Cable Overheating Localization Using Gas Sensor Array With M-Mamba Network

Low-voltage cables are widely used in power transmission and distribution systems. However, overload or high environmental temperatures can cause excessive operating temperatures, leading to insulation material aging and potential fire hazards. Traditional temperature measurement methods suffer from issues such as contact measurement, high costs, and monitoring blind spots. To overcome these limitations, this paper proposes a cable overheating temperature prediction and localization model based on machine olfaction. The method utilizes a distributed array of gas sensor units, which can be practically deployed at intervals of 10 meters, to detect responses to cable pyrolysis gases. These responses are then processed by an M-Mamba network. The network predicts temperature with a mean error of 6.57°C and distance with a mean error of 0.026m. This approach outperforms common neural networks, with lower computational resource usage. Experiments with random distances and overcurrent also demonstrate its high generalization performance.