A two-stage industrial load forecasting scheme for day-ahead combined cooling, heating and power scheduling

Authors	Sungwoo Park, Jihoon Moon, Seungwon Jung, Seungmin Rho, Sung Wook Baik, Eenjun Hwang
Title	A two-stage industrial load forecasting scheme for day-ahead combined cooling, heating and power scheduling
Publication	Energies
Volume	13
Issue	2
Pages	x
Year	2020
DOI	https://doi.org/10.3390/en13020443

Introduction

Background

• There are growing concerns about environmental problems caused by carbon dioxide emissions and energy shortage problems.
• Smart grid technologies are gaining attention as they help to solve these problems by enabling more efficient use of energy.
• A smart grid is an intelligent power grid that combines information and communication technology with the existing power grid.

Previous Research

• Many previous studies have suggested a two-stage STLF (Short-Term Load Forecasting) model that uses LR (Linear Regression) in the second stage.
• However, there are deficits in the linearly combined model as it can ignore potential nonlinear terms and give poor results when there is a strong nonlinear relationship between predictors and outcomes.
• Despite sufficient studies on electric load forecasting models, there are not many cases of configuring a power system in conjunction with CCHP (Combined Cooling, Heating, and Power).

Proposed Model

• A novel two-stage STLF scheme based on nonlinear combination of forecasting methods is proposed.
• In the first stage, two STLF models are built using XGBoost (eXtreme Gradient Boosting) and RF (Random Forest).
• In the second stage, a DNN (Deep Neural Network)-based STLF model is built to combine the predicted values from the first stage.
• An economic analysis-based operation scheduling scheme for CCHP is proposed to effectively utilize the results of the STLF.

Significance

• The proposed model addresses the limitations of previous studies and provides a bi-directional benefit to power suppliers and users.
• The results of the economic analysis show the impact of electric rates and contract demand on the efficiency of CCHP.
• The model focuses on the features of the Korean power system and aims to improve energy efficiency in one of the highest energy consumption countries.

Proposed Model

Overall system architecture

• The proposed STLF model consists of two stages: constructing two STLF models and combining them using a DNN.
• In the first stage, two STLF models are built based on XGBoost and RF, which are tree-based ensemble models in time series prediction.
• The XGBoost and RF models are chosen because of their better predictive accuracy and highest correlation with actual power consumption compared to other boosting and Bagging (Bootstrap Aggregating) algorithms.
• In the second stage, the results of the two STLF models from the first stage are combined using a DNN model that takes into account time factors, weather data, historical electric energy consumption data, and electric rate as input variables.
• The DNN model uses the SELU(Scaled Exponential Linear Unit) function as an activation function and has five hidden layers, with the number of neurons in the hidden layer set by two thirds of the number of input variables.

• Algorithm shows the generation of an operational schedule for maximizing annual economic benefits.
• The schedule tells how much energy should be generated by CCHP and how much energy should be supplied by the public power system for each scheduling hour.

Experiment

CVRMSE comparison for each month

MAPE comparison for each month

Results of Wilcoxon test and Friedman test

• The proposed model showed better prediction performance than other forecasting models according to the CVRMSE (Coefficient of Variation of the Root Mean Squared Error) and MAPE (Mean Absolute Percentage Error) performance indicators.
• The proposed model was proven to be superior to other models through the Wilcoxon test and Friedman test.

Annual economic benefit of each electric rate based on contract demand

Example of CCHP operation scheduling

• Five different electric rates were compared in the experiment.
• "Industrial service (A) II / high voltage A / Option II" electric rate with 160 kW contract demand can make the highest annual economic benefit.
• Using this electric rate, economic benefits of more than USD 14,000 annually can be achieved by using CCHP with the public power system.