Cold load prediction method based on support vector machine parameters optimized with cat swarm algorithm

Abstract

The invention relates to a cold load prediction method based on support vector machine parameters optimized with the cat swarm algorithm. The method includes the following steps: (1) selecting prediction features of a cold load, (2) pre-processing historical data of the cold load, (3) analyzing the historical data of the cold load, (4) automatically optimizing the support vector machine parameters with the cat swarm algorithm, and (5) performing cold load prediction with an optimized support vector machine. The parameters of the support vector machine are optimized through the local searching capability and the global searching capability possessed by the cat swarm algorithm, so that the prediction capability of the support vector machine is promoted, and the effect of promoting prediction accuracy is achieved. Due to the cat swarm algorithm is applied to the optimization process of the support vector machine parameters, automatic optimization of the support vector machine parameters is achieved, and finally the prediction accuracy of the cold load prediction is promoted. The cold load prediction method based on the support vector machine parameters optimized with the cat swarm algorithm is high in practicality and strong in popularization capacity.

Description

Technical field [0001] The invention relates to a prediction method based on data mining, in particular to a cooling load prediction method based on a cat colony algorithm to optimize support vector machine parameters. Background technique [0002] The current cold forecasting techniques are mainly statistical regression forecasting methods. Statistical regression forecasting methods include multiple linear regression methods, exponential smoothing (time series forecasting), grey forecasting, neural networks and support vector machines (SVM). [0003] The cooling load forecasting model based on multiple linear regression technology is better for forecasting 4 hours in advance, but it cannot handle the longer-term forecasting problem well. The seasonal air conditioning exponential smoothing cooling load forecasting model has a low average forecast error during the entire forecast period, but this method is suitable for forecasting the air conditioning load of office buildings. When...

AI Technical Summary

Problems solved by technology

[0006] To sum up, the existing cooling load forecasting methods have the following problems: ①Multiple linear regression forecasting accuracy is not high, and long-term forecasting cannot be done; ②The time series forecasting model requires high stability of the original data, while the actual cooling load is generally None of them meet the requirements of stationarity, resulting in unsatisfactory prediction accuracy; ③The neural network is prone to fall into local minimum points in the cooling load prediction, the generalization ability is poor, and the prediction accuracy is not ideal in practical applications; ④Based on the support vector machine The (SVM) cooling load prediction method can achieve better prediction accuracy than the above methods, especially the prediction method based on genetic algorithm, ant colony algorithm and particle swarm algorithm to optimize the SVM parameters

However, its application in cooling load forecasting still has problems such as low accuracy and high time complexity, so it is necessary to further improve the existing SVM forecasting algorithm

Images