In lead-bismuth eutectic (LBE) cooled fast reactors, the primary coolant average temperature directly determines the safety margin with respect to the core outlet and cladding temperature limits, and is therefore a key controlled variable for plant operation. A disturbance in the air-cooler air flow rate must propagate through the secondary side and the primary heat exchanger before it affects the primary-loop temperature; the resulting long transport path imposes an inherent lag on single-loop feedback control. To address this problem, a system-level model of the ENEA NACIE-UP LBE test loop, comprising the primary LBE circuit, the pressurized-water secondary loop and the air-cooling circuit, is developed in Modelica, with all model parameters taken from the open literature. Taking the average temperature of the heated section as the controlled variable, an integrated temperature controller is designed in which the proportional gain is derived from the actuator range, the feedforward term follows analytically from the steady-state energy balance of the loop, and a first-order dynamic compensation is introduced to match the thermal inertia of the disturbance channel. Comparative simulations are carried out under identical initial and boundary conditions, with a −40% step reduction of the air flow rate as the disturbance.

The results show that, without dynamic compensation, the feedforward output acts ahead of the thermal propagation of the disturbance, and the maximum absolute temperature deviation increases relative to feedback-only control: with the gain determined from the steady-state energy balance (1.005×10⁵ W/(kg·s⁻¹)), the maximum absolute deviation rises from 0.275 K (feedback-only) to 1.823 K (a factor of 6.62), and the reverse undershoot appears as early as 6.5 s after the disturbance, far ahead of the thermal response of the facility. When the same gain is applied with a first-order dynamic compensation matched to the thermal inertia of the disturbance channel (time constant 300 s), the reverse undershoot is reduced by 78.9% and the maximum absolute deviation falls to 0.384 K. Furthermore, after the feedforward gain is calibrated to 5.8×10⁴ W/(kg·s⁻¹) (58% of the analytical value) according to the measured power variation, the maximum absolute temperature deviation under the −40% flow step decreases from 0.275 K (feedback-only) to 0.178 K, a reduction of 35.2%, with a negligible steady-state error.

This paper reveals the time-mismatch mechanism between the static feedforward and the thermal effect of the disturbance. The proposed method requires only steady-state design data to determine the feedforward parameters, without identification of dynamic transfer functions, and thus provides a practical reference for engineering tuning.