Dr. Yukun BaoSchool of Management, HUST
Business Forecasting: Experiments and Case Studies
Dr. Yukun BaoSchool of Management, HUST
Case 3: Load Forecasting
April 18, 2023 Business Forecasting: Experiments and Case Studies
3
Contents
1. Problem Statement
2. Modeling tasks
3. Data Analysis
4. Experimental Results
5. Summary
April 18, 2023 4
1. Problem Statement
Business Forecasting: Experiments and Case Studies
April 18, 2023 5
1. Problem Statement
Load Forecasting Predict the future electric demand based on
historical load, climate factors, seasonal factors, social activities, and other possible factors.
Typical applications Short-term: from one hour to one week ahead
forecasts Medium-term: a week to a year ahead Long-term: Longer than a year Forecasts for different time horizons are
important for different operations within a utility company
Business Forecasting: Experiments and Case Studies
April 18, 2023 6
1. Problem Statement
Benefits of accurate forecasting of Load demand Utilities/ System Operators/Generators/ Power
Marketers/ other participants in electric generation, transmission, distribution, and markets
automatic generation control, safe and reliable operation, and resource dispatch
Energy transaction in deregulated and competitive electricity markets
infrastructure development …
Business Forecasting: Experiments and Case Studies
April 18, 2023 7
1. Problem Statement
Goal of this case study Primary experimental study in day-ahead load
forecast (Short-term Load forecasting) Data
Hourly load and temperature data from North-American electric utility
Forecasting Methods ( by Matlab/R) Support Vector Regression Artificial Neural Network ARIMA ES MA
Business Forecasting: Experiments and Case Studies
April 18, 2023 8
2. Modeling Tasks
Step1: Data Analysis (SPSS/Matlab) Preprocess Visualize and Analysis
Step2: Constructing Model Input features selection Parameters Optimization
Step3: Experimental Results and Analysis Run Model Results and comparison
Business Forecasting: Experiments and Case Studies
April 18, 2023 9
3. Data Analysis (1)
Testing period: January in 1991
Training period: The previous three months hourly data
Preprocess: Zero values [0,1]
Business Forecasting: Experiments and Case Studies
April 18, 2023 10
3. Data Analysis (1)-Descriptive
SPSS:
Business Forecasting: Experiments and Case Studies
Descriptive Statistics
N Range
Minimu
m
Maximu
m Mean
Std.
Deviation Variance Skewness Kurtosis
Statisti
c Statistic Statistic Statistic Statistic Statistic Statistic Statistic
Std.
Error Statistic
Std.
Error
Load 2904 3285.0
0
1350.00 4635.00 2623.7999 616.25958 379775.8
76
.180 .045 -.417 .091
Temperature 2904 54.00 12.00 66.00 42.9490 9.33553 87.152 -.854 .045 .928 .091
Valid N
(listwise)
2904
April 18, 2023 11
3. Data Analysis (1)-ScatterPlot
In SPSS: GraphsLegacy DialogsScatter/Dot…Simple Scatter
Business Forecasting: Experiments and Case Studies
April 18, 2023 12
3. Data Analysis (2)
Hourly load from 01, May,1990 --- 05, July,1990 load demands have
multiple seasonal patterns including the daily and weekly periodicity.
load level in the weekend days and holidays is lower than that in working days
Business Forecasting: Experiments and Case Studies
0 500 1000 1500
1000
1500
2000
2500
3000
3500
Hour
Load
Val
ue
Fig.3 Hourly load from 01, May,1990 to 05, July,1990
April 18, 2023 13
3. Data Analysis (3)
Average hourly load during 24 hours varies from hour to
hour working days
except Friday have similar shapes and similar magnitude
weekend days < working days
Business Forecasting: Experiments and Case Studies
2 4 6 8 10 12 14 16 18 20 22 241400
1600
1800
2000
2200
2400
2600
2800
Hour
Load
Val
ue
SundayMonday
Tuesday
Wednesday
Thursday
FridaySaturday
Fig.4 Hourly load during a day
April 18, 2023 14
3. Data Analysis (4)
Temperature v.s. Load Demand nonlinear relationship
Business Forecasting: Experiments and Case Studies
10 20 30 40 50 60 70 80 90 1001000
1500
2000
2500
3000
3500
4000
4500
5000
Temperature
Load
Fig.5 Correlation between the load and temperature.
April 18, 2023 15
3. Data Analysis (4)
Temperature v.s. Load Demand Only for training and testing period
Business Forecasting: Experiments and Case Studies
Fig.5 Correlation between the load and temperature.
Correlations
Load Temperature
Load Pearson Correlation 1 -.574**
Sig. (2-tailed) .000
N 2904 2904
Temperature Pearson Correlation -.574** 1
Sig. (2-tailed) .000
N 2904 2904
**. Correlation is significant at the 0.01 level (2-tailed).
April 18, 2023 16
3. Data Analysis (5)
Input features for SVR/ANN hourly load values of the previous 12 hours, and similar
hours in the previous one week Temperature variables for time point that the load was
included, plus the forecasted temperature for the forecasting hour.
daily and hourly calendar indicators
Business Forecasting: Experiments and Case Studies
( 1), ( 2),..., ( 12), ( 24), ( 48),..., ( 168),
( ) ( ), ( 1), ( 2),..., ( 12), ( 24), ( 48),..., ( 168),
( ), ( )
L t L t L t L t L t L t
Input t T t T t T t T t T t T t T t
DI t HI t
April 18, 2023 17
4. Experiments
Forecasting Methods ( by Matlab/R) Support Vector Regression Artificial Neural Network ARIMA ES MA
Input features: all the above features
Parameter optimization: Grid search, PSO
Business Forecasting: Experiments and Case Studies
April 18, 2023 18
4. Experiments
Evaluation measures
Business Forecasting: Experiments and Case Studies
Metrics Formula
1
11
2
2
1 1
ˆ1100
ˆ11
1
1001
ˆ1, 0
0,
Nt i t i
i t i
Nt i t it
ij j
j
N
ii
t i t i t i t ii
y yMAPE MAPE
N y
y yMASE MASE
Ny y
t
dDS DS
N
if y y y yd
otherwise
April 18, 2023 19
4. Experiments
Results
Business Forecasting: Experiments and Case Studies
MAPE(%) MASE DS(%)
SVR_GS 6.95 0.77 89.23
SVR_PSO 7.01 0.79 90.19
NN 8.55 0.86 85.15
ARIMA 9.24 0.95 76.91
ES 10.11 1.792 61.24
MA 13.62 2.42 45.09
April 18, 2023 20
4. Experiments
Results
Business Forecasting: Experiments and Case Studies
0 100 200 300 400 500 600 700 800-500
0
500
1000
1500
2000
2500
3000
3500
4000
4500
Hour
Load
Actual
Forecast
Error
0 100 200 300 400 500 600 700 8001500
2000
2500
3000
3500
4000
4500
Time
Dem
and
MAForecast set original data
MAForecast set forecast
0 100 200 300 400 500 600 700 8001500
2000
2500
3000
3500
4000
4500
Time
Dem
and
ESForecast set original data
ESForecast set forecast
April 18, 2023 21
Summary
Electricity load forecasting is an important issue to operate the power system reliably and economically. In this case study, support vector regression (SVR) is applied for short-term load forecasting. Characteristics of the hourly loads are firstly analyzed to select the input features. Then forecasting results of SVR with two parameter optimization methods are compared with several benchmark forecasting models.
Further topics: features selection method, separated modeling for each day and special days.
Business Forecasting: Experiments and Case Studies
Top Related