The rapid acceleration of digital transformation has increased the need to maintain the operational integrity of data centers. Conventional monitoring methods rely heavily on isolated threshold-based alarms, often failing to provide a holistic or predictive view of the system’s health. By introducing a system-level health index model that integrates multisource data by fusing internal uninterruptible power supply operational logs with external wireless sensor networks, we develop a hierarchical evaluation method using power system health, environmental health, and equipment status as important criteria. To normalize heterogeneous sensor dimensions, a Gaussian mapping function is applied to score industry-standard optimal values. Weight coefficients are derived using the analytic hierarchy process. The results show that power system health (weight of 0.539) and average temperature (0.193) are the most influential factors of systemic stability. The model developed effectively detects transient anomalies, such as a 10 V voltage sag accompanied by an 80 kW load surge, through a sensitive decrease in the health index. The model exhibits strong agreement with expert assessments (𝑅 = 0.96) and shows a low mean absolute error (0.04). The results of this study contribute to the development of a multimodal sensor deployment strategy and provide a quantifiable, data-driven tool for proactive facility management. Whereas the developed model’s applicability is limited by subjective expert weighting and transient data gaps, objective entropy weight methods and deep learning architectures, such as long short-term memory, need to be introduced to enhance long-term predictive accuracy and system resilience.