Power load prediction method, system and equipment based on longitudinal federated learning, and medium

Abstract

The invention discloses a longitudinal federated learning-based power load prediction method, system and device, and a medium, and solves the problem of power load prediction on the premise that data of a power company and data of an industrial and information bureau cannot be shared. The industrial information bureau without the label and the electric power company with the label obtain a common sample ID of the two parties through privacy security intersection; the labeled electric power company calculates a first-order derivative and a second-order derivative of each sample, then uses a k-means clustering algorithm to group the samples, calculates the sum of the first-order derivative and the second-order derivative of each group, and uses a homomorphic encryption technology to send the ciphertext of the sample ID and the derivative of each group to the industrial information bureau without the label; calculating approximate derivative information of each sample by the label-free industrial information bureau, calculating gain of each characteristic division and sending a result to an electric power company; and the labeled electric power company sets a splitting point according to the maximum gain. According to the longitudinal federal XGboost algorithm based on homomorphic encryption and clustering, the privacy security of information transmission of the two parties is ensured, and the communication overhead is remarkably reduced.

Description

technical field [0001] The invention belongs to the field of data privacy and security, and relates to a method, system, device and medium for electric load forecasting based on longitudinal federated learning. Background technique [0002] Power load forecasting is a series of forecasting work aimed at power load, including the forecast of future power demand, power consumption and load curve. Power load forecasting can provide reliable decision-making basis for power system planning and operation, which has important practical significance. In the traditional power load forecasting method, a specific machine learning model is usually used to analyze the historical power load and power consumption of the enterprise, and the power load is predicted based on historical data. Power load forecasting is affected by factors including historical power consumption data, meteorological data, corporate benefits, holidays, and major emergencies. The traditional method of only using p...

AI Technical Summary

Problems solved by technology

The root cause of the above problems is the indiscriminate processing when using federated learning technology, but the power load forecasting model is not designed based on the application scenario

Images