Short-term power load prediction method based on SFO-TSVR

Abstract

The invention relates to a short-term power load prediction method based on SFO-TSVR. The method comprises the following steps: obtaining historical load data, meteorological data and corresponding date type data to serve as original sample data; dividing the original sample data into a training set and a test set; setting an input data sequence, and constructing a TSVR model; based on the training set, training the TSVR model by adopting an SFO algorithm; and based on the test set, verifying the trained TSVR model, and if the verification is passed, inputting related data of an actual to-be-predicted day into the trained TSVR model according to the set input data sequence to obtain a power load prediction value of the actual to-be-predicted day, otherwise, returning to carry out model training again. Compared with the prior art, the method can efficiently and accurately obtain a short-term load prediction result.

Description

technical field [0001] The invention relates to the technical field of power system load forecasting, in particular to a short-term power load forecasting method based on SFO-TSVR. Background technique [0002] Short-term load forecasting is the basic component of power system economic dispatch, and it is also an important guarantee for the safe operation of power system. With the rapid development of social economy, the demand for electric energy in all walks of life is increasing day by day. At the same time, the requirements for the accuracy of power load forecasting are gradually increasing. Improving the accuracy of load forecasting can rationally arrange power production and dispatch operation plans. It also has an important impact on the economic benefits of the power grid. [0003] With the development of computer technology, artificial intelligence and other fields, the current short-term power load forecasting methods mainly include support vector regression, rand...

AI Technical Summary

Problems solved by technology

Among them, the deep neural network has the advantages of strong learning ability and good feature extraction effect, but it has high requirements for data scale and hardware; support vector regression has the advantages of being good at processing small sample data and strong generalization ability, but it has the disadvantages of slow running speed, The disadvantage of relatively low accuracy; in addition, twin support vector regression (twin support vector regression, TSVR) is based on support vector regression, by using two non-parallel hyperplanes to convert a complex convex optimization problem into two simple The convex quadratic programming problem greatly improves the training efficiency and fitting ability of the model. However, twin support vector regression is a parameter-dependent model. The selection of parameters will directly affect the prediction accuracy of the model, and the method of manual parameter adjustment will inevitably Easily affected by subjective experience, traditional intelligent optimization algorithms also have certain limitations in terms of optimization ability and convergence speed

All of the above will affect the accuracy of forecast results and forecast efficiency

Images