In power grid supply chain networks, anomalous states of nodes often propagate downstream along the network. Specifically, when some nodes operate abnormally (e.g., due to equipment failures or delivery delays), the downstream nodes may also exhibit abnormal states as a result of this influence. The complexity of power grid construction and maintenance leads to intricate inter-node dependencies, making it challenging to accurately identify the source nodes that trigger downstream anomalies. Moreover, the time at which node anomalies are detected and reported often lags behind the actual occurrence, making it unreliable to infer source nodes based solely on the temporal order of observed anomalies. In this work, we propose a novel method for inferring anomalous source nodes in power grid supply chain networks that does not rely on the exact timing of anomalies. Instead, the method infers the most likely source nodes causing downstream anomalies solely based on whether nodes exhibit abnormal states. Specifically, we first define the rules of anomaly propagation among grid supply chain nodes and construct a probabilistic model describing the influence of potential source nodes on the occurrence of anomalies in other nodes. The method then computes the probability that observed anomalies are caused by each potential source node. Finally, the top-k nodes with the highest probabilities are identified as the anomalous source nodes. Experiments on both synthetic networks and real-world power grid supply chain networks demonstrate that, compared with baseline methods, the proposed approach achieves significantly higher inference accuracy.