As one of the electrical parameters measuring devices, capacitor voltage transformer (CVT) is widely used in the grid with a voltage of 110kV and above. It provides reliable basis for electric energy measurement, state monitoring and relay protection, thus related to the safety, stability and economic operation of the power system. The existing measurement error evaluation methods are not conducive to timely evaluate the error of CVTs. Therefore, a real-time error evaluation method for CVTs based on same-phase voltage correlation and large-dimensional random matrices (SPVC-LDRM) is proposed. Collect same-phase voltage measurement data of CVTs, extract the error state characteristic parameters for CVTs based on the same-phase voltage correlation, and construct a large-dimensional random matrices model, then the evaluation index and its threshold are calculated to evaluate the error anomaly of CVTs in real time according to single-ring law. The simulation experiment verifies the effectiveness of the method in evaluating the measurement error of the CVTs.