Malta
Optimal Management of Water Resources and Assets for a Water Stressed Mediterranean Nation using AI Prediction and Multi-Objective Optimization
In this project, NOAH Global Solutions working with their European subsidiary, was awarded a groundbreaking contract by the European Union to help protect the future of an island nation.
Because of increasing water scarcity, many areas depend on a combination of different water sources when available. Different sources, however, typically have different costs and benefits relative to numerous factors, including environmental, financial, social, technological, and even cultural. Furthermore, there may be different risks or vulnerabilities associated with each water source. For example, while groundwater in coastal areas is often vulnerable to salt-water intrusion, surface water is often vulnerable to algae blooms. Climate change and the associated uncertainty that it imposes on both water demand and water supply is another complicating factor that must be considered to reduce risk and increase water security.
The objective of this project, executed under the contract, entitled Development of a Predictive Water Abstraction and Production Modeling Framework, was to study the feasibility of implementing NOAH’s Patented water management decision support system to optimize conjunctive management of Malta’s complex water assets and operations. NOAH executed the work in partnership with Malta’s Energy and Water Agency (EWA) and Water Services Corporation (WSC).
Malta’s vision is to implement a data driven AI and optimization decision support system for optimizing their complex water assets and operations. The system would help Malta’s water managers identify the optimal balance between minimizing energy consumption costs, maximizing the quality of consumer provided water, and maximizing protection of the groundwater system from over-pumping.
As part of this study, NOAH:
- Assessed Malta’s Supervisory Control and Data Acquisition System (i.e., SCADA) system for automatically measuring and transmitting critical resources and operational data to a centralized database.
- Assessed Malta’s software system architecture for integration with a real-time AI and optimization-based decision support system.
- Analyzed data to increase understanding of critical relationships between different operating conditions and system responses.
- Assessed the existing data collection strategy within the context of Malta’s natural water resources and water infrastructure assets for implementing a prediction and optimization modeling system necessary for optimal water management.
- Using representative Malta data sets, developed and assessed AI capability for predicting important time changing variables of interest like water demand and chloride concentrations in groundwater.
- Using representative Malta data sets, developed and assessed optimization models for reducing energy consumption costs, optimizing customer delivered water quality, and minimizing salt-water degradation of groundwater from over-pumping. Stochastic models included the element of weather uncertainty, which is becoming even more prevalent with climate change.
NOAH’s singularly unique system combines real-time data with artificial intelligence (AI) based prediction and optimization management models. This intelligent data-driven system generates important predictions like water demand, water quality, and groundwater elevations. Accurate forecasting is essential for improving water utility operations and protecting water resources. However, prediction models alone will not necessarily identify good, let alone optimal management solutions. What separates NOAH’s system from others is our formal optimization modeling feature, which takes decision making to an entirely different level. Optimization can achieve any number of important objectives, like maximizing water resources protection, minimizing operational costs, and maximizing consumer provided water quality to the extent possible. And because objectives often conflict, an optimal balance between conflicting objectives is a common but typically unattained management goal.
For example, minimizing operational costs may result in poorer water quality. NOAH’s system can also perform multi-objective optimization, where the optimal trade-off between competing objectives is computed. In this way, decision makers ensure that they are operating their system where all management objectives are optimally balanced. The system when directly integrated with real-time data streams computes optimal solutions at the desired frequency that reflect real-time conditions. Furthermore, using advanced statistical algorithms, the decision support system can continuously account for uncertainty like weather, which drives important system conditions like water demand and water supply availability.
Malta, a nation of natural beauty and cultural richness, represents the ultimate challenge in water management for the numerous reasons outlined below.
BACKGROUND
Google Earth Image of Malta and Surrounding Sicily and Northern Africa in Mediterranean Sea
Malta, the most densely populated European nation, is an archipelago consisting of one main island and several smaller ones located in the Mediterranean Sea between Sicily and Africa. It now ranks among the most water scarce nations in the world.
The ships from the sea-trading Phoenicians once heralded from its waters. In the 1500s the legendary Knights of Malta were granted ownership by Charles V, Holy Roman Emperor, to whom they swore allegiance, presenting him with a prized Maltese falcon each year. Even during the reign of the Knights, water scarcity was a problem for Malta’s inhabitants.
Main Harbor, Valletta, Malta
CHALLENGE
Today, a thriving tourist destination with a booming economy, Malta is experiencing a ‘perfect storm’ all too typical across the globe: too little natural freshwater for their growing demand, made worse by more frequent and intense droughts from climate change.
Due to over-pumping of their scarce groundwater resources, Malta’s limited limestone aquifers have been degraded to some extent by salt-water from the surrounding Mediterranean Sea, and remain at risk to this threat.
Island Comino, Malta
Malta derives almost all its water from two sources; a deep limestone aquifers 400 feet below ground, and three reverse osmosis plants that desalinate Mediterranean seawater. Given the complexity of their water resources and assets, combined with their stressed groundwater aquifers, growing population, and increasingly dryer climate, Malta represents the ultimate challenge in water management.
Conceptual Depiction of Malta’s Major Water Sources and Distribution System
Sixty percent (60%) of Malta’s water supply comes from three (3) reverse osmosis (RO) plants. While RO is of superior quality, containing far less chloride or salt water, its cost is five times more than groundwater.
The remaining water demand is met through extraction of threatened groundwater via 120 production boreholes or wells and several large galleries. The largest groundwater source is stored in massive karst limestone aquifers characterized by large fissures, caverns, and sinkholes. Historical over-pumping of the deep limestone aquifers has caused waters from the Mediterranean Sea to partially move into these irreplaceable water supply formations, degrading their water quality with salts like chloride.
NOAH staff visiting the Control Room at one of Malta’s Reverse Osmosis facilities.
Conceptual Cross-Section of Malta’s Hydrogeologic System and Groundwater Production Assets
The massive limestone aquifers with complex flow pathways and seemingly random interconnections to the Mediterranean Sea constitute a highly complex natural system. Superimposing the effects of pumping extractions from hundreds of wells further complicates the complex movement and interaction of groundwater with seawater. As such, it is impossible to accurately predict and therefore optimally manage, in real-time, these complex freshwater/seawater interactions with conventional physics-based groundwater computer models.
Alternatively, with the right combination of data and expertise, AI models can learn to accurately predict complex systems far better than traditional physical-based models, particularly for real-time predictions at specific locations. Based on NOAH’s previous work and in discussions with Malta’s water experts, it was assumed that AI models with proper data collection and assimilation could accurately predict important dynamic groundwater responses in the massive complex limestone aquifer in Malta, like time changing salinity (i.e., chloride) concentrations at specific locations.
Wellhouse for a public production borehole or well
DATA SYSTEM AND WATER MANAGEMENT
A Water Services Corporation operator monitors data from the main control room.
Malta deploys an advanced data collection system which monitors critical operational and state variables like water production rates, groundwater elevations, and water quality conditions in their assets, including groundwater production sources, RO plants, water reservoirs, and monitoring wells. WSC’s operators analyze much of the data in real-time to evaluate system conditions and manage the enormously complex network of RO plants, groundwater extraction wells, and reservoirs. A subset of this data was used to develop and assess preliminary AI prediction and multi-objective optimization management models.
STUDY APPROACH AND RESULTS
NOAH developed preliminary AI models to accurately predict chloride concentrations in groundwater that dramatically change over space and time due to variable extraction rates and other conditions. A particularly difficult and unique groundwater extraction source was selected to test the AI prediction capability in Malta.
The Ta’ Kandja pumping station is an ingeniously designed water extraction system. Located 97 meters below ground, it consists of 6.2 kilometers of long tunnels or galleries incised into limestone, radiating out from a central shaft like spokes of a wheel. By laterally skimming water off the top of the Mean Sea Level aquifer along these long galleries, groundwater extractions are distributed over a larger and shallower aquifer area, reducing the potential for saltwater upconing. Still, this station experiences highly variable salinity concentrations due to freshwater/seawater interface fluctuations induced by variations in extractions by both the station and other nearby groundwater extraction assets.
A tunnel or gallery in the Ta’ Kandja pumping station
Results
The AI model achieved relatively accurate prediction performance for most of the events, including ones that exhibited large monthly increases or decreases in chloride concentrations. Further testing of this AI model demonstrated that it was learning complex system dynamics, and not simply over-fitting the data.
Measured versus AI Predicted Chloride Concentrations at Ta' Kandja Pumping Station
This analysis included a comparison of training, testing, and validation model performance for different model trials.
Similarly, NOAH developed accurate AI models for forecasting water demand due to changing weather conditions. Based on the data analysis and modeling, NOAH identified ways to further improve data collection and assimilation strategies to further support development of more accurate and robust AI prediction models.
Using formal optimization, the following three water management objectives were simultaneously optimized to delineate their trade-off surface:
- Final delivered water quality of the combined RO and groundwater sources;
- Combined energy consumption of these two sources;
- Saltwater degradation of the limestone aquifer due to over-pumping.
Generated 3-D trade-off surface for Malta
NOAH staff with meeting with Malta’s Agency for Energy and Water
The trade-off surface depicts in three dimensions how the values for each of the three objective functions increase or decrease as one moves along the surface, with a zero-value representing worst-case and a 1.0 representing the best case for each objective along each respective axis. Among the infinite set of points, there is no single point along the trade-off surface where all three objectives simultaneously achieve the best value of 1.0. For example, increasing energy consumption costs does not always increase water quality. In theory, no single point on this “non-dominated” surface is superior to any other point on the surface; it is all a matter of priorities.
Using this generated trade-off surface, the decision makers can visualize the complex interplay between the three objectives, from which the optimal water contributions can be identified for each of the individual RO and groundwater production sources.
That is, each point along the trade-off surface represents in a transformed space the individual operational rates of the reverse osmosis and the groundwater production rates for that particular point. Using the trade-off surface, the decision makers can identify from among the infinite set of points the optimum point or management solution which best represents their priorities for balancing the three objectives. There are numerous mathematical techniques for identifying the optimal trade-off point; Game Theory is one such method.
An extremely important but often over-looked benefit to performing a formal multi-objective analysis is the generated trade-off surface excludes the infinite set of inferior solutions, all of which lie outside of this surface. Therefore, the decision makers are at least guaranteed to not select a sub-optimal or inferior solution, which without conducting a formal optimization analysis for a relatively complex system, is unfortunately the norm.
Both deterministic and stochastic models were used in this study. Deterministic models assume that there is no uncertainty in future system conditions, while stochastic models recognize the inherent statistically uncertainty of real-world conditions like water demand due to changing weather conditions. Accounting for statistical uncertainty produces a more robust solution set where not only expected variability but also possible extreme conditions are also accounted for. With climate change creating more extreme weather patterns, NOAH’s algorithms further reduce risk and costs.
Future Work
As Malta’s water supply system increases in complexity and needs, the AI and optimization-based decision support system can augment their years of hard-earned expertise, reducing costs, improving consumer provided water, and protecting critical groundwater resources.
NOAH continues to work with Malta water resources and utility experts to implement innovative solutions to help manage their daunting water challenges. A nation with a storied history and rich culture, optimal water management is an imperative to allow this Mediterranean jewel to flourish far into the future.