Control-System Purpose, Feedback and Controller Modes

In power and industrial heating plants, maintaining process variables—such as boiler drum level, main steam header pressure, deaerator storage temperature, and fuel-to-air combustion ratios—at precise operational targets requires automated control loops. Operating engineers must understand how control signals are processed, how different controller mathematical modes respond to process upsets, and how final control elements physically regulate mass and energy flows under normal and emergency conditions.

Fundamentals of Industrial Control Loops

An automated control loop consists of five core elements working in continuous coordination:

1. Process: The physical operating system containing the fluid or energy being regulated (e.g., a boiler steam drum, feedwater heater, or deaerator).
2. Primary Sensing Element & Transmitter: Measures the condition of the Process Variable (PV)—such as pressure, temperature, level, or flow—and converts it into a standardized electronic or pneumatic transmission signal.
3. Setpoint (SP): The desired target value at which the process variable is to be maintained, established by the operating engineer or programmed by supervisory plant control logic.
4. Controller: An electronic microprocessor or pneumatic mechanism that compares the measured PV against the target SP, calculates the magnitude and direction of the Error ($e$), and executes a mathematical control algorithm to determine the required Controller Output (CO):

$$e = \text{SP} - \text{PV} \quad \text{(or } e = \text{PV} - \text{SP} \text{ depending on loop action)}$$

1. Final Control Element (FCE): The physical mechanical device that directly alters the process energy or mass flow in response to the controller output signal. In thermal power plants, the FCE is almost universally an automated control valve, though variable frequency drives (VFDs) and combustion air dampers are also common.

Open-Loop vs. Closed-Loop Control

• Open-Loop Control: A system in which the control action is completely independent of the process output. The controller executes an action (e.g., an operator opening a manual fuel valve by $25\%$ or a timer running a sootblower for 10 minutes) without measuring whether the desired process outcome was actually achieved. Open-loop systems cannot compensate for external load disturbances.
• Closed-Loop (Feedback) Control: A continuous, self-correcting system in which the process variable is constantly measured, fed back to the controller, compared with the setpoint, and corrected. If a disturbance alters the PV, the resulting error drives the controller to adjust the final control element until the error is eliminated.

Feedback vs. Feedforward Control

• Feedback Control (Reactive): The traditional control architecture. The controller only takes action after an external disturbance has already entered the process and caused a measurable deviation between the PV and SP. While robust and capable of correcting unexpected disturbances, feedback control inherently permits transient errors during load changes.
• Feedforward Control (Predictive): Measures incoming process disturbances before they alter the primary process variable. Using a mathematical model of the process, the feedforward controller calculates the exact corrective action needed and adjusts the FCE simultaneously with the disturbance arrival. For example, in three-element boiler drum level control, an increase in steam flow (demand disturbance) immediately commands an increase in feedwater valve position while level feedback trims inventory errors, including misleading shrink/swell effects. Because feedforward cannot correct for internal model inaccuracies or unmeasured disturbances, power plants combine feedforward predictive action with feedback trim.

Controller Action Modes: P, I, and D

Controllers use mathematical algorithms to translate error into output. The four standard control modes are On/Off, Proportional (P), Integral (I), and Derivative (D).

1. On/Off (Two-Position) Control

The simplest control mode. The controller output switches abruptly between two discrete states: fully energized ($100\%$) or fully de-energized ($0\%$):

• Differential Gap (Deadband): To prevent rapid, destructive cycling (chattering) of contactors or valves when the PV hovers near the setpoint, an intentional deadband is built into the switch. For example, a commercial heating sump pump may switch ON when water rises to $1.2\text{ m}$ and switch OFF only when level drops to $0.6\text{ m}$.
• Applications: Building heating thermostats, air compressor pressure switches, and boiler low-water fuel cutoffs (where safety demands positive emergency shutoff rather than throttling).

2. Proportional (P) Control

In proportional control, the controller output is linearly proportional to the instantaneous magnitude of the error:

$$\text{CO} = K_p \cdot e + b = \frac{100}{\text{PB}} \cdot e + b$$

Where $K_p$ is the controller proportional gain, $\text{PB}$ is the Proportional Band (in percent), $e$ is error, and $b$ is the manual bias (output at zero error, typically $50\%$).

• Proportional Band (PB): Defined as the percentage change in the process variable required to drive the controller output through its full range ($0\%$ to $100\%$):

$$\text{PB (\%)} = \frac{100}{K_p}$$

• Narrow PB (High Gain): Small error causes large valve movement. Response is fast, but if PB is too narrow, the loop becomes unstable, resulting in continuous hunting and severe oscillation.
• Wide PB (Low Gain): Large error causes small valve movement. Response is generally less aggressive; stability still depends on process dynamics and tuning.
• The Offset (Droop) Phenomenon: The fundamental operational limitation of pure proportional control is offset (permanent steady-state droop). When a plant load change occurs (e.g., steam demand increases from $50\%$ to $80\%$), the boiler feed valve must remain open at an expanded steady-state position to balance the higher mass flow. However, by definition ($CO = K_p \cdot e + b$), the controller can only produce an output different from $50\%$ if a continuous, non-zero error persists. Therefore, many self-regulating processes using proportional-only control do not return the process variable completely to setpoint following a sustained load change.

3. Integral (I) Control (Reset Action)

Integral control eliminates offset by continuously driving the controller output at a rate proportional to the time integral of the error:

$$\text{CO}_I = K_i \int e(t)\,dt$$

Here $K_i$ is the integral gain. In the ideal PID form, $T_i$ is integral time, expressed as time per repeat; some instruments instead use reset rate, its reciprocal, in repeats per unit time. Integral action accumulates error. In a stable, correctly configured loop with sufficient output authority, it can remove steady offset from a constant disturbance; saturation, deadband, faults and dynamics can prevent exact setpoint tracking.

• Reset Windup Hazard: If a massive process upset occurs or a system is started from cold, the PV may be physically unable to reach the SP for an extended duration (e.g., an oversized load change where the control valve is forced wide open to $100\%$). The integral algorithm continues integrating the uncorrected error, "winding up" the internal calculation to mathematical saturation. When the PV finally crosses the setpoint, the controller output remains pinned at $100\%$ until the error of opposite polarity slowly unwinds the accumulated integral sum. This causes extreme overshoot and potential high-pressure safety valve lifting. Many controllers use anti-windup strategies such as limiting integration or tracking actual output; implementation and configuration vary.

4. Derivative (D) Control (Rate Action)

Derivative control responds to the rate of change of the error, rather than its magnitude:

$$\text{CO} = K_d \frac{de}{dt} = T_d \frac{de}{dt}$$

Where $T_d$ is the derivative time constant. If the process variable begins moving rapidly away from setpoint, derivative action immediately produces a large opposing output spike to apply dynamic braking. As the error trend slows, derivative output drops to zero. Derivative control cannot eliminate offset and is normally combined with other modes for ordinary process regulation.

• Benefits & Limitations: Derivative action adds phase lead and damping, stabilizing loops with long thermal or dead-time lags (e.g., superheater steam temperature attemperation). However, derivative action is hyper-sensitive to electrical noise or fluid turbulence; high-frequency measurement noise generates violent valve hunting. Derivative action is commonly avoided on fast noisy flow loops unless a justified design and filtering arrangement calls for it.

5. Combined PID Three-Term Control

Combining all three modes provides complete, robust loop regulation:

$$\text{CO}(t) = K_p \left( e(t) + \frac{1}{T_i} \int_0^t e(\tau)\,d\tau + T_d \frac{de(t)}{dt} \right) + b$$

• P (Proportional): Provides immediate corrective effort proportional to current error.
• I (Integral): Continuously pushes output to eliminate steady-state offset.
• D (Derivative): Anticipates future error trajectory and damps overshoot.

Source notes

Official scope and technical references checked 2026-10-17: Reference 1; Reference 2. Archived DOE handbooks are used for physical principles, not current Canadian legal requirements.

For the ideal PID form shown, Kᵢ = Kₚ/Tᵢ. Controller manufacturers also use parallel and other forms, so gain and reset settings must be interpreted with their units and algorithm. PB = 100/Kₚ assumes error and output are both expressed on compatible percentage spans. The equation does not authorize changing a plant tuning constant without the approved process.