SOINN E-1 Energy-Saving AI for HVAC and Heat Sources

SOINN Inc. is a Tokyo-based AI company registered in 2014 by Dr Osamu Hasegawa, a former associate professor at Tokyo Institute of Technology. It builds very lightweight AI based on its patented SOINN algorithm, which learns from small amounts of data on ordinary CPUs, and focuses on energy management, anomaly detection and equipment control.

Its E-1 energy-saving AI divides a building into tens to hundreds of small zones, assigns a dedicated model to each and retrains all models daily on a single CPU without GPUs. Precise predictions feed optimisation, including stopping air conditioning when possible, for both heat sources and air handling equipment. E-1 connects through BACnet, Modbus, LonWorks, SLMP and other standards, works with most equipment brands without retrofitting, and can run in the cloud or on an on-site PC. Users include large venues, data centres, chain stores and factories.

• Zone-level models: splits buildings into tens to hundreds of zones with a dedicated AI model for each.
• Lightweight daily learning: all models retrain automatically every day on one CPU without GPUs.
• Heat source and HVAC control: optimises both primary heat sources and secondary air conditioning equipment.
• Multi-protocol connectivity: supports BACnet IP and MS/TP, Modbus RTU and TCP/IP, LonWorks, SLMP, FINS and Syslog.
• Fast start-up: launches from two to three months of data and reuses trained models for new sites.
• Flexible operation: from demand forecasts and saving recommendations to fully automatic control, plus anomaly detection.

Exhibition centres, shopping malls and hospitals in Vietnam with BACnet-based BMS can add automatic zone-level HVAC optimisation without replacing equipment. Japanese-invested factories can optimise chillers and boilers around production plans. Data centre operators can test the AI on cooling systems under a shared-savings fee.

In a verification at Tokyo Big Sight, E-1 cut electricity and heat use by about 40%, according to the company. The initial investment is usually recovered in a few months to one year, and the running fee is in principle 40% of the cost savings achieved, with flat fees possible for chains. Operation is fully automatic after start-up, input data and trained models need not be disclosed to SOINN, and anomaly detection can run in parallel.

Large venues, data centres and factories in Asia want AI-based HVAC optimisation but often lack clean data, budgets for GPUs or permission to share operational data with vendors. Lightweight edge AI that learns from small datasets and works with existing BACnet or Modbus systems lowers these barriers, while performance-based fees match the shift toward energy-saving services.

Registered in 2014 in Machida, Tokyo, and led by CEO Dr Osamu Hasegawa; Seibu Shinkin Capital is a major shareholder. The company holds patents on the SOINN algorithm, was selected for Tokyo's green startup support and Google for Startups Cloud Program, and its energy-saving AI is a certified trial product of Machida City.