The present invention discloses a conditional probability adjustment-based electric energy metering device short-term demand prediction method and belongs to the metering device demand prediction technical field. According to the method of the invention, the demand prediction of electric energy metering devices of a power operation expansion project is adopted as an object; overall demand prediction is performed through using a time series decomposition model based on the historical data of a marketing business system in the past years; the project process data of key business nodes such as metering demand determination and demand actual production in power operation expansion are considered comprehensively; an on-progress project conditional probability adjustment model is constructed according to a conditional probability distribution theory; an obtained preliminary prediction result is adjusted dynamically in real time; and therefore, a more accurate demand prediction result can be obtained. With the conditional probability adjustment-based electric energy metering device short-term demand prediction method adopted, original demand prediction results can be adjusted, optimized and improved in real time according to resent project demand data change, and accurate demand plans can be provided for purchase, the arrival of goods, production, verification, warehousing and distribution.