Most DLE deployments today operate on the same basic control philosophy as industrial pH control or level control from the 1970s: set a target, measure the process variable, make a correction when you deviate from target. That approach works in systems with relatively slow dynamics and stable inputs. Electrochemical DLE is neither.
The Fixed-Setpoint Inheritance Problem
When DLE electrode technology was being commercialized, the control engineers involved drew from what they knew: industrial SCADA and DCS systems with PID loops. This is an entirely reasonable starting point. The problem is that PID control applied to an electrode cycle treats the electrode potential (or current) as the controlled variable and the desired extraction efficiency as the setpoint, and assumes that a fixed relationship between the two holds across operating conditions. That relationship does not hold when brine composition varies, when temperature shifts, or when the electrode ages.
The commissioning-era setpoint problem is specific: during commissioning, the system is tuned against the brine composition and temperature conditions present on the commissioning dates. Those conditions are not the steady-state operating average. They are a snapshot. For salar brine sources, the brine composition during commissioning may be substantially different from the seasonal average the plant will run against for the rest of its life. For geothermal sources, the fluid temperature on commissioning day depends on which time of year it is and what the production rate was in the preceding weeks.
Operators who have run DLE pilots with fixed setpoints report a common experience: the system performs well for the first few months after commissioning, and then begins to show declining cycle efficiency or increasing purity excursions as the brine composition drifts from its commissioning-era value. The typical response is a manual setpoint adjustment, which is then held until the next time performance degrades. This is not a control system; it is a step-function approximation of a continuously variable process that needs continuous adaptation.
What a Closed-Loop System Requires
A genuine closed-loop control system for an electrochemical DLE cell stack requires three functional elements: sensing (continuous measurement of the variables that drive electrode behavior), a control model (a representation of how electrode behavior depends on those variables), and actuation (the ability to change electrode operating parameters in response to model outputs).
Sensing in this context means more than measuring electrode current and potential. It includes measuring the brine feed composition (Li+, Mg2+, conductivity, temperature, pH), measuring the electrode effluent composition (Li+ and Mg2+ at the cell outlet, which reflects what the electrode has retained), and measuring the electrode state of charge (which can be inferred from the ratio of current to a reference state at the same potential, or directly estimated from the voltage-time profile during adsorption). Each of these measurements carries lag and uncertainty, and the control model must account for that.
The control model does not need to be a first-principles electrochemical simulation. First-principles models require calibration parameters (crystal structure, surface area, diffusion coefficients) that vary with electrode manufacturing lot, age, and fouling state, and keeping them current in a production system adds maintenance overhead that most operators do not have capacity for. A semi-empirical model that captures the key input-output relationships (how does adsorption rate vary with Li+ concentration, temperature, and electrode state of charge?) and is calibrated from operational data during commissioning and updated through continuous operation is more practical and nearly as accurate for control purposes.
Multi-Cell Stack Coordination
A multi-cell DLE stack introduces a coordination problem that single-cell operation does not have. In a 4-cell stack where two cells are in adsorption and two are in elution (a common configuration for maintaining continuous feed processing), the cells are never in exactly the same state. Small differences in electrode material properties, piping hydraulics, and prior operating history mean each cell has a slightly different state of charge trajectory, different fouling level, and different effective capacity at any given moment.
A fixed-cycle schedule treats all cells as if they were identical, running each on the same timing protocol. In practice, the cell with the most fouling accumulation will reach its adsorption saturation point before the others, meaning it is running in the late-stage adsorption phase (with its higher competing cation co-intercalation risk) while the healthier cells are still at optimal loading rates. A schedule-based rotation that treats all cells equally misses this inter-cell variation.
The closed-loop approach assigns an independent state estimate to each cell and uses that estimate to trigger the cycle transition for each cell independently, rather than using a shared timer. The operational benefit: each cell runs an adsorption phase appropriate to its current condition. The cell with higher fouling terminates adsorption slightly earlier (protecting both purity and the remaining active surface area) while the healthier cells run slightly longer. The net stack throughput is maintained, but the distribution of work across cells is more even, extending the effective service life of the most heavily used cells.
Sensor-to-Actuator Loop Latency
The loop latency of the control system, the time from when a sensor detects a change to when the electrode setpoint has been adjusted in response, matters because electrode behavior can change faster than slow control loops can follow. The relevant timescale for electrode cycle control is on the order of 2 to 10 minutes per cycle phase, not hours. A control loop that takes 15 minutes from detection to actuation is not useful for managing a cycle phase that only lasts 8 minutes.
Our target loop latency for the adaptive setpoint control function is under 90 seconds from inlet sensor reading to updated electrode setpoint. That is achievable with inline sensor placement at the cell inlet, direct digital communication to the electrode power supply over the industrial Ethernet connection to the existing PLC or DCS, and a control model that runs its computation in under 10 seconds. The 90-second figure includes the typical 60 to 70 second response time of an ion-selective electrode sensor returning a stable reading after a composition step change in the feed, which is the dominant latency component in the loop.
The feedforward component of the control system, which uses inlet sensor readings to predict what the electrode will see after the hydraulic transit time, partially decouples the control loop latency from the transit delay. By the time the new brine composition arrives at the electrode, the setpoint adjustment has already been in place for most of the transit time. The residual latency that matters is only the time from when the inlet sensor detects the change to when the feedforward model computes and applies the setpoint change, not the full transit time.
Integration with Existing SCADA Infrastructure
The EELI control system is designed to sit as a decision layer on top of existing plant SCADA or DCS infrastructure, not to replace it. Plant operators at DLE facilities already have SCADA systems managing their pumps, valves, flow meters, and process alarms. Replacing that infrastructure with a new control system introduces project risk and vendor dependency that operators are understandably reluctant to accept.
The integration model is: existing SCADA continues to handle the operational logic it already handles (valve sequencing, pump control, safety interlocks, process alarms). The EELI system reads process data from the SCADA historian or directly from sensor field buses via OPC-UA or Modbus TCP, runs the electrode optimization model, and writes setpoint recommendations back to the electrode power supply setpoints in the SCADA system. From the plant operator's perspective, the setpoints in the SCADA are being updated more frequently and more intelligently than they were before, but the SCADA interface they work with day-to-day does not change.
This architecture also means the EELI system can be deployed in a monitoring-only mode first (reading data, logging model outputs, but not writing setpoint changes), which allows the operator to validate model predictions against plant behavior before committing to closed-loop operation. That commissioning path substantially reduces the risk of the first deployment compared to a system that starts in closed-loop mode from day one.