Universal artificial intelligence workflow for factory energy saving: Ten case studies

Abstract

Numerous studies have affirmed that artificial intelligence (AI) can effectively enable energy savings in factories. However, there is currently a lack of explicit research that identifies the energy-saving effects of AI methods as compared to the conventional practices employed in factories, which involve the replacement of equipment with high energy efficiency (non-AI method), with respect to variance in the effectiveness. Thus, this ignited the motivation for this study, wherein the research team developed a “Universal AI workflow for energy saving” which adopted a standard workflow of “learning, Modeling, and Prediction.” Additionally, a three-year research project was launched to conduct empirical tests in ten factories, comparing the efficacy differences between AI and Non-AI methods. In 2021, research team established the energy consumption baselines for each factory through data collection; the entirety of 2022 was dedicated to AI integrations; and energy-saving benefits were evaluated in 2023. During the research period, each factory also implemented other non-AI energy-saving methods which were juxtaposed with AI approaches for comparison. Our study results showed that average AI-enabled energy savings across the ten factories amounted to 106,124 kWh/yr, while average energy savings from non-AI methods amounted to 1,231,625 kWh/yr. Despite the inferior energy-saving results, the AI methods had an average return on investment (ROI) of 0.46 years while non-AI methods had an average ROI of 6.22 years, indicating that AI methods could recover the investment costs through energy savings in less than one year. It is also worth noting that the ten factories produced a diverse range of products encompassing wafers, steel, and foods, but the AI methods used for energy saving employed the same workflow. By contrast, the non-AI methods necessitated distinct workflows and detailed engineering adjustments. It can be inferred from these empirical findings that AI techniques for achieving energy-saving control in factories can be integrated using standardized processes. In summary, AI methods can facilitate standardized workflows in factories and offer substantial energy-saving benefits that can typically cover investments within a year. AI represents a highly viable energy conservation option for factories seeking to maintain uniformity in product processes and realize efficient ROIs.