This paper presents a modeling approach which can effectively improve meter accuracy. This model, which combines the improved gray system and particle swarm optimization, uses a small amount of the instrument known error data as training data to get the full-scale error fitting model. Then the model can be used to amend all values in the range. The system has the advantages of economy, practicality and high accuracy proved by experiments.