Model Predictive Control for the Hera Group treatment plant in Modena
The Hera Group is one of the largest Italian Multiutilities, number one in terms of volume of waste treated annually and second in terms of volumes of water sold, with over 35 thousand kilometers of managed water networks and more than 3.6 million customers of the water service alone.
As part of the purification service, the company has identified for the Modena plant opportunities for further energy efficiency - compared to what has already been optimized - which cannot be resolved with existing controllers. For this reason, the need arose to explore new control techniques which would facilitate operators' activities.
For this purpose, we have developed a controller for one of the oxidation tanks of the purification plant, which uses predictive logic of artificial intelligence, to predict the evolution of pollutants and regulate them in advance, thus enabling the optimization of energy consumption and improving the quality of the outlet water, further reducing, compared to the legal limits, the concentration of substances that are inevitably present, such as nitrogen.
THE CASE STUDY
Objectives of the project- To prevent pollutant load peaks
- To promote energy savings
- To support personnel in the development of a “data driven” approach based on an in-depth analysis of process data
We used a predictive controller characterized by artificial intelligence models capable of predicting the evolution of the system and optimally adjusting the process set points. By taking advantage of Machine Learning algorithms, the controller understands the dynamic nature of the system and adjusts the optimal oxygen value to be introduced in the purification process, according to the forecast of the nitrogen concentrations leaving the plant.
Results achieved-8.1% Reduction of the total nitrogen concentration in effluent
-15% Reduction of energy consumption
- Active involvement of all operators in the process
WOLF – Wastewater optimization load forecast
This technology integrates the technical-engineering know-how necessary to model the system, allowing the physic-chemical logic that governs it to be kept under control. After the process of preparation and data analysis, different predictive models are tested, including all the variables influencing the system, in order to identify the model with the best forecasting performance and the maximum descriptive capacity of the phenomenon.