PURPOSE: To eliminate inaccuracy of a simulator due to a calculated value at the time of designing by making a learning part which learns system characteristic values of a simulator part from measurement data start learning when the state signal of a plant meets specific requirements.

CONSTITUTION: The plant state predicting device 1 has the simulator part 2, the learning part 3 which learns the system characteristic values used for simulation, a model data part 4 which stores a simulator model expression and model data used for the simulation, a process signal input part 5 which inputs the process signal of the plant, a signal input part 6 for a digital signal indicating the state of equipment or a system, and a display and input device 7. Then the learning part 3 learns the system characteristic values based on the measurement data of the plant system and the actual plant state. Consequently, assumptions of calculation, inaccuracy of a calculated value, and influence from other systems are excluded and the system characteristic values matching the actual state can be set.