A Behavior theory integrated ANN analytical approach for understanding households adoption decisions of residential photovoltaic (RPV) system

Abstract

As residential photovoltaic (RPV) system gradually replaces fossil energy, it is imperative to conduct the RPV adoption decision analysis to investigate factors that drive RPV growth. Non-linear relationships are common in analyzing adoption behaviors, but they are often ignored in the existing studies due to the limitations of conventional data analysis approaches. On the other hand, Artificial Neural Networks (ANNs) are robust in dealing with non-linear relationships, but they lack interpretation capability, making ANNs unsuitable for RPV adoption analysis. To resolve the interpretation issue when considering non-linear relationships, this study proposes a six-step analytical procedure based on hybrid ANNs. The behavior theory was first integrated into the ANN to improve the model's performance. Afterward, the network weight-based method is implemented to calculate the importance of factors from the trained ANN. The proposed approach was used to analyze RPV adoption in Singapore, and the results indicated that the hybrid-ANN outperforms existing models in predicting and explaining adoption behaviors. The study demonstrates the suitability of adopting ANNs in decision analysis by developing a novel way to construct the hybrid ANN in considering non-linear relationships. In addition, the case study has identified that unfamiliarity with RPV hinders the adoption, emphasized the role of social support in promoting RPV, and proposed three practical policy implications for RPV development in Singapore, which can be extended to residential buildings across the globe.