Existing detection methods cannot accurately locate a false data injection attack (FDIA). Thus a location detection method based on a hybrid chimp optimized extreme learning machine (ELM) is proposed for FDIA in a cyber-physical power system. First, an ELM is used as a classifier to extract the features of power data and detect the attacked state of each bus in the system. Then, a hybrid chimp optimization with global search and faster speed of local convergence is adopted to optimize the number of hidden layer neurons of the ELM. Thus, a detection method is established to realize the accurate location detection against FDIA. This is conducive to the implementation of subsequent defense measures. Finally, a large number of simulation experiments are carried out in IEEE14 and IEEE57 bus systems. The results show that the proposed method has better accuracy, precision, recall and F1 score. This means this method can carry out more accurate location detection against FDIA.