The invention discloses a multi-parameter integration real-time normally-running condition prediction method for a high voltage power supply of a radar transmitter. The method comprises the following steps of: monitoring input voltage, current, output voltage, and current signals of the high voltage power supply in real time and on line, calculating to acquire four parameter values, namely the output voltage, output resistance, output ripples, and power supply efficiency in real time, and constructing a four-dimensional vector which is used as a fault characteristic vector of the high voltage power supply; performing phase-space reconstruction of four parameters by a genetic algorithm and a least square support vector machine algorithm according to historical time series data of the four parameters, namely the output voltage, output resistance, the output ripples, and the power supply efficiency; obtaining four parameter values at certain moment in future and the corresponding fault characteristic vector by a multivariate time series locality predication integration method; and finally after the fault characteristic vector of the high voltage power supply at certain moment in the future is normalized, and calculating to obtain the Euclidean distance between the normalized fault characteristic vector and the standard fault characteristic vector, wherein the Euclidean distance is used for estimating the running state of the high voltage power supply. By the method, the state of the high voltage power supply of the radar transmitter can be monitored in real time, and the fault of the high voltage power supply can be predicted, and accurate and real-time information is provided for estimation of the running state and the on-condition maintenance of the high voltage power supply of the radar transmitter.