Feed-Forward Control Implementation in DCS Platforms

When Feed-Forward Control Is the Right Choice

Feed-forward control is effective when three conditions are true. First, the disturbance is measurable in real time. Second, the disturbance acts before or simultaneously with its effect on the controlled variable. Third, the process dead time from disturbance to controlled variable is longer than the disturbance ramp rate. Combustion air control in boilers meets all three conditions — fuel flow demand changes are fast, measurable via the fuel flow transmitter, and the oxygen sensor response has a dead time of 8–15 seconds.

Cascaded feedback alone produces O2 swings of ±1.5% under load ramps. Adding feed-forward reduces this to ±0.3–0.5%. However, feed-forward is not appropriate when the disturbance measurement is noisy or unreliable. Apply a first-order filter with a time constant of 2–5 seconds to the disturbance measurement before using it as a feed-forward input.

Lead-Lag Filter Design

The core of a feed-forward design is the lead-lag dynamic compensator. The transfer function is:

G_FF(s) = K_FF × (T_lead × s + 1) / (T_lag × s + 1)

Calculate K_FF from the process gain ratio: K_FF = (K_process_disturbance) / (K_process_manipulated). In a combustion air loop, if a 1% increase in fuel flow demand requires a 0.95% increase in air flow, then K_FF = 0.95.

Identify T_lead and T_lag from step test data. If fuel flow change reaches the burner in 2 seconds and affects O2 in 12 seconds, while an air damper change affects O2 in 8 seconds, the required lead is approximately 12 − 8 = 4 seconds. Set T_lead = 4 s. Set T_lag = the process time constant of the air-to-O2 path, typically 5–8 seconds. Start with T_lag = 6 s and adjust during commissioning.

Implementing in Emerson Ovation

Emerson Ovation uses the function block diagram (FBD) environment for control strategy configuration. The Ovation OCC100 controller library includes a LEADLAG block and a FFWD_ADDER block. Connect the disturbance measurement (fuel flow PV) to the LEADLAG block input. Set the LEAD parameter to T_lead (4 s) and the LAG parameter to T_lag (6 s). Connect the LEADLAG output and the PID output to a FFWD_ADDER block. Set the GAIN_FF parameter to K_FF (0.95).

Configure feed-forward enable/disable logic carefully. Add a LOGIC block that disables the LEADLAG output when the disturbance measurement signal quality is BAD or UNCERTAIN. In Ovation, check the STATUS output pin of the disturbance measurement AI block. When STATUS is not GOOD, set the LEADLAG output to zero through a MUX block. This prevents the Ovation controller from applying a corrupted feed-forward correction.

Implementing in GE Mark VIe

GE Mark VIe uses Toolbox ST application environment. The discrete-time lead-lag equation is:

y[n] = (T_lead / (T_lead + T_scan)) × (x[n] − x[n-1]) + (T_lag / (T_lag + T_scan)) × y[n-1] + K_FF × x[n]

In a 100 ms task frame, for T_lead = 4 s and T_lag = 6 s, the coefficients are: lead coefficient = 0.976, lag coefficient = 0.983. Store x[n-1] and y[n-1] in RETAIN variables to preserve filter state across controller restarts on the Mark VIe UCSC controller.

Use the Mark VIe FFWD_GAIN parameter block to scale the lead-lag output before summing it with the PID output. The Mark VIe PID block has a dedicated FFWD input pin. Connect the scaled lead-lag output to this pin. Mark VIe sums the FFWD input with the PID controller output internally and applies bumpless transfer during mode changes automatically.

Commissioning Validation

• Step 1: Run a disturbance step test with feed-forward disabled. Record the peak PV deviation and recovery time. This is the baseline performance of feedback-only control.
• Step 2: Enable feed-forward. Repeat the disturbance step. Target: peak deviation reduced by at least 50% and recovery time reduced by at least 30%. If improvement is less than 30%, adjust K_FF (+10% if correction insufficient) or T_lead (+2 s if correction peaks too late).
• Step 3: Test the feed-forward signal quality fault handling. Force the AI block quality to BAD in the engineering workstation. Confirm the feed-forward output switches to zero within one controller scan cycle (100 ms maximum).
• Step 4: Document the final K_FF, T_lead, and T_lag values in the instrument data sheet and DCS configuration management system. Record the step test results as the as-commissioned baseline for future performance audits.

Conclusion and Action Advice

Feed-forward control is a powerful complement to PID feedback in processes with fast, measurable disturbances. First, calculate K_FF, T_lead, and T_lag from process step test data before entering any values — guessed parameters produce poor results. Second, implement disturbance signal quality monitoring in both Emerson Ovation and GE Mark VIe to prevent noise injection during transmitter faults. Validate performance with quantified step test data — a feed-forward implementation that does not reduce peak deviation by at least 50% should be re-tuned rather than left in service. Review the feed-forward gain and lead-lag parameters during annual instrument calibration — a K_FF value valid at commissioning may be off by 15–20% after three years of equipment wear.

Author: Guo Peilin is an industrial automation engineer with over 10 years of experience in PLC, DCS, and control systems.