Spacecraft in-orbit operations exhibit characteristics such as prolonged autonomous functioning and complex environmental conditions. Anomaly detection in such situations must satisfy stringent real-time requirements under dual constraints of incomplete prior knowledge and limited resource allocation. Consequently, lightweight unsupervised anomaly detection algorithms are commonly employed, with distance-based methods being a prevalent approach. However, when addressing anomaly detection tasks involving high-dimensional, complex telemetry data, traditional distance-based methods suffer significant performance degradation, which arises from their inability to capture the rich anomaly features inherent in the spatio-temporal coupling of parameters. To address these limitations, a spatio-temporal joint clustering framework is proposed. Building upon the lightweight advantages of conventional distance-based approaches, this framework introduces a novel spatio-temporal distance metric. This metric dynamically perceives and quantifies the temporal dependencies within parameter evolution and the coupling relationships between different parameters. To demonstrate the effectiveness of the proposed framework, the classical inductive monitoring system (IMS) algorithm is enhanced by incorporating the proposed spatio-temporal distance metric. The enhanced algorithm is then validated using both simulated and real operational data from spacecraft power systems. Results demonstrate that while meeting the real-time requirements for spacecraft anomaly detection, the enhanced algorithm achieves a detection balance accuracy improvement of 20.34% and 5.87% over traditional methods on two datasets respectively. The enhanced algorithm effectively distinguishes between normal and abnormal data, providing both substantial theoretical support and a technical pathway for applying distance-based methods to detect complex spatio-temporally coupled anomalies in spacecraft.