Analysis and Prediction Method of Power Supply Cost Based on Prophet-lstnet Combination Model

Abstract

The invention discloses a power supply cost analysis and prediction method based on the Prophet-LSTNet combination model, which solves the deficiencies of the prior art, and includes the following steps: Step 1, obtaining historical daily power supply cost data within a period of time, and inputting the power supply cost data into Prophet model; step 2, decompose the power supply cost data into nonlinear trend component data, seasonal component data and holiday component data through the Prophet model; step 3, carry out feature engineering construction, and obtain multi-dimensional characteristics of power supply cost data; step 4, power supply cost data The multi-dimensional features of the cost data are input into the LSTNet model as parameters, and the nonlinear trend component data is input into the LSTNet model, and the LSTNet model is trained to find the dependence between the multi-dimensional features of the power supply cost data and the nonlinear trend component data, train and update alternately LSTNet model and nonlinear trend component data weight parameters; step 5, predict the power supply cost, and get the prediction result.

Description

technical field [0001] The invention relates to the technical field of distribution networks, in particular to a method for analyzing and predicting power supply costs based on a Prophet-LSTNet combination model. Background technique [0002] Grid power supply cost prediction refers to the analysis of the future form based on past operation and maintenance costs, including labor costs, maintenance and operation costs, marketing and operation and maintenance costs, and other operating expenses, so as to realize the estimation of cost consumption in a specific period of time in the future. Its accurate prediction is of great significance to the State Grid's overall grasp of the cost of use, capital deployment and investment construction. At present, there are few related researches. The main reason is that the grid scale is large, the distribution area is vast, and the operating conditions are cumbersome, which causes the data fluctuation of the grid power supply station to be...

AI Technical Summary

Problems solved by technology

[0006] However, in fact, LSTM cannot capture very long-term sequence relationships, so related researchers designed the LSTNet model to solve this problem. LSTNet includes convolutional components, cyclic neural network components, jumping cyclic neural network components, and autoregressive components, which can capture data. Scale period law, but LSTNet can only mine the sequence features of different periods, and cannot smooth the sequence noise, etc.

Images