Power equipment temperature prediction method based on PSO-LSSVM online learning

Abstract

The invention discloses a power equipment temperature prediction method based on PSO-LSSVM online learning, and the method carries out the online evaluation and screening of a training sample: firstlytransmitting the collected real-time temperature data of power equipment to a background system, and carrying out the preprocessing of the sampling data; then judging whether incremental learning iscarried out on newly added samples or not according to KKT conditions of an LSSVM model, Meanwhile, rejecting the samples with the maximum characteristic difference in the current training sample setthrough an adjusdesk cosine similarity method; and then, predicting the temperature of the power equipment after the moment t by using an online trained PSO-LSSVM model, and carrying out different levels of high-temperature early warning on the power equipment which may exceed a temperature threshold. According to the method, the contribution of the newly added samples is more emphasized, and thetraining sample set capable of best reflecting the current state is obtained, so that the generalization ability of the model is improved, the prediction operation efficiency is improved, the temperature at the next moment can be effectively predicted, and the safety of using electrical equipment is improved.

Description

technical field [0001] The invention relates to the technical field of power equipment fault monitoring, in particular to a method for predicting the temperature of power equipment based on PSO-LSSVM online learning. Background technique [0002] With the development of modern power systems towards large units, large capacity, and high voltage levels, the requirements for reliability and stability of power supply are also getting higher and higher. Temperature is a very important parameter of high-voltage electrical equipment in operation and needs to be strictly monitored. The information on equipment temperature changes can correctly reflect whether the high-voltage equipment is in a normal state. If the temperature rises too high, it will easily cause equipment damage and cause huge economic losses. Since the heating of equipment is a continuous process, the temperature in the next stage can be predicted through the trend of the collected temperature data, which plays a r...

AI Technical Summary

Problems solved by technology

Since the system is dynamic and time-varying, this method not only cannot obtain the temperature of all electrical equipment in real time, but also is not easy to predict the development trend of equipment temperature

In recent years, some scholars have proposed online methods, but they often ignore the pruning of samples or use the sliding time window method to prune. These methods cannot solve the problem of data redundancy well, resulting in high computing efficiency.

Images