Comparing a Graph-Grammar Approach to Genetic Algorithms for Computational Synthesis of PV Arrays

Abstract

The research presented in this paper compares two different computational approaches to topology synthesis: graph grammars (GG) and a genetic algorithm (GA). Both are applied to the domain of solar panel networks, which are configured to maximize power under a variety of sunlight conditions. Networks mostly connected in series create high voltages but are susceptible to shading. On the other hand, highly parallel topologies tend to create too low voltages. Thus, there is an opportunity to optimize photovoltaic (PV) array topologies given sunlight conditions. The paper describes methods to model the PV topology optimization problem using both a synthesis and an optimization approach. The focus is on computationally representing the design problem as well as the problem specific knowledge. Implementations of two different grammar rule sets in the GG are compared to four different encodings of the design problem in the GA. Results are shown for both approaches and a discussion of their advantages and drawbacks is given.

A Two-Tiered Grammatical Approach for Agent-Based Computational Design

“…That study both successfully demonstrated the ability of a graph grammar to be used in design optimization and suggested that its implementation allows for complete search within a design space. Later works utilized graph grammar-based design in gearbox design synthesis [25,29], passive dynamic brachiating robots [23], and optimization of photovoltaic arrays [30], among others.…”