Abstract

With the fast development of natural gas in China, short-term gas load forecasting plays a vital role in gas companies, which guides gas purchases. Gas usage is usually affected by various factors such as temperature, humidity, holidays, etc., which bring instability and randomness to gas consumption and increase the difficulty of gas load forecasting. This paper introduces a short-term gas load forecasting method based on Temporal Convolutional Network (TCN) and a Bidirectional gated recurrent unit (BiGRU) neural network. This paper adopts the sliding window method. The sliding window has a fixed length. It is used to process gas data, combining Temporal Convolutional Network and Bidirectional gated recurrent unit. Finally, a fully connected layer is applied to export the load prediction results. The proposed method is employed to predict daily gas consumption based on the real data and compared with the most advanced way. The result shows that the prediction error of the developed method is the lowest, thus outperforming the other methods. The method can be also applied to learn other amount of energy such as wind and electric energy.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call