Amid global warming, the COVID-19 pandemic, and armed conflicts around the world, energy prices have been soaring and everyone is searching for ways to reduce energy consumption and costs.
Hitachi Air Conditioning assigned the problem for the innovative energy conservation division of the AI Wave: Taiwan Generative AI Applications Hackathon, held on May 18-19, 2024 by DIGITIMES. They challenged the participating teams to utilize generative AI to develop smart interactive applications for air conditioning systems that enhance user experience.
Hitachi Air Conditioning also provided resources and data for the teams to use. The team named "Hao I Coding," comprising Zi-Zun Fang, Ying-Rong Lin, Yun-Hao Chang and Ruo-Yu Wu from the graduate program of the Department of Information Management at the National Central University as well as An-Chi Wang from Institute of Information Management, Department of Industrial and Information Management, National Cheng Kung University, built a solution that uses smartphones to keep an eye on electricity usage and leverages generative AI for electricity usage analyses. The team stood out from the competitors in the Hitachi Air Conditioning- innovative energy conservation division and won the honorable mention award, voted unanimously by all the judges.
Coping with rising electricity prices, households are taking measures to reduce power consumption. After in-depth research, Hao I Coding found three major challenges hindering households' energy conservation efforts - inaccurate responses from chatbots, lack of air conditioning settings, and no way of knowing the electricity usage of traditional home appliances.
To help address these challenges, the team decided to start by connecting smartphones with air conditioning systems to collect complete operation data and detect system malfunctions as soon as they occur. Based on that foundation, the team used generative AI to analyze air condition systems' electricity usage to help users find a perfect energy conservation balance, thereby resolving the two challenges - lack of air conditioning settings and no way of knowing the electricity usage of traditional home appliances.
Although there are many generative AI solutions with Large Language Models (LLMs), they have difficulty providing correct responses to energy conservation questions without appropriate training using domain-specific data. Because of this, Hao I Coding conducted complete testing and decided to use the Anthropic Claude 3 Sonnet model in AWS Amazon Bedrock, a fully managed service that makes high-performing foundation models on Amazon Web Services (AWS) coupled with energy conservation big data, Hitachi Air Conditioning user manuals, and power-consuming appliance lists for training. They were able to build a chatbot that provides information-laden, logical and practical advice.
Hao I Coding noted that the AWS workshop was a very valuable part of the hackathon as they did not have any opportunity to learn about generative AI and LLM before.
The team named "Hao I Coding Team" won honorable mention for their smartphone-based AI solution for electricity usage analysis