After the emerging energy is incorporated into the power system, the harmonic problem of the power system is becoming more and more serious. How to accurately measure electric energy is the key to ensure the stable operation of the power system. Voltage transformer is one of the key equipment of electric energy measurement. Whether its operating state is normal or not plays a vital role in its measurement performance. Therefore, an on-line state monitoring method of voltage transformer based on digital twins is proposed. Select appropriate state characteristic parameters (actual transformation ratio, angular difference, relative capacitance and grounding current coefficient of voltage transformer), and construct the digital twin model of voltage transformer in combination with the requirements of state monitoring. According to the model, the twin data of voltage transformer under different state backgrounds are obtained, and the sensitivity of state characteristic parameters is analyzed to determine the change law of state characteristic parameters. Based on the change law, the state types and corresponding labels of voltage transformer are divided, and the on-line state of voltage transformer is evaluated through network probabilistic neural network (PNN), so as to realize the monitoring of on-line state of voltage transformer. The experimental data shows that the state monitoring parameters of voltage transformer obtained by the proposed method are small, and the accuracy of state monitoring is as high as 90.00%, which proves that the proposed method has better state monitoring performance.