The invention relates to a short-term forecast method and system for the temperature of a photovoltaic assembly. According to the invention, meteorological element forecast data of a specified location is generated by using a mesoscale weather research forecast (WRF) model, and the meteorological element forecast data such as the temperature, relative humidity, wind speed, wind direction and totalradiation is calculated based on a model result; a historical time sequence of meteorological element data at the location of the photovoltaic assembly and temperature data of the photovoltaic assembly at the corresponding location, time period and resolution are collected, a multi-dimensional Kalman filtering short-term forecast model is built by taking the total radiation, temperature, relativehumidity, wind speed and wind direction at the moment t as multi-dimensional Kalman filtering factors and taking the assembly temperature at the moment t as an output factor, and training is performed on the model; the day-ahead forecast data of the meteorological elements at the location of the photovoltaic assembly is taken as the input of the multi-dimensional Kalman filtering short-term forecast model, and the output of model is a temperature value of the assembly. The technical scheme of the invention obtains high forecast precision.