Abstract: Pressurizer water level is an important monitoring parameter to marine pressurized water reactor for operator to estimate operation transient of the reactor. However, pressurizer often takes on problems of false water level, over-range measurement of water level and the loss of measuring. A method based on support vector regression was used to reconstruct the pressurizer water level according to the coupling relationship between pressurizer water level and other thermal-hydraulic parameters, such as the average temperature between reactor core inlet and outlet, pressure and temperature of pressurizer, coolant inventory of main loop system, and charge and drainage flow. Simulation analysis shows that the method can quickly, accurately and efficiently reconstruct the pressurizer water level signal under normal operating conditions.