Integrating processes accumulate the net differences between inflows and outflows—one of, if not the most challenging, integrating processes is boiler drum level control. The drum level acts as an integrator because it accumulates the net difference between feedwater inflow and steam outflow.

Some of the reasons for this challenge are the inverse response from the shrinkage and swelling effects. A sudden increase in steam demand causes a temporary pressure drop, leading to bubble expansion and a resulting swell, which can produce an increase in level, even when the water mass has decreased in volume.

The opposite occurs with a reduced steam demand, causing shrinkage. A consequential increase and decrease in firing rate for an increase and decrease in steam demand, respectively, causes formation and collapse of bubbles in the downcomer (tubes carrying drum liquid to the bottom for heating), causing a change that is the opposite of the final change in level.

This inverse response can outpace the level controller. An even greater inverse response is seen from level controller action when the feedwater is much colder than the drum level due to insufficient feedwater preheating. A standard PID level controller will take the opposite action from what is required to attain the setpoint level and is consequently detuned, often excessively. To help the drum level controller deal with changes in steam demand, three-element drum level control is used, where a steam flow feedforward is added to the level controller output to proactively change the feedwater flow control setpoint.

However, an increase and decrease in steam demand lead to an increase and decrease in boiler feedwater, causing the collapse and formation of bubbles, respectively. An overreaction to the shrink and swell can be somewhat mitigated by simply adding a delay to the steam feedforward signal. For cases where the feedwater is particularly cold or the drum is undersized, consider including an additional feedforward signal with a lead-lag and opposite sign that enables a temporary correction for shrink and swell.

Given the boiler’s role in the process, deviations can lead to boiler tube overheating or water carryover into turbomachinery, posing safety risks, efficiency losses, and equipment damage.

I examined some of the process automation hall of fame member Greg McMillan’s writings for the recommendations he has provided over the years to address these challenges. In particular, see page 301 in Momentum Press Tuning and Control Loop Performance, Fourth Edition, 2014, and pages 341 and 342 in McGraw-Hill Process/Industrial Instruments and Controls Handbook, Sixth Edition, 2019.

• Implement Three-Element Drum Level Control: This is a core strategy combining feedback and feedforward. The drum level acts as the primary process variable in an outer loop, with steam flow serving as a feedforward signal to preemptively adjust the feedwater flow setpoint (matching water input to steam output). A secondary inner cascade loop controls feedwater flow to eliminate valve nonlinearities and uncertainties. This setup maintains material balance, reduces variance from disturbances, and is effective for about 80% of boiler drums, particularly where swell/shrink effects are moderate. For trimming, use PI control on the level loop with emphasis on proportional action to provide quick corrections. A model predictive control strategy may be helpful for highly undersized boiler drums that exacerbate swell and shrink.
• Incorporate Feedforward for Disturbance Rejection: Forward the steam flow signal directly to the drum level controller to set feedwater demand, allowing proactive response before level deviations occur. For firing rate changes, delay feedforward signals using a deadtime block to synchronize timing—e.g., offsetting swell from increased firing (bubble expansion) with shrink from added cold feedwater. This is crucial for fast steam flow ramps (up to 20-25% per minute) that could otherwise overwhelm the controller. In cases where the feedwater temperature is close to the drum temperature (difference <100°F), pair this with aggressive tuning and aids such as feedwater heaters or economizers to minimize swell.
• Tune for Fast Response and Integrating Dynamics: Prioritize tuning the flow and combustion control loops that support the boiler pressure and level control loops. For the integrating nature of level processes (where imbalances ramp indefinitely), apply Lambda tuning methods: set the integral time based on the arrest time (Lambda) that is the desired time to stop the excursion for reasonable load rejection, using firing rate changes as test disturbances during tuning. This handles inverse response from swell/shrink better than overly conservative settings.
• Enhance Measurement Flow Measurement Performance: Use Coriolis meters for precise and high rangeability steam/feedwater flow measurements. These improvements enhance the signal-to-noise ratio, allowing for tighter control bands and more accurate detection of imbalances. Online diagnostics and compensation in smart devices further reduce maintenance issues in harsh boiler environments.
• Address Configuration and Anti-Windup Issues: Ensure proper PID setup, including scaling and back-calculate (BKCAL) signals in cascade arrangements, to prevent integral windup during mode switches or saturation. This requires experienced tuning to enable bumpless transfers and sustained performance.

These recommendations emphasize proactive control to mitigate the inherent challenges of these boiler dynamics.

You can find many of Greg’s writings on Amazon, ControlGlobal.com, ISA Interchange, and on this blog.