Global warming, the COVID-19 pandemic, and armed conflicts around the world have led to rising energy prices, with everyone looking for ways to reduce their energy consumption and costs.
Hitachi Air Conditioning presented a challenge to the Innovative Energy Saving category of the AI Wave: Taiwan Generative AI Application Hackathon held by DIGITIMES on May 18-19, 2024. Participating teams were given the challenge to utilize generative AI to develop a smart interactive application for air conditioning systems that would enhance the user experience.
Hitachi Air Conditioning also provided the resources and data used by the team. The “Hao I Coding” team, consisting of Zi-Zun Fang, Ying-Rong Lin, Yun-Hao Chang, and Ruo-Yu Wu from the Graduate School of Information Management at National Central University, and An-Chi Wang from the Information Management Institute at the School of Industrial Information Management at National Cheng Kung University, built a solution that uses smartphones to monitor power usage and leverages generative AI to analyze it. The team outshone their competitors in the Hitachi Air Conditioning Innovative Energy Saving category and received an Honorable Mention award by unanimous vote of the judges.
To cope with rising electricity bills, households are taking measures to reduce power consumption. After thorough research, Hao I Coding discovered that inaccurate responses from chatbots, inadequate air conditioner settings, and not knowing the power usage of traditional home appliances are three major challenges hindering households' energy-saving efforts.
To solve these challenges, the team first decided to connect a smartphone to the air conditioning system to collect complete operating data and detect system failures as soon as they occur. Based on that foundation, the team used generative AI to analyze the power usage of the air conditioning system and help users find the perfect energy-saving balance, thereby solving two problems: undersetting the air conditioner and having no way to know the power usage of traditional home appliances.
Although there are many generative AI solutions using large language models (LLMs), it is difficult to correctly answer questions about energy conservation without proper training with domain-specific data. For this reason, Hao I Coding conducted thorough testing and decided to train the Anthropic Claude 3 Sonnet model from AWS Amazon Bedrock, a fully managed service that creates high-performance foundational models on Amazon Web Services (AWS), by combining energy conservation big data, Hitachi air conditioner user manuals, and power-consuming appliance lists. This enabled us to build a chatbot that provides informative, logical, and actionable advice.
Hao I Coding said the AWS workshop was an invaluable part of the hackathon, as he had never had the opportunity to learn about generative AI and the LLM before.
The team “Hao I Coding Team” won the Excellence Award for their smartphone-based AI solution for analyzing electricity usage.
