Using artificial intelligence to improve Energy-from-Waste plant performance

Non-conforming waste can cause blockages in energy-from-waste incinerators and lead to unplanned shutdowns, costing a single plant up to 100,000€ per shutdown due to lost production capacity.  Moreover, the blockages caused by non-compliant items must be removed, which presents a health and safety risk and often requires repair to restart normal plant operations.  

As part of its waste recovery activity, SUEZ has set up an industrial sorting process: waste is sent to a transfer station where it is monitored remotely by operators, via screens, before being sent to a plant where it will be incinerated to produce electricity. This process requires the constant focus of the operators, who must observe up to 8 screens simultaneously to detect non-compliant items in the continuous waste stream. These are typically bulky objects, such as mattresses, metal pipes, supermarket shopping trolleys or rubbish bins.

An application of artificial intelligence (AI) called computer vision can provide operators with more accurate monitoring tools and thus limit the risk of unplanned shutdowns.  The AI-enabled solution acts as an extra set of eyes, raising alerts when non-conforming items are detected and “augmenting” the operator in their day-to-day responsibilities.

The solution was developed in a three-part collaboration between the Group Digital team, SUEZ in the UK, and a startup partner called Deepomatic.  A Proof of Concept (POC) was first completed in 2019 as part of the SUEZ Digital HUB Acceleration Program, and the solution was later industrialized and deployed with the support of the Group Data Office in 2020.  

The solution analyses images from surveillance cameras and alerts operators to the potential presence of non-compliant objects in the waste stream. For each detection, the operator accesses images of the non-conforming item, processes an alert, and can then take necessary actions to remove the item safely.  

In 4 months of operation, the system detected 11 non-conforming items that were missed by the operator and could have caused blockages.  Since deploying the solution, the EfW plant experienced a sharp decline in shutdowns, and the Group is looking to scale up the solution to other EfW sites in the UK, France, and Belgium.