Estudo exploratório da biblioteca Python Pandapower e sua utilização em conjunto com Reinforcement Learning para análise e otimização de sistemas de energia

• Main Author: Chella, Marcelo
• Publication Date: 2022
• Format: Bachelor thesis
• Language: por
• Source: Repositório Institucional da UTFPR (da Universidade Tecnológica Federal do Paraná (RIUT))
• Download full: http://repositorio.utfpr.edu.br/jspui/handle/1/31714
• topic: Aprendizado do computador; Inteligência artificial; Python (Linguagem de programação de computador); Machine learning; Artificial intelligence; Python (Computer program language); CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

Nowadays, most of the research and project in the energy field are based on software as a primary tool. The use of this type of approach makes it possible to research, simulate, study and easily change the parameters of a given system, enabling easy and less expensive research on the matter. Within this scope are the economic, environmental and engineering aspects of electrical networks. Following this model, the pandapower tool was developed to be a simulation tool for researchers and educators, easy to use and modify, keeping the created code simple to understand and modify. This toolbox is an open-source Python tool for modeling, analyzing and optimizing power systems with a high degree of automation, and also based on a tabular data structure from the Python pandas library. Applications for this tool are calculation of three-phase AC load flow, configuration/execution of system state estimation and AC load balancing (improving load distribution or resizing transformer points). Additionally, Reinforcement Learning (RL) algorithms can be used in tasks related to demand response analysis. This work aims to explore the applications of pandapower by modeling and calculating power flow on energy networks, as well as demonstrating the prediction of behavior of these networks in response to triggered events. Some Artificial Intelligence models based on Artificial Neural Networks (ANN) and Reinforcement Learning will be used for demonstrations of grid state prediction and exploration of photovoltaic generation scenarios in response to variable demand, and their results for the optimization of energy systems will be presented.