With the digital transformation and development of the petrochemical industry, power quality issues are becoming increasingly serious. Due to production needs, a large number of nonlinear, asymmetric, and impactful power loads are connected to the power grid, resulting in a series of power quality disturbances that affect the safe operation of the petrochemical industry. On this basis, a deep-learning-based method for identifying power quality disturbances in the petrochemical industry is proposed. A multi-scale temporal convolutional network (TCN) is used to extract global and local features of power quality time-series signals, and a multi-head self-attention mechanism is introduced to highlight important features and improve the recognition performance of the model. Based on simulation data and power grid data from a petrochemical industry in a certain region, the experimental results show that the proposed method can achieve high-precision identification of single and composite power quality disturbance signals, and has strong anti-noise and generalization capabilities.

Petrochemical industry  /  power quality disturbances  /  deep learning  /  multi-scale TCN  /  multi-head self-attention