关键词: 火灾探测, 粒子群算法, BP神经网络, GA-BP, 极限学习机

Abstract: In order to improve the accuracy of fire detection and avoid the standard ELM falling into local optimization, this paper constructs a fire detection model based on the fire characteristic value CO concentration, smoke concentration and temperature, and optimizes the ELM input layer and the hidden layer weight and bias through PSO. The best optimal value is used to train the extreme learning machine network, and the trained network is used to predict the test samples and verify the effectiveness of the method. The study shows that the mean square root error (RMSE) of PSO-ELM is 1.403%, the average absolute error (MAE) is 1.055%, and the average absolute percentage error (MAPE) is 1.183%. Compared with BP, GA-BP and ELM models, the algorithm accuracy and generalization ability are obviously improved. At the same time, PSO-ELM model training speed is faster, can improve the fire detection ability more efficiently.

Key words: fire detection, particle swarm algorithm, BP neural network, GA-BP, extreme learning machine