DOI QR코드

DOI QR Code

A Model of Four Seasons Mixed Heat Demand Prediction Neural Network for Improving Forecast Rate

예측율 제고를 위한 사계절 혼합형 열수요 예측 신경망 모델

  • Received : 2019.11.01
  • Accepted : 2019.12.09
  • Published : 2019.12.31

Abstract

In this study, a new model is proposed to improve the problem of the decline of predict rate of heat demand on a particular date, such as a public holiday for the conventional heat demand forecasting system. The proposed model was the Four Season Mixed Heat Demand Prediction Neural Network Model, which showed an increase in the forecast rate of heat demand, especially for each type of forecast date (weekday/weekend/holiday). The proposed model was selected through the following process. A model with an even error for each type of forecast date in a particular season is selected to form the entire forecast model. To avoid shortening learning time and excessive learning, after each of the four different models that were structurally simplified were learning and a model that showed optimal prediction error was selected through various combinations. The output of the model is the hourly 24-hour heat demand at the forecast date and the total is the daily total heat demand. These forecasts enable efficient heat supply planning and allow the selection and utilization of output values according to their purpose. For daily heat demand forecasts for the proposed model, the overall MAPE improved from 5.3~6.1% for individual models to 5.2% and the forecast for holiday heat demand greatly improved from 4.9~7.9% to 2.9%. The data in this study utilized 34 months of heat demand data from a specific apartment complex provided by the Korea District Heating Corp. (January 2015 to October 2017).

본 연구에서는 기존 열수요 예측 시스템이 공휴일과 같은 특정 일자의 열수요 예측율이 저하되는 문제점을 개선하기 위해 새로운 모델을 제안한다. 제안된 모델은 사계절 혼합형 신경망 모델(Four Season Mixed Heat Demand Prediction Neural Network Model)로서 열수요 예측율 상승하였고, 특히 예측일 유형별(평일/주말/공휴일) 열수요 예측율이 크게 증가하였다. 제안된 모델은 다음과 같은 과정을 통해 선정되었다. 특정 계절에 예측일 유형별로 고른 오차를 갖는 모델을 선정하여 전체 예측 모델을 구성한다. 학습 시간의 단축과 과도학습을 방지하기 위해 구조적으로 단순화된 서로 다른 4개의 모델을 각각 학습한 후에 다양한 조합을 통해 최적의 예측 오차를 보여주는 모델을 선정하였다. 모델의 출력은 예측일의 24시간의 시간대별 열수요이며 총합은 일일 총열수요이다. 이 예측값을 통해 효율적인 열공급 계획을 수립 할 수 있으며, 목적에 따라 출력값을 선택하여 활용할 수 있다. 제안된 모델의 일일 열 총수요 예측의 경우, 전체 MAPE(Mean Absolute Percentage Error, 평균 절대 비율 오차)가 개별 모델의 5.3~6.1%에서 5.2%로 향상되었고, 공휴일 열수요예측은 4.9~7.9%에서 2.9%로 크게 개선되었다. 본 연구에서는 한국 지역난방공사에서 제공한 특정 아파트 단지의 34개월 분량의(2015년 1월~ 2017년10월) 시간단위 열수요 데이터를 활용하였다.

Keywords

References

  1. Seo, B. S., Shim, S. Y., 2012, Forecasting the Demand for Local District-Heating Systems, Korean Energy Economic Review, Vol. 11, No. 2, pp. 27-55
  2. Baek, J. K., Han, J. H., 2011, Forecasting of Heat Demand in Winter Using Linear Regresson Models for Korea District Heating Corporation, Korean Academy Industrial Cooperation Society, Vol. 12, No. 3, pp. 1488-1494 https://doi.org/10.5762/KAIS.2011.12.3.1488
  3. Magnus Dahl, Adam Brun, Oliver S. Kirsebom, Gorm B. Andersen, 2018, Improving Short-Term Heat Load Forecasts with Calendar and Holiday Data, energies, Vol. 11, No. 7, pp. 1-16
  4. Krzysztof Wojdyga, 2014, Predicting Heat Demand for a District Heating Systems, International Journal of Energy and Power Engineering, Vol. 3, No. 5, pp. 237-244 https://doi.org/10.11648/j.ijepe.20140305.13
  5. Vincent Bakker, Albert Molderink, Johann L. Hurink and Gerard J.M. Smit, 2008, Domestic Heat Demand Prediction Using Neural Networks, 19th International Conference on Systems Engineering, pp. 189-194
  6. Etienne Saloux, Jose A. Candanedo, 2018, Forecasting District Heating Demand Using Machine Learning Algorithms, 16th International Symposium on District Heating and Cooling, Energy Procedia, Vol. 149, pp. 59-68
  7. Kimburm Song, Jinsoo Park, Yun-Bae Kim, Chul-Woo Jung and Chan-Min Park, 2011, Heat Demand Forecasting for Local District Heating, IE Interfaces, Vol. 24, No. 4, pp. 373-378 https://doi.org/10.7232/IEIF.2011.24.4.373
  8. Eriksson, Niclas, 2012, Predicting demand in districtheating systems: A neural network approach, Uppsala University Publications, pp. 43