Layout Optimization for a Large-Scale Grid-Connected Solar Power Plant

A solar power plant provides green electricity to the public via a power grid. As governments worldwide have pledged to reduce carbon emissions and achieve carbon neutrality, large-scale grid-connected solar power plants are booming. Developing such a plant requires significant investment, a large proportion of which covers construction costs. Such costs, together with the energy yield, critically depend on the plant’s layout. The layout planning of a solar power plant involves a series of complex optimization problems such as district partitioning, photovoltaic (PV) component location, and cable routing problems in a solar power plant. These problems have received limited attention in the literature and are highly challenging because they involve large-scale instances, complex design principles, and complicated physical constraints. Motivated by our collaborative projects with an electrical engineering company in China, this paper specifically focuses on the integrated location and routing (ILR) problem, which involves locating service ways, inverters, combiner boxes, and routing cables to connect them. We develop exact algorithms to effectively solve the ILR problem via a decomposition framework (leading to a variant of Benders decomposition (BD)), which is proven to produce an optimal solution. We also develop an exact branch-and-cut scheme to solve each subproblem in the decomposition framework by incorporating cutting planes and separation algorithms. Our solution approach is evaluated on 50 real-world data instances via extensive numerical experiments. Compared with the manual method based on greedy heuristics used in practice, our approach reduces the total cost by approximately 20%. Our decomposition method also achieves an average gap of 0.02% between the obtained lower and upper bounds, significantly smaller than the 16.08% gap achieved with the traditional BD. History: Accepted by David Alderson, Area Editor for Network Optimization: Algorithms & Applications. Funding: This work was partially supported by the Research Grants Council of Hong Kong [Grant 15501221], the National Natural Science Foundation Program of China [Grants 72122006, 72471100, and 72131008], Huazhong University of Science and Technology Double First-Class Funds for Humanities and Social Sciences (Digital Intelligence Decision Optimization Innovation Team), the Interdisciplinary Research Program of HUST [Grant 5003300129], and the Fundamental Research Funds for the Central Universities [Grant 2023WKFZZX101]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.0223 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2023.0223 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .