AI-Enhanced Energy Harvesting Materials For Self-Sustaining Civil Structures

Abstract

The combination of the artificial intelligence (AI) and the energy harvesting materials is a promising solution to the creation of self-sustaining civil architecture. This paper evaluates the use of AI models, Artificial Neural Network (ANN), Support Vector Regression (SVR), Random Forest Regression (RFR) and Genetic Algorithms (GA), that can be used to optimize and predict performance of the piezoelectric, thermoelectric and photovoltaic energy harvesting systems fixed in infrastructure. With a sample of 300 and varying environmental conditions (load, temperature, light intensity, and vibration frequency), AI model was trained to ensure the power output was high and with high accuracy. Of them, RFR recorded the best result with the following performance: R 2 score of 0.96, RMSE of 0.58 0 G and MAE of 0.39 0 G. GA also helped optimize structural parameters which gave a 15.8 percent increase in power output. The piezoelectric and thermoelectric systems registered the lowest average power output of 3.1 3 1 and 8.3 3 1 respectively whereas the average power output of photovoltaic systems was recorded as 11.3 3 1. As shown in the study, intelligence and flexibility of infrastructure enabled by AI-enhanced frameworks can accomplish much more than energy efficiency only, as intelligent and adaptive infrastructure can sustain embedded sensors and systems on their own. The results lead to long-term, AI-based solutions in smart city and Infrastructures.