Industrial Problem Solution
Development of demand forecast model for city gas
2020-06-121. Company introduction
It is a comprehensive energy service corporation that supplies city gas throughout Daejeon Metropolitan City and Gyeryong City since its establishment in 1985.
※ In October 2017, the name was changed from Chungnam Urban Gas Co., Ltd. to CNCITY Energy Co., Ltd.
2. Problem Background and Summary
Local city gas suppliers, including CNCTITY Energy Co., Ltd., are selling gas supplied by Korea Gas Corporation after predicting the supply in advance a year ago. When signing a contract with Korea Gas Corp. and changing the contract, there is a problem that the supply must be accurately predicted at least two months ago.
Precise demand forecasts are needed to improve the limits of different meter dates and long-term temperature forecasts for different generations.
The goal is to establish a more accurate demand forecasting model by analyzing the effects of sudden variables such as climate change, recession, and population fluctuations on gas demand, along with verifying the adequacy of existing data analysis (linear recovery analysis).
3. Solving Process
For mid- to long-term forecasting, we developed a gas usage prediction model two months later by utilizing LSTM, one of the methods of deep learning.
After receiving data such as sales volume by generation/use and daily temperature-linked supply, it was confirmed that the results of the existing linear return analysis were significant.
4. Ripple effects and future plans
Based on mathematical theories such as fourier series and ARIMA, the daily/monthly gas demand rate is similar to regression analysis, but it is expected to be efficient when supplemented.
Various variables affecting demand will be added and used as a mathematical basis for the actual demand forecast and order quantity calculation by the requesting company.
1. Company introduction
It is a comprehensive energy service corporation that supplies city gas throughout Daejeon Metropolitan City and Gyeryong City since its establishment in 1985.
※ In October 2017, the name was changed from Chungnam Urban Gas Co., Ltd. to CNCITY Energy Co., Ltd.
2. Problem Background and Summary
Local city gas suppliers, including CNCTITY Energy Co., Ltd., are selling gas supplied by Korea Gas Corporation after predicting the supply in advance a year ago. When signing a contract with Korea Gas Corp. and changing the contract, there is a problem that the supply must be accurately predicted at least two months ago.
Precise demand forecasts are needed to improve the limits of different meter dates and long-term temperature forecasts for different generations.
The goal is to establish a more accurate demand forecasting model by analyzing the effects of sudden variables such as climate change, recession, and population fluctuations on gas demand, along with verifying the adequacy of existing data analysis (linear recovery analysis).
3. Solving Process
For mid- to long-term forecasting, we developed a gas usage prediction model two months later by utilizing LSTM, one of the methods of deep learning.
After receiving data such as sales volume by generation/use and daily temperature-linked supply, it was confirmed that the results of the existing linear return analysis were significant.
4. Ripple effects and future plans
Based on mathematical theories such as fourier series and ARIMA, the daily/monthly gas demand rate is similar to regression analysis, but it is expected to be efficient when supplemented.
Various variables affecting demand will be added and used as a mathematical basis for the actual demand forecast and order quantity calculation by the requesting company.