Case Study
Optimizing water pumping process
The project
The system was applied to a pumping station with a water tank that supplies a neighborhood with a mixture of residential and commercial customers. The tank has an average daily consumption that varies depending on the load. A pump with a VFD installed supplies the tank with water from a nearby water source, and the water reaches the end-user by gravity. The goal was to improve the pumping operation and reduce overall cost.
​
We used our platform to build a Digital Twin of the pump station - the supply, the demand, and the pump itself. Тhe Digital twin accurately represented the real conditions under which the pump station operates. It accurately predicts how much water would be needed in the next few hours.
In order to optimize the process, we trained a Reinforcement learning algorithm to decide at what frequency the pump should work at any moment, and when should it work (based on electricity prices).The system dynamically controls the pumps and effectively automates the pumping process.
​
When using the system in real conditions, we managed to successfully bring an average of 29.25% savings from electricity consumption and 34.38% in periods with lower consumption.
For the same period, combining the reduction of electricity consumption, the system also selects the time intervals with the lowest price of electricity on the free market. Thus, we further reduced the price per m3 of water and we saved 12.24% of the price compared to the base price.
Key concepts
-
Digital Twins
-
Energy-efficiency
-
Artificial Intelligence
-
Water pumping
-
On-premise
Results
-
30% average reduction in electricity consumption
-
Digitalized and automated process