The growth in adoption of communication technologies in the power system results in an increase in its vulnerability towards cyberattacks. This paper presents a novel state partition-particle filter (SP-PF) based detection algorithm that dynamically adapts to varying operating conditions. The algorithm partitions state variables into size-restricted blocks, effectively grouping highly correlated variables to enhance computational efficiency and detection accuracy. Our approach consists of two main steps: (i) state-partition estimation of variables and (ii) detection based on likelihood conditions. The proposed detection algorithm was tested in a real-time cyber-physical environment using a real-time digital simulator (RTDS) in hardware-in-loop configuration with PMUs and a synchronization clock, all connected via standard TCP/UDP protocols. Experimental results demonstrate successful detection of false data injection attacks, replay attacks, and hybrid attacks under various operating conditions. Comparative analysis with extended Kalman filter shows that our approach achieves significantly improved accuracy in state estimation with reduced mean square error, enhancing the overall robustness of the detection mechanism.