数据信息，影响优化效果。本项目拟分析分布式光伏系统的复杂环境特性，围绕功率优化问题开展以下研究：（1）定量分析阴影环境下系统输出特性，研究非线性风格归一化变换方法，使“多风格”特性数据转换为满足独立同分布的“单风格”数据；（2）寻求面向 “新风格”数据的迁移学习方法，构建基于风格-支持向量回归框架的功率优化模型，以精准预测最大功率点；（3）建立模型算法控制机制，构造环境自适应光伏控制器，实现实时高效的系统功率优化。本项目的研究对揭示环境因素与最大功率点的内在联系，阐明风格-支持向量回归理论有重要意义，有望进一步改善光伏系统能源转换效率，为分布式光伏系统提供设计思路。

提高复杂环境下光伏系统能源转换效率一直以来是光伏应用领域的研究热点与难点。现有功率优化方法多依据采样数据进行最大功率点跟踪控制，但常因忽略输出特性“风格”差异而混合数据信息，不能达到理想的功率控制效果。本项目在分析分布式光伏系统的环境特性基础上，围绕功率优化问题开展以下研究：（1）定量分析阴影环境下系统输出特性，研究非线性风格归一化变换方法，使“多风格”特性数据转换为满足独立同分布的“单风格”数据；（2）寻求面向 “新风格”数据的数据学习方法，构建基于风格-支持向量回归框架的功率优化模型，以精准预测最大功率点；（3）建立模型算法控制机制，构造环境自适应光伏控制器，实现实时高效的系统功率优化。.项目组通过研究阴影遮挡下光伏系统电气特性，定义了面向分布式光伏系统的环境定量分析参数，研究了局部阴影检测方法，发现通过光伏系统电气特性可较精准地评估环境参数；依据不同阴影遮挡条件下光伏功率曲线呈现不同“风格”曲线，建立了究非线性风格归一化变换方法，通过去风格处理可精确地预测最大功率点位置；仿真实验与硬件实验均已验证通过模型指引的最大功率点跟踪方法可有效提高光伏系统输出功率。研究方法可应用于光伏控制器设计，有较好的应用前景。