Saroj Dhital Department of Business and Economics University of Wisconsin-Superior FORECASTING COAL CONSUMPTION IN THE UNITED STATES
Dec 24, 2015
Saroj Dhital
Department of Business and Economics
University of Wisconsin-Superior
FORECASTING COAL CONSUMPTION IN THE UNITED STATES
INTRODUCTION
• Coal is the most exclusively used and most abundant fossil fuel in the United States
• Coal Accounts for about 30% of World’s total energy production and consumption
• Coal is the only fuel capable of offsetting any shortage of energy created by petroleum
• Most Coal producing countries will soon be reaching Peak Coal
• Hence, Necessity arises to account for total coal production and consumption
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 11765000
70000
75000
80000
85000
90000
95000
100000
105000
110000
Historical Data
Historical Data
METHODOLOGY
• Collect Data
• Develop Model
• Combine selected Models
• Test for Significance and Errors
• Developing Final Forecast
DATA
• Coal Consumption in the US (Million Btu) - CC
• Electricity Generation by Coal (million Kilowatt Hours) - EG
• Total Inventory of Petroleum and Coal Products (Million USD) - TI
• Cost of Coal Receipts at Electric Plants (USD per Btu) - Cost
• Unemployment Rate - UR
• Decomposed Seasonality Index - DS
DATA STUDIED BUT NOT USED
• Electricity End Use Consumption
• Price Index for Purchasing Fuel
• Gross Domestic Products
• Elasticity Coefficient for Coal Consumption
• Coal Consumption as a percentage of total energy used
• Average Temperatures in various Months in US
• Average Price of Petroleum Products
MODEL
• Winter’s Multiplicative Method
• Ft = αAt + (1-α)Ft-1
• 10% Trend, 10% Seasonality, 10% Cyclical Patterns and 12 Seasonal Cycles
• Multiple Regression• 5 Independent Variables
• CC=β0 + β1*EG + β2*TI + β3*Cost + β4*UR + β5*DS
• Combined Model• Multiples Regression of two above mentioned models, forced through the origin
• Forecast = β1*Regression + β2*Winter’s
OTHER MODEL CONSIDERED
• ARIMA
• Box-Jenkins
• Linear Exponential Smoothing Model
TEST OF SIGNIFICANCEModel F-Stat Significant F DW
Multiple Regression296.2293 1.45E-63 1.699334046
Winter’s Model -- --
Combined Model812.1991 2.49E-69
Electricity Total Inventory Cost of CoalUnemployment
rate Seasonality
Index
Electricity 1
Total Inventory 0.371811949 1
Cost of Coal 0.024545714 0.771339499 1
Unemployment Rate -0.34639723 0.153368528 0.665052144 1
Seasonality Index 0.526544261 0.020218988 -0.0359247 -0.02990803 1
ANOVA TABLE FOR REGRESSIONS
Summary output and ANOVA table for Multiple Regression
Summary Output and ANOVA table for Combined Regression
ERROR ANALYSIS
Error Multiple Regression
Winter’s Method Combined Model
MAD 1504.72 2387.48 1659.93
MPE -0.05% 2.20% -0.05%
FINAL FORECASTDECEMBER 2010: 84234.43 MILLION BTU
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 10110510911311755000
65000
75000
85000
95000
105000
115000
Historical DataFinal Forecast
THANK YOU
ANY QUESTIONS?