top of page

It is common industry knowledge that reverse osmosis (RO), the process of forcing water through a semi-permeable membrane to remove dissolved and suspended solids, is a power-hungry process. In fact, RO is the main consumer of electricity in most seawater and brackish water RO plants. This represents a major opportunity for treatment facilities as energy use accounts for the majority of (44%) of the total annual operating expenditure at a typical SWRO facility, according to DesalData.

Want to get to the ❤️ heart of the matter? Click here to see our in-depth RO Whitepaper

With such a clear problem statement, why then has it been so difficult to optimize the energy consumption of the reverse osmosis process?


Process optimization is defined as the method of choice for improving the performance of an industrial process while enforcing the satisfaction of safety and quality constraints. In the context of reverse osmosis, what are the optimal controls (pressures and flowrates) and train configuration (on/off) to minimize energy use while reliably meeting product water requirements.

The Status Quo – Constant Recovery

The industry standard for maintaining performance and product quality is to run RO trains at constant recovery. While reliable, running at constant recovery does not account for variations in influent water, changing demand, and degradation in equipment and membrane conditions – leading to inefficiencies and foregone savings.

Manual Analysis

Proactive facilities may opt for manual analysis of individual trains to quantify membrane degradation and reactively tune performance for additional savings. This approach is practical for single train facilities but is not economic for multi-train facilities as parameterization becomes less feasible as complexity increases (see figure below):

Figure 1 - Optimization vs. system complexity

Let us take a trivial analogy to better understand the challenge in optimizing more complex RO systems. Imagine you enter a grocery store for the first time and are looking for a single item - bananas. In this case, you would likely identify the produce section, locate bananas, and identify the optimal size and ripeness for your needs. Now imagine Christmas has arrived and you have a larger shopping list of 50 items and a few friends to help you. You and your friends enter a new grocery store and are tasked with organizing your list into categories (assuming the grocery store is organized like other stores) and finding the optimal path to procure and select the optimal list items in the least amount of time.

The above-described complexity has limited RO optimization efforts to capital changes (e.g., procuring a new energy recovery device), reliance on operator experience, and expensive manual analysis of data to guide future performance. Both operators and managers have long known that RO set-points can be adjusted for operational savings, but the impracticality of manual analysis and absence of real-time computational tools have limited innovation in ongoing RO operations.


Recent developments in cloud computing and IoT technology paired with machine learning (ML) techniques have allowed for practical improvements in RO performance.

Pani’s AI Coach™, a web-based platform designed for reverse osmosis applications seamlessly integrates with existing systems (SCADA and PLC systems) to centralize data and prescribe optimal set-points (pressures, flow rates, on/off cycles) for optimum energy consumption and product water quality at different levels of demand:

Figure 2 - Prescribed set-points with RO Optimization

Want to learn more? Get the full RO Whitepaper






Post category

Plant-wide Risk Mitigation for Reverse Osmosis Facilities: Digital Twins for the Win

In our recent post, Top Three Opportunities for Machine Learning to Improve Reverse Osmosis Plans, we introduce the case for automated plant-wide risk mitigation in reverse osmosis facilities.

Feb 22, 2021


min read



Get the latest Pani insights and news in your inbox. 

bottom of page