A water treatment plant’s operations branch for a North American city is using Emagin AI by Innovyze to reduce the operational expenditure (OPEX) required to deliver high-quality drinking water to its approximately 150,000 residents.
The drinking water distribution system consists of around 4,000 hydrants and close to 650 kilometers of water main and transmission mains. The system features a Surface Water Treatment Plant (SWTP) with a current capacity of over 21,000 Mega Liters per year (ML/yr), together with associated low-lift pumping stations (LLPS), high-lift pumping stations (HLPS), groundwater wells, in-ground storage, booster stations, and storage towers.
For any drinking water distribution concerns, water quality is of paramount importance, and this – together with energy consumption – form the main cost areas. Reducing these costs helps deliver greater value to consumers and the municipality. Initially, the water operations department contacted Innovyze to investigate its energy usage with a view to saving costs. Innovyze’s team looked at other key cost areas where data could provide insight into additional cost savings.
To do this, they explored the process being used at the SWTP and identified membrane optimization as an area where substantial differences could be made to OPEX.
Four LLPS pump raw water from the primary water resource, a nearby lake, to the treatment plant. Here it undergoes a range of treatments; primary screening, flocculation, membrane filtration, carbon contractors, and disinfection.
The raw water first undergoes a pre-treatment process where chemicals such as coagulants are added to enhance the formation of flocculants. These settle at the bottom of the pre-treatment tank and are removed. This is the first stage of the water treatment process.
The pre-treated water is then sent into a two-stage membrane filtration unit, which is made up of a primary and secondary membrane. This part of the process removes bacteria, microorganisms, particulates and other material which can affect the color, taste and quality of water. The recovery of treated water from the membranes is based on the water operation’s desired recovery setpoint, typically around 70-80%.
After leaving the membrane units, the treated water is disinfected in the chlorine contact tanks and initially stored in the reservoirs located at the SWTP. Six constant speed, high lift pumps then transfer the water to a larger reservoir located outside the SWTP. The volume of water transferred is based on control logic that constantly monitors the reservoir levels. At the final part of its journey, the water feeds two DMAs and a storage tower to deliver clean drinking water to the city’s population.
With rising energy costs, the customer needed to reduce their energy consumption while ensuring that the condition of the membranes did not compromise water quality. They defined a two-phase deployment designed to achieve three main objectives – reducing energy costs, ensuring membrane effectiveness and maximizing membrane life, again reducing costs.
In the first phase, a reduction in energy consumption was to be achieved by intelligently scheduling and controlling of the high lift pumps’ run status. This would reduce overall energy consumption while still meeting defined system constraints. At the same time, the membrane integrity would be monitored in real-time, and the operator notified immediately of any failure in the membrane.
Phase two involves maximizing the life of the membrane while reducing the cost of maintaining it. This would be achieved by developing an optimized maintenance and cleaning regime, reducing unproductive downtime as well as the amount of chemicals consumed in the cleaning process.
To achieve the first objective of reducing energy consumption, the SWTP system was modeled to create an AI-enabled Digital Twin which was used to predict reservoir levels, demand and pump flow. This involved understanding the operational energy bill, which comprised two segments: Global Adjustment (GA) and the Hourly Local Energy Price (HLEP). GA is a fixed cost, populated at the end of each month while the HLEP is a dynamic energy tariff that changes every hour. GA makes up 80 percent and HLEP 20 percent of the total energy cost.
Capturing the unique energy system meant clustering the HLEP, with each cluster having a distinct pattern, and generating exemplar dates as well as objective function factors such as total energy consumption and hourly energy price. This was based on the fixed price GA as it made up the major portion of the cost. It was found that controlling pump run status was effective in minimizing the overall plant energy consumption. Emagin AI was able to show how energy cost savings of 15 to 20 percent were achievable.
The second objective involved real-time monitoring of the integrity of the membrane. Every membrane undergoes cleaning and a Membrane Integrity Test (MIT) every day. Skipping this daily cleaning process adversely affected the integrity and life of the membrane. By observing Transmembrane Pressure (TMP) and Log Reduction Value (LRV) trends, the system could tell whether cleaning was carried out or not. Cleaning the membranes showed up as spikes in TMP of >75kPa and a change in LRV. Emagin created an algorithm to detect this trend within a 24-hour window. If the trend was not observed, the operator received a notification specifying the date and particular membrane where the anomaly occurred. This alert triggered the need to clean the affected membrane and preserve its integrity and effectiveness.
The third objective, to extend membrane life while reducing the cost of maintenance activity is currently in the planning stage with the customer.
In summary, as a result of the work undertaken by the Emagin AI deployment for this particular water operations department, Emagin AI was able to achieve 15%-20% energy cost savings, and proactively identify when membrane maintenance was not undertaken over a 24-hour period, notifying the customer that maintenance is required.
As this is an ongoing deployment, with phase two imminent, the customer expects to see additional results and even more learning from the work. This is a success story that we will revisit to further demonstrate the capabilities of AI to reduce OPEX within surface water treatment plants.