At present, there is a key problem in the field verification of electricity meters, which is how to obtain multi-dimensional experimental errors by using the electricity energy meter error from many unidimensional experiments which are independent of each other. In this paper, three independent experimental platforms, including temperature, harmonics and magnetic field, are built, and each platform can control the load parameters of the electricity meter. Then many virtual multi-dimensional experimental intersection are designed, and a semi-supervised learning method is proposed to obtain the electricity meter error under the multi-dimensional condition experiment. Finally, the effectiveness and generalization of the method are verified in a comprehensive experimental platform.