Application of a grey combined prediction model with weighted compensation

The real-time response and the rapidity are the special requirements for an online prediction system. The original data sequence of this system may have some features, such as a few sample data, non-index, and strong randomness, which bring some difficulties to the online prediction. In this paper, a solution is proposed to improve the shortcomings and deficiencies of a traditional grey prediction model. This solution is built based on the optimized Verhulst model. It is also combined with some other methods, such as the normalization, residual error correction, dynamic weighted compensation, to enhance the reliability and the prediction accuracy in a medium-long term forecasting. Finally, this model was applied to a simulated short-circuit situation in a low voltage distribution system to achieve the real-time online prediction. The comparative tests of several different kinds of models are given. And the results show the practicality and effectiveness of this proposed model.