EMAGIN Inc.

Water utilities extract water from a source; chemically treat it at a treatment plant; and then use pumps to distribute that water through a complex network of pipes to consumers. These consumers produce sewage that is then treated at a waste recovery plant and then discharged back at the source. In today’s industry standard, this entire process from source to tap and back is controlled manually by operational staff – who rely on their judgement when rendering operational decisions. These decisions range from determining chemical treatment dosages to scheduling pumps to supply demand. The problem is that this form of manual operations leads to decision-making that is not only energy inefficient but is fundamentally reactive. For example, water distribution in Toronto consumes more electricity than the Toronto Transit Commission (TTC) and 5 times more electricity than all the city’s streetlights and traffic lights. Moreover, operators are unaware of anomalies in their system, such as low water quality and pipe bursts, until after they occur. For example, the City of Toronto experiences over 1,000 pipe bursts a year - resulting in a tremendous volume of water lost. This form of manual operations also persists in industrial facilities leading to inefficiencies at scale - including but not limited to asset downtime and production losses, higher energy costs and reduced productivity.

EMAGIN aspires to address these issues by introducing artificial intelligence into the water sector. Specifically, EMAGIN provides water and wastewater utilities with an artificial intelligence driven platform, HARVI, that supports operational decision-making when controlling critical assets in real-time. By creating real-time recommendations on critical performance drivers, HARVI helps utilities shift their operations from reactively responding to issues to proactively managing them, thereby reducing operational costs, enhancing reliability and emergency preparedness.


HARVI works in 4 simple steps:
1) it integrates with the utility’s SCADA system and exploits existing sensors - we do not require any additional sensors.
2) it learns about the utility’s system via historical data.
3) it then generates hourly predictions on how the system will behave for the next 24-hours.
4) it uses those predictions to generate optimized recommendations for the operators to view on a dashboard and implement.