Grey-related least squares support vector machine optimization model and its application in predicting natural gas consumption demand

Predicting energy demand is of great significance for governments to formulate energy policies and adjust industrial structures. Energy data, such as demand of natural gas, are small samples. In this paper, data with small sample size, nonlinearity, randomness and fuzzy influence factors are considered. A least squares support vector machine model based on grey related analysis (GRA-LSSVM) is proposed, and weighted adaptive second-order particle swarm optimization algorithm (WASecPSO) is designed to optimize the model's parameters. The second-order particle swarm optimization (SecPSO) algorithm updates particles velocity and position weights dynamically, which can balance global search ability and local improvement, and further improve the accuracy of optimization. In addition, the GRA-LSSVM optimized by the WASecPSO algorithm predicts the annual consumption of natural gas in China. The results show that GRA-LSSVM has better generalization ability and training effect, and GRA-LSSVM optimized by WASecPSO algorithm has higher prediction accuracy than PSO algorithm and SecPSO algorithm optimized model.