This paper proposes a hybrid system combining Multi-Layer Perceptron (MLP) and Gated Recurrent Unit (GRU) for detecting electricity theft in smart grids. The system is applied to data from the Chinese National Grid Corporation (CNGC) to analyze and solve electricity theft. The proposed hybrid system involves data preprocessing, balancing using k-means SMOTE, and applying both MLP and GRU models to the purified data. The performance of the system is evaluated using various metrics such as accuracy, F1-score, precision, recall, and Matthews Correlation Coefficient (MCC). The results show that the proposed hybrid system outperforms other models like Alexnet, GRU, BGRU, and RNN in terms of accuracy and efficiency. The system is effective in detecting electricity theft due to its ability to handle imbalanced data and extract relevant features from both time and frequency domains. The proposed model is also efficient in terms of execution time, with the hybrid MLP-GRU model performing better than other benchmark models. The results demonstrate that the proposed hybrid system is a promising solution for detecting electricity theft in smart grids.This paper proposes a hybrid system combining Multi-Layer Perceptron (MLP) and Gated Recurrent Unit (GRU) for detecting electricity theft in smart grids. The system is applied to data from the Chinese National Grid Corporation (CNGC) to analyze and solve electricity theft. The proposed hybrid system involves data preprocessing, balancing using k-means SMOTE, and applying both MLP and GRU models to the purified data. The performance of the system is evaluated using various metrics such as accuracy, F1-score, precision, recall, and Matthews Correlation Coefficient (MCC). The results show that the proposed hybrid system outperforms other models like Alexnet, GRU, BGRU, and RNN in terms of accuracy and efficiency. The system is effective in detecting electricity theft due to its ability to handle imbalanced data and extract relevant features from both time and frequency domains. The proposed model is also efficient in terms of execution time, with the hybrid MLP-GRU model performing better than other benchmark models. The results demonstrate that the proposed hybrid system is a promising solution for detecting electricity theft in smart grids.