Trend Analysis of Power Quality Basing on Deep Belief Network

Accurate power quality trend prediction provides basic guarantee for power quality dynamic evaluation and early warning, which is conducive to the safe and stable operation of the new power system. Aiming at the problems of many correlation indicators, few related research and low prediction accuracy of power quality trend prediction, a power quality trend prediction method for energy storage users based on refined model input and deep belief network is proposed. By mining the correlation between multi type with correlation indicators and the trend change sequence of power quality at energy storage users through information gain, the vertical refinement of prediction model input is realized on a spatial scale; By constructing the similarity set, the multi day correlation indicator with the highest similarity within the day to be predicted and the historical data of power quality with trend change are selected to update the input data sequence of the prediction model, and the horizontal refinement of the input of the prediction model is realized from the time scale. Finally, the deep belief network is used to establish the trend prediction model of power quality, and the whale optimization algorithm is used to optimize the key parameters of the prediction model to obtain a better prediction results with dynamic adjustment of the input of the prediction model. The example simulation shows that the proposed prediction method has better accuracy and high universality.