Non-invasive load identification method based on collaborative training and test system thereof

Abstract

The invention discloses a non-invasive load identification method and a test system thereof. Based on wavelet design and data mining cooperative training, through a pair of complementary algorithms and in combination with a decision tree model and a k-nearest neighbor classifier, cooperative training is performed on a transient current signal data set to classify the transient current signals in the data set, and types of loads are predicted. Therefore, the following beneficial effects are achieved: 1, the wavelet transform can better locate the waveform characteristics of the signal to be identified, has the characteristics of multi-resolution, and has the ability to characterize the local characteristics of the signal in the time domain and frequency domain, which is conducive to extracting the effective characteristic information of the load; 2, the algorithm based on k-nearest neighbor classifier and decision tree model not only reduces the computational complexity, but also improves the prediction accuracy; 3, with the increase of load, the method has the advantages of short computation time and high reliability.

Description

technical field [0001] The invention belongs to the field of smart grid and signal processing, in particular to a non-invasive load identification method based on collaborative training and a test system thereof. Background technique [0002] Non-intrusive load identification technology is a kind of electricity consumption information obtained by users at different levels of fineness at outdoor distribution boards or smart meters. It is an important research direction in the field of intelligent electricity consumption. Establishing an accurate time-varying load model is of great significance for power grid companies to achieve refined demand-side management, accurate load forecasting and analysis of the dynamic characteristics of active distribution systems. [0003] The key to solving the non-intrusive load identification technology is to analyze the transient or steady-state signals extracted during the use of electrical appliances. Extensive and in-depth research and di...

AI Technical Summary

Problems solved by technology

Such as short-time Fourier transform, S transform, wavelet transform and other technologies, but they all have their own shortcomings: (1) Although Fourier transform has the frequency information of the signal, it cannot determine the time when these frequency signals appear, resulting in the loss of time localization information , and can not reflect the characteristics of high frequency and low frequency at the same time, there are limitations; (2) Using the S-transform method to extract features for non-intrusive load monitoring, the detection accuracy is high, and the classification is relatively accurate, but the S-transform has a large amount of calculation , the real-time performance is difficult to guarantee; (3) wavelet transform can be better positioned to identify the waveform characteristics of the signal, has the characteristics of multi-resolution, and has the ability to characterize the local characteristics of the signal in both the time domain and the frequency domain, but the traditional practice Wavelet calculation is slow and time-consuming

However, in the case of data without class labels, supervised machine learning algorithms usually cannot predict their labels, which may lead to misclassification of one or more loads in NILM applications, which is highly dependent on manual intervention and not practical.

Images