SDG 6: Clean Water and Sanitation
SDG 9: Industry, Innovation and Infrastructure
SDG 11: Sustainable Cities and Communities
SDG 14: Life Below Water
2. Project Details
Company or Institution
Artificial intelligence models for water conservation
General description of the AI solution
The ecosystem of environmental services tends to adopt innovative solutions very slowly. In wastewater treatment there are two separate goals which should be jointly pursued: nutrient removal and energy saving. We have developed AI models thanks to neural network algorithms capable of predicting the load evolution inside the oxidation tanks. This algorithm has been field tested on several both municipal and industrial plants, with relevant results in terms of better effluent quality and energy saving.
Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.
We have applied AI techniques in a field where it is difficult to modernize the approach to problems, because we are talking about the oxidation tank of the wastewater treatment plant, a system where chemical and physical variables converge. The goal was to radically innovate the control of oxidation tanks, the most energy-intensive part of wastewater plants: we applied AI systems in order to overturn the current control systems based on a retroactive logic and applying ML and DL models to create forecast models. We have gone from a retroactive logic (values measured in the tank) to a predictive logic (expected values at the outlet of the tank). In particular, our focus was on monitoring and forecasting nitrogen components (ammonium and nitrates). By predicting the evolution of these values in the output of the tank about half an hour in advance, it is possible to act preventively modifying the dissolve oxygen to ensure compliance with regulatory standards on water quality and to save energy. In this way, the control logic is radically transformed, from feedback to prediction. Briefly: we used machine learning and deep learning prediction models to predict ammonium and nitrate leaving the tank. We have created an optimization algorithm to identify the dissolved oxygen setpoint to be introduced into the tank, to comply with the control objectives: water quality standards and reduction of energy consumption. In September 2019, we presented the first results of the project at the 10th IWA symposium in Copenhagen – https://iwa-network.org/events/10th-iwa-symposium-on-modelling-and-integrated-assessment/, in 2020 we published a paper on the IWA journal “Water science and Technology” – https://iwaponline.com/wst/article/81/11/2391/74854/Real-time-model-predictive-control-of-a-wastewater. As the results have confirmed the expectations, we have been further investigating new techniques to facilitate the deployment and the dissemination of the solution.
Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.
The solution responds directly to four SDGs: the 6th, for the significant improvement in water treatment and saving capacity; the 9th, for the improvement of industrial waste treatment capacity; the 11th, for the improvement of local policies in water treatment, preserving a common good and promoting holistic sustainability; the 14th, for its ability to prevent the risk of eutrophication and ensure greater control of the environmental impact of treated water. The current type of control technologies could not guarantee compliance with all the quality standards for treated water required by law: being based on a retroactive control logic it was not possible to prevent spillage risks. In return, the introduction of our solution within the plants allows to predict the water values well in advance and enables to prevent the risk of dispersion of pollutants into the environment. The introduction of this solution allows to generate a direct impact on two levels: 1. through the modernization of the plants, it is possible to treat civil and industrial wastewater with a greater forecasting capacity, a better treatment capacity and an infinitely lower environmental impact; 2. the result allows the treated water to be reintroduced into the water cycle, ensuring the protection of the physiological environmental balance. In this circular economy perspective, it is possible to reuse treated water in agriculture, or for other uses, depending on sectoral regulatory standards, in the various Countries. What is certain is that the introduction of this AI solution applied to a model predictive control system allows to preserve and restore the value of water that would otherwise be destined to be lost, or, worse still, to be reintroduced into the environment, with harmful effects. The impact of this solution will have a profound effect on global water treatment scenarios, with a multilevel and multisectoral impact.
Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.
We have created a controller that has already been tested in several plants, and implemented on a full scale, in the field, in a large wastewater treatment plant (500k population equivalent). We obtained the results in two ways: 1. carrying out alternating tests: alternating our AI-based controller and the traditional controller, for periods of 3-6 months, on a weekly basis, in order to normalize the incoming load. The results we have obtained are: (a) compliance with the legal limits and improvement of the treatment result (8% reduction of the limits on total nitrogen); (b) significant energy saving (15% reduction in consumption, compared to an advanced controller, based on Fuzzy logic; more than 30% reduction compared to more common controllers, with fixed setpoint). 2. the other added value has been the decision to implement a dashboard to support the control, to allow the operator to know the system directly and to interact with it. The human being does not play a passive role, but is directly involved in the process, enabled to analyze trends and change some constraints based on different conditions for the seasonality. The human being becomes an integral part of the developed controller, while still guaranteeing the possibility of a simple supervisory role. Our control system is open, not a black box, but through the dashboard it highlights the measured and forecasted variables but also the logics that lead the selection of the best dissolve oxygen setpoint to be set in the tank. The solution has been very successful because it allows to respect all the regulation limits and at the same time to provide those who manage the plants with a greater knowledge of the systems, a deeper understanding of the processes, being able to interact with the plants through user friendly tools.
Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.
The solution was conceived starting from an ethical framework, to ensure that the impact could be positive for humanity, in a global perspective for integral sustainability. In the short term we expect the solution to spread within the market, to support a progressively better implementation, but at the same time the results will be made available to the global scientific community. In the long term, we expect the possibility of significantly improving the quality standards of treated water to influence international politics, through awareness building campaigns, including lobbying and advocacy actions. The solution, supported by scientific evidence, makes it possible to translate the possibilities deriving from the responsible application of AI into a tool for the overall improvement of the ability to promote policies for sustainability, converging society, business and policy makers in the recognition of the common interest. Starting from the Italian context, this solution is also part of our cultural development plan, based on overcoming the traditional concept of the so-called human-centered AI. Anthropocentrism is not the centrality of human rights as opposed (or superior) to the rights of the planet, but the centrality of human responsibilities with respect to the environment and all the creatures that inhabit it. In return, for two years we have been promoting a widespread strategy of social literacy, to enable end users to understand how AI works and its impact in daily life. In addition to media coverage, we have been disseminating the solution impact within our educational programs, for schools, society and business, with the aim of making people understand the central role AI can play. For over a year we have been actively participating in international networks of scientific diplomacy and public diplomacy, especially with Latin America, to promote and enhance our solutions within international institutional networks.