AbstractIn recent proposals for next generation superconducting tokamaks, such as the ITER project, the nuclear burning plasma is confined by magnetic fields generated from a large set (up to 100 GJ stored energy) of superconducting magnets. These magnets suffer heat loads in operation from thermal and nuclear radiation from the surrounding components and plasma as well as eddy currents and AC losses generated within the magnets, together with the heat conduction through supports and resisitive heat generated at the current lead transitions to room temperature. The initial cryoplant for such a tokamak is expected to have a steady state capacity of up to about 85 kW at 4.5 K, comparable to the system installed for LHC at CERN. Experimental tokamaks are expected to operate at least initially in a pulsed mode with 20–30 short plasma pulses and plasma burn periods each day. A conventional cryoplant, consisting of a cold box and a set of primary heat exchangers, is ill-suited to such a mode of operation as the instantaneous heat loads can amount to over three times the average over a plasma pulse and even the average heat load over a short pulse can be 2–3 times the long term average. It is not possible to adjust the cryoplant compressors quickly enough to follow the pulse loads and even if they could be adjusted on a longer timescale of 100 s or so, some phases of operation would force the cryoplant to operate in regimes with very low efficiency or where there is a tendancy to instability. The magnets of a superconducting tokamak offer a very large heat storage buffer, with capacity to store heat both in the helium cooling the conductor (where the temperature rise must be strictly controlled) and in the mechanical structures (where larger temperature ranges can be tolerated). In this report, the various modes of tokamak operation are simulated, from the initial phase of non-nuclear plasmas with short duration plasma pulses and high electromagnetic heat loads, to long duration burning plasmas. Control algorithms are derived that show that, if there is some flexibility in the scheduling of plasma pulses, the cryoplant can be designed to provide only the steady state heat load (averaged over a 24-h period). The use of such a control system will significantly reduce the cost of the cryoplant and permits a staged approach to cryoplant construction with limited non-nuclear operation possible with about two thirds of the final plant capacity.