A growing number of people and businesses all over the world already rely on desalination plants for their water. But what about the future, as climate change increases the need to produce water in places where scarcity was previously unknown? Here we review the evolution of desalination technology and give some insights on the potential to improve its efficiency and resilience through digitalization. In particular, high-fidelity predictive modeling can help desalination plants increase their performance and thus the security of the water supply for end users.
The desalination market has been growing for the last 20 years. What has changed in the last few years is the consciousness that, due to climate change, this technology is needed not only in deserts and hot countries but even in places like central Europe or the UK; we will be seeing more and more plants in the coming years. Therefore, following the evolution of this technology over the last 20 years is of high interest. It moved from thermal desalination coupled with power plants, over multi-stage flash technology, to reverse-osmosis membranes using top-quality membrane material. The driver for these developments was always a reduction in energy consumption, which dropped from over 5 kWh/m3 to 2–2.5 kWh/m3, increasing the energy efficiency and therefore the sustainability of the treatment as well. The technology of forward osmosis has led to several publications in the scientific field but has not yet evolved to an industrial level. In all cases, the thermodynamic minimum depends on the recovery rate as well as on physical and chemical conditions, but it lies between 0.8 and 1.5 kWh/m3. This shows that the strong developments in process technology have brought desalination close to the limits. Two recent trends can be observed in making desalination even more energy efficient and sustainable: digitalization and the use of renewable energy for desalination.
If we focus on digitalization, we need to keep in mind that a continuous, stable, and efficient operation is needed to deliver constant water quality to consumers. If you look closer, however, this task becomes increasingly complex. To achieve this goal, plant operators need to optimize multiple operational KPIs and parameters. Moreover, “optimum performance” can have different meanings for different plants and different modes of operations – so KPIs include consumption of energy and chemicals in relation to water output, production costs, and energy consumption. At the process level, inefficiencies in operation can also have multiple causes: excessive use of energy and chemicals due to suboptimal process conditions, high maintenance requirements due to module failures caused by fouling, and inconsistencies in water quality that impact consumers. Therefore, plant operators need effective strategies for monitoring and managing process conditions, and they need tools to support informed decision making for operation and maintenance of the entire desalination process.
Why a digital twin of performance?
Digital twins of performance help to make plants more energy efficient and at the same time more reliable, resulting in a more resilient water supply for water consumers. One essential part of these strategies and tools is process simulation. Using model-based optimization software, plant operators can capture fundamental knowledge about a process – its physics, chemistry, control philosophy, operating policy, feedstock and energy costs, and product prices – in the form of mathematical models and their associated data. A system of such models can then be used in conjunction with state-of-the-art mathematical techniques to analyze and optimize process design or operation, improve plant efficiency, reduce energy consumption, and even optimally schedule cleaning processes. In desalination plants, such models can help optimize multiple operational KPIs and parameters – for example, the consumption of energy and chemicals in relation to water output, production costs, and resource efficiency. Moreover, modeling facilitates identifying optimal operation and cleaning parameters to maximize water output while preventing module failures.
What simulation to use? A quick comparison
At the core of such a digital twin of performance are complex mathematical models. Broadly speaking, there are two general principles that can be applied to process modeling: artificial intelligence (AI) methods and specifically machine learning (ML) solutions as the most commonly used AI method employed for process analysis and optimization on the one hand, and model-based solutions using high-fidelity mathematical models (either static or dynamic) to simulate and optimize processes on the other hand. While AI/ML methods are currently what most people think of first when they hear “data-driven optimization,” I want to point out that high-fidelity models do have benefits for process applications. Following are some reasons why.
First, let’s look at the prerequisites. High-fidelity models are based on the equipment configuration, physical and chemical equations, and operating procedures of the respective plant. The model is calibrated based on selected process data, and creating the model requires detailed process know-how and a high level of expertise. Nevertheless, after the calibrated model has been achieved, the optimization algorithms help us get to the highest possible level of performance. On the other hand, AI methods will always work in existing conditions and optimize to the best “experienced” level of performance.
Next, let’s look at lifetime considerations. Process models improve as more knowledge about the real process becomes available, and the digital twin can be amplified successively to more plant sections. Process models can be maintained and updated as well as reused for other purposes – for example, for operator training or to explore hazardous or theoretical scenarios. Also, the models can be transferred to similar plant types and therefore reduce engineering work.
How to set up a high-fidelity simulation?
With the right simulation software and high-fidelity modeling, users can create a robust, reliable digital process twin that supports a dynamic simulation and optimization of the process. For example, the gPROMS process-modeling environment encompasses the entire physical and chemical process in rigorous, equation-based, high-fidelity mathematical (white-box) models. The gPROMS environment can be used for end-to-end simulations and optimizations both online, with a connection to the plant process-control system providing optimal setpoints, and offline. The physical sensor portfolio can be reduced due to its capacities in soft sensing. When used online, the digital performance twin can run parallel to plant operation and use all current distributed data to make recommendations for operators – for example, on the best strategy to improve a given KPI (such as minimizing energy consumption). The gPROMS environment provides the required process visibility needed for effective process optimization.
The software enables informed decision making and helps operators improve the overall operational effectiveness of desalination plants. It also supports upgrades and plant modifications. Modeling with gPROMS can also help identify optimal operation and cleaning parameters to maximize water output while preventing module failures.
What else can digitalization do for desalination?
This is just one example of how digitalization can help achieve higher performance and efficiency, contributing to a more sustainable and resilient water production. The capabilities of digitalization are not limited to optimizing plant performance. The gPROMS environment is part of the Siemens Digital Enterprise portfolio for the water industry. With comprehensive industry experience and digital solutions, Siemens helps implement innovative business solutions for the water industry that work to secure a sustainable, reliable water supply. Curious about what else digitalization has to offer for desalination plants? Follow Siemens on smartwatermagazine.com and/or visit our landing page to get all the news.