With the profound changes in transportation and energy, the integration of new energy electric vehicles into the power grid will generate a large amount of data. Sensors are deployed in the coupling environment of a transportation network and a power grid to transmit accurate monitoring data. Aiming at sensors that generate faults under the coupling interaction between a distribution network and a transportation network, in this paper, we propose a fault sensor node judgment method based on the spatial-temporal correlation fusion algorithm (FA). First, the cubic exponential smoothing (CES) algorithm of the time attribute and the piecewise least squares (PLSE) algorithm of the spatial properties are used to predict the temperature, humidity and voltage data monitored by the sensors. Then, according to the error size, the adaptive weight adjustment method is used to find the optimal weight value, and the FA model is obtained, so as to gain more accurate detection results. Finally, by comparing the predicted value with the set confidence interval, the identification of the fault sensor node is demonstrated. The results showed that the detection model proposed in this study has excellent fault sensor node detection performance. For the prediction results of the temperature data of the sensor, the fit accuracies of FA are 45.1 and 77.4% higher than those of ES and PLSE, respectively, which has certain practical significance.