Assessing climate sensitivity of hourly electricity demand in Japan - - PowerPoint PPT Presentation

▶
assessing climate sensitivity of hourly electricity
SMART_READER_LITE
LIVE PREVIEW

Assessing climate sensitivity of hourly electricity demand in Japan - - PowerPoint PPT Presentation

Apr 09, 2024 25 likes •230 views

The 25th International Workshop Assessing climate sensitivity of hourly electricity demand in Japan Yuki HIRUTA, Lu GAO and Shuichi ASHINA November 18-19, 2019 National Institute for Environmental Studies Center for Social and Environmental

slide-1
SLIDE 1

Assessing climate sensitivity of hourly electricity demand in Japan

Yuki HIRUTA, Lu GAO and Shuichi ASHINA

International Workshop The 25th November 18-19, 2019

Center for Social and Environmental Systems Research

Yuki Hiruta

Research Associate

National Institute for Environmental Studies

slide-2
SLIDE 2

■Motivations

It is important to clarify the relationship between weather conditions and the hourly electricity demand

  • Warming urban environment in summer
  • Highly fluctuating electricity demand
  • Multiple problems such as heat strokes

International Workshop

Global warming Urbanization

  • Decrease the efficiency of electricity supply system
  • Increase the fossil fuel consumption
slide-3
SLIDE 3

■Overview

Target Areas

  • Jurisdiction of 10 electric power companies in Japan

(EPC)

Method over view

Build regression models for each EPC Multiple variables

Hourly electricity consumption

Predictor Explained variable

Understanding the relationship between weather conditions and the hourly electricity demand Represent the Temperature response functions Based on simulation by the constructed models

Simulation

  • weather conditions
  • daily cycle of human activities

Related to

slide-4
SLIDE 4

■Temperature response function(TRFs)

Electricity consumption Temperature() Comfort Zone Base Demand

U-shaped

Electricity consumption Temperature() BPT(Balanced point temperature)

V-shaped

Base Demand

4

*Also called Energy signature

  • BPT → Reference temperature for HDD/CDD
  • BPT, Slope → Electricity demand projection

Application:

Important! TRFs → Basic Units

Balance point temperature

Affects assumptions of

  • ther models!

BPT:

slide-5
SLIDE 5

■Data and Variables

Names （Abbreviations） Units Data Description Data Sauce Explained Variable Historical electricity demand（EC） MWh

Hourly electricity demand in each EPC jurisdictional area. Organization for Cross-regional Coordination of Transmission Operators Japan,

Predictors Historical weather Temperature（TEMP） ℃

Hourly averaged air temperature in FY2016 and FY2017 Japan Meteorological Agency

Humidity（HUM） %

Hourly averaged relative humidity in FY2016 and FY2017

Solar radiation（SUN） MJ/m

Hourly averaged total radiation in FY2016 and FY2017

Wind speed（WIND） m/s

Average wind speed in ten minutes before each hour in FY2016 and FY2017

Rainfall amount（RAIN） mm

Hourly total rainfall amount in FY2016 and FY2017

Snow depth （SNOW） cm

Hourly total snow depth in FY2016 and FY2017

Thermal index Discomfort Index（DI） ­

Discomfort Index.

DI 0.81 TEMP 0.01 HUM 0.99 TEMP 14.3 46.3

Derived from Historical Weather Data Above

Wind Chill（WCI） ­

Wind Chill Index.

WCI 33 TEMP 10.45 10 WIND. WIND

Human activity Holidays and weekends dummy （HDD） ­

Weekends, holidays, the New Year, and the Obon Festival were set as 1, and all other days were set as 0 in FY2016 and FY2017 Calendars in FY2016 and FY2017

Working people （WORK%） %

The percentage of people who are working at the hour. NHK Broadcasting Culture Research Institute.

Awake people （WAKE%） %

The percentage of people awake in their homes at the hour.

Sleeping people （SLEEP%） %

The percentage of people who are sleeping at the hour.

5

slide-6
SLIDE 6

■Human activity data

Hour Human activity in each hour

100 80 60 40 20 90 70 50 30 10

1 24 12 18 6

%

6

(Data source: NHK Broadcasting Culture Research Institute, 2015)

slide-7
SLIDE 7

■Model construction MARS：multivariate adaptive regression splines (Friedman, 1991)

  • Captures the complexity of the potential model by applying a locally

linear models.

  • Selects important variables during the model building process.
  • Showed excellent prediction performance in short-term power

consumption modeling

7

  • Can model nonlinearities in

high-dimensional data: flexible

  • Highly generalizable

Intuitive understanding for MARS performance

Flexible and Highly generalizable modeling is expected

(Sigauke & Chikobvu, 2010; Al-Musaylh, Deo, Adamowski, & Li, 2018)

slide-8
SLIDE 8

■Constructed models

Hokkaido Tohoku Tokyo Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa R

0.922 0.909 0.935 0.902 0.890 0.933 0.908 0.903 0.873 0.953

generalized R

0.921 0.907 0.934 0.900 0.887 0.932 0.907 0.901 0.870 0.953

Number of terms

35 39 29 37 39 25 29 32 35 30

Number of predictors adopted

9 9 8 10 9 7 7 7 8 9

Number of input predictors

11 11 11 11 11 11 11 11 11 11

Adopted predictors TEMP

〇 〇 〇 〇 〇 〇 〇 〇 〇 〇

HUM

  • SUN

〇 〇 〇 〇 〇 〇 〇 〇 〇 〇

WIND

  • 〇

〇

  • 〇

〇

RAIN

〇

  • 〇

〇

  • 〇

SNOW

〇 〇

  • 〇
  • DI

〇 〇 〇 〇 〇 〇 〇 〇 〇 〇

WCI

  • 〇
  • 〇
  • 〇

HDD

〇 〇 〇 〇 〇 〇 〇 〇 〇

  • WORK%

〇 〇 〇 〇 〇 〇 〇 〇 〇 〇

WAKE%

〇 〇 〇 〇 〇 〇 〇 〇 〇 〇

SLEEP%

〇 〇 〇 〇 〇 〇 〇 〇 〇 〇

Performed well in all power company models. 0.870(Kyushu ) 〜 0.953(Okinawa)

8

slide-9
SLIDE 9

■Model performance

〇 ：In-sample result ＋ ：Out-of-sample result x-axis：Actual （MWh）y-axis：Prediction（MWh）

  • ：The coefficient of determination of OLS (in-sample)
  • ：The coefficient of determination of OLS (out-of-sample)

：OLS regression line for in-sample result ：OLS regression line for out-of-sample result

  • 0.922
  • 0.909
  • 0.902
  • 0.882
  • 0.908
  • 0.917
  • 0.953
  • 0.960
  • 0.909
  • 0.888
  • 0.890
  • 0.895
  • 0.903
  • 0.896
  • 0.935
  • 0.934
  • 0.933
  • 0.922
  • 0.873
  • 0.896

5,000 4,500 4,000 3,500 3,000 2,500 5,500 5,000 4,500 4,000 3,500 3,000 2,500 5,500 14,000 12,000 10,000 8,000 6,000 14,000 12,000 10,000 8,000 6,000 50,000 40,000 30,000 20,000 50,000 40,000 30,000 20,000 25,000 20,000 15,000 10,000 25,000 20,000 15,000 10,000 25,000 20,000 15,000 10,000 25,000 20,000 15,000 10,000 10,000 9,000 8,000 7,000 6,000 5,000 11,000 10,000 9,000 8,000 7,000 6,000 5,000 11,000 5,000 4,000 3,000 2,000 5,000 4,000 3,000 2,000 14,000 12,000 10,000 8,000 6,000 16,000 14,000 12,000 10,000 8,000 6,000 16,000 1,400 1,2,00 1,0,00 800 600 1,400 1,2,00 1,0,00 800 600 5,000 4,000 3,000 2,000 5,000 4,000 3,000 2,000

9

High-quality models in terms of both fitting and generalization

Hokkaido Tohoku Tokyo Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa Observed values vs estimated values

slide-10
SLIDE 10

Simulating Temperature Response Functions

10

slide-11
SLIDE 11

■Hourly simulation

Weather predictors

TEMP

Regular sequences of the value Weather predictors

HUM, SUN, WIND, RAIN, SNOW

Average values of each time period at each location in weekdays. The thermal indicators

DI, WCI

Calculated from the weather predictors . Holiday dummy

HDD

Weekdays: 0 Human activity predictors

WORK%, WAKE%, SLEEP%

Values of each time period in weekdays. Hokkaido Tohoku Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa x-axis：Temperature（） y-axis：Power consumption（MWh）

Hour

〇 Observation Hourly simulation

10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 5,000 4,500 4,000 3,500 3,000 2,500 14,000 12,000 10,000 8,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 10,000 9,000 8,000 7,000 6,000 5,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 14,000 12,000 10,000 8,000 6,000 1,400 1,2,00 1,0,00 800 600 50,000 45,000 40,000 35,000 30,000 25,000 20,000 55,000

Settings for the simulation 11 Tokyo

slide-12
SLIDE 12

x-axis：Temperature（） y-axis：Power consumption（MWh） 〇 Observation Day-time simulation (from 10:00 to 18:00) Night-time simulation (from 1:00 to 5:00) Hokkaido Tohoku Tokyo Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa

10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 5,000 4,500 4,000 3,500 3,000 2,500 14,000 12,000 10,000 8,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 10,000 9,000 8,000 7,000 6,000 5,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 14,000 12,000 10,000 8,000 6,000 1,400 1,2,00 1,0,00 800 600 50,000 45,000 40,000 35,000 30,000 25,000 20,000 55,000

■Simulation in day-time and night-time

Weather predictors

TEMP

Regular sequences of the value Weather predictors

HUM, SUN, WIND, RAIN, SNOW

Average values in day‐time and night‐time at each location in weekdays. The thermal indicators

DI, WCI

Calculated from the weather predictors . Holiday dummy

HDD

Weekdays: 0 Human activity predictors

WORK%, WAKE%, SLEEP%

Average values during day‐time and night‐time in weekdays. Settings for the simulation 12

slide-13
SLIDE 13

■Approximation

Hokkaido Tohoku Tokyo Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa

10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30

x-axis：Temperature（） y-axis：Power consumption（MWh） Approximate function breakpoints

5,000 4,500 4,000 3,500 3,000 2,500 14,000 12,000 10,000 8,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 10,000 9,000 8,000 7,000 6,000 5,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 14,000 12,000 10,000 8,000 6,000 1,400 1,2,00 1,0,00 800 600 50,000 45,000 40,000 35,000 30,000 25,000 20,000 55,000

“Temperature response functions” are approximated by piecewise linear function using MARS.

〇 Observation Day-time simulation Night-time simulation 13

slide-14
SLIDE 14

Hokkaido Tohoku Tokyo Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa

10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30

■Approximate function

x-axis：Temperature（） y-axis：Power consumption（MWh） 〇 Observation Approximate function for day-time Approximate function for night-time

5,000 4,500 4,000 3,500 3,000 2,500 14,000 12,000 10,000 8,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 10,000 9,000 8,000 7,000 6,000 5,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 14,000 12,000 10,000 8,000 6,000 1,400 1,2,00 1,0,00 800 600 50,000 45,000 40,000 35,000 30,000 25,000 20,000 55,000

Parameters(the coordinates of breakpoints,

the coefficients of each linear function)

can be obtained from the piecewise linear function

Provide the parameters to other models

14

slide-15
SLIDE 15

■The effect of humidity on power consumption

Weather predictors

TEMP,HUM

Regular sequences of the value Weather predictors

SUN, WIND, RAIN, SNOW

Average values in day‐time at each location in weekdays. The thermal indicators

DI, WCI

Calculated from the Weather predictors . Holiday dummy

HDD

Weekdays: 0 Human activity predictors

WORK%, WAKE%, SLEEP%

Values for each time period in weekdays. Hokkaido Tohoku Tokyo Chubu Hokuriku Kansai Chugoku Shikoku Kyushu Okinawa 〇 Observation

10 20 30 40 50 60 70 80 90 100

Relative humidity %

x-axis：Temperature（） y-axis：Power consumption（MWh）

10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 10 10 20 30 5,000 4,500 4,000 3,500 3,000 2,500 14,000 12,000 10,000 8,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 10,000 9,000 8,000 7,000 6,000 5,000 25,000 20,000 15,000 10,000 5,000 4,500 4,000 3,500 3,000 2,500 2,000 14,000 12,000 10,000 8,000 6,000 1,400 1,2,00 1,0,00 800 600 50,000 45,000 40,000 35,000 30,000 25,000 20,000 55,000

15 Settings for the simulation

slide-16
SLIDE 16

■The effects of humidity in detail

A simulation for Tokyo in daytime as an example

16

Temperature 30℃ Humidity 80% Temperature 30℃ Humidity 27%

Power consumption reduced by 13.8%.

27% 27% 80% 80%

An actual humidity range in Tokyo when the temperature is higher than 30℃

slide-17
SLIDE 17

■ BPT(Balance point temperature)

%

BPT decrease as humidity rises

:Under the conditions wherein the humidity is high, the power consumption for cooling begins to increase at lower temperatures.

BPT(Balance point temperature)

The bottom of the V-shaped temperature response function (Amato et al. 2005)

17

slide-18
SLIDE 18

Suggestions Summarized results

■Summary

we proposed a series of methods to understand the relationship between weather conditions and the hourly electricity demand

  • The constructed models are of high-quality in terms of

1)fitting, 2)generalization capability, and 3) simulating accurate temperature response functions

  • Two different Temperature Response Functions were identified in a day;

for day-time (from 10:00 to 18:00) and for night-time (from 1:00 to 5:00).

  • Humidity affects electricity consumption significantly when temperature

is high

  • Under the condition wherein the humidity is high, the power consumption

for cooling begins to increase at lower temperatures.

  • The proposed method is recommended for identifying temperature

response functions; especially the consideration of multiple factors are important.

  • The effect of humidity should not be ignored.
slide-19
SLIDE 19

■References

19

  • Japan Ministry of Economy, Trade and Industry: Annual Report on

Energy for Fiscal Year 2018 (Energy White Paper 2019), Gyosei, 2019. (In Japanese)

  • Federation of Electric Power Companies: Electric power load leveling –

the power situation, (In Japanese) http://www.fepc.or.jp/enterprise/jigyou/juyou/index.html.

  • Heat Island Countermeasure Promotion Conference: Heat Island

Countermeasure Outlines, 2013. (In Japanese)

  • Wang, Yaoping, Bielicki, Jeffrey M: Acclimation and the response of

hourly electricity loads to meteorological variables, Energy, vol. 142, pp. 473–485, 2018.

  • Hirano, Yujiro, Yoshida, Yukiko, Narumi, Daisuke: Examination of the

effects of climatic conditions upon power consumption in the commercial sector by region and year, Proceedings of the Air‐ Conditioning and Sanitation Engineering Conference, vol. 2016.9, pp. 9–12, 2016. (In Japanese)

  • Hirano, Yujiro, Gomi, Kei, Nakamura, Shogo, Yoshida, Yukiko, Narumi,

Daisuke, Fujita, Tsuyoshi: Analysis of the impact of regional temperature pattern on the energy consumption in the commercial sector in Japan, Energy and Buildings, vol. 149, pp. 160–170, 2017.

  • Yokoyama, Makoto, Tanaka, Takahiro: Evaluation of future CO2

emissions using power consumption temperature sensitivity, and land‐ use scenario proposal and evaluation using sea breeze, Proceedings of the Air‐Conditioning and Sanitation Engineering Conference, vol. 2016.10, pp. 45–48, 2016. (In Japanese)

  • Narumi, Daisuke, Kishimoto, Takuya, Sakaguchi, Katsutoshi, Shimoda,

Yoshiyuki, Mizuno, Minoru: Temperature Sensitivity related to power consumption targeting Osaka Prefecture D‐35, Proceedings of the Air‐ Conditioning and Sanitation Engineering Conference, vol. 2004.2, pp. 1071–1074, 2004. (In Japanese)

  • Kittaka, Kosuke, Miyazaki, Hiroshi: Temperature sensitivity of power

consumption over a wide area, The Architectural Institute of Japan’s Journal of Environmental Engineering, vol. 79, no. 704, pp. 891–899,

  • 2014. (In Japanese)
  • Ihara, T., Genchi, Y., Sato, T., Yamaguchi, K. & Endo, Y: City‐block‐scale

sensitivity of electricity consumption to air temperature and air humidity in business districts of Tokyo, Japan, Energy, vol. 33, pp. 1634–1645, 2008.

  • Amato, Anthony D., Ruth, Matthias, Kirshen, Paul, Horwitz, James:

Regional Energy Demand Responses To Climate Change: Methodology And Application To The Commonwealth Of Massachusetts， Climatic Change, vol. 71, no. 1, pp. 175–201, 2005.

  • Mukherjee, Sayanti, Nateghi, Roshanak: Climate sensitivity of end‐use

electricity consumption in the built environment: An application to the state of Florida, United States, Energy, vol. 128, pp. 688–700, 2017.

  • Mukherjee, Sayanti, Vineeth, C. R., Nateghi, Roshanak: Evaluating regional

climate‐electricity demand nexus: A composite Bayesian predictive framework, Applied Energy, vol. 235, pp. 561–1582, 2019.

  • Ohashi, Yukitaka: Thermal Index, Tenki, vol. 57，no. 1，pp. 57–59，2010.

(In Japanese)

  • Friedman, Jerome H.: Multivariate Adaptive Regression Splines, The

Annals of Statistics, vol. 19, no. 1, pp. 1–67, 1991.

  • Organization for Cross‐regional Coordination of Transmission Operators:

Wide Area Engine System, (In Japanese) http://occtonet.occto.or.jp/public/dfw/RP11/OCCTO/SD/LOGIN_login#.

  • Japan Meteorological Agency: Past weather data, (In Japanese)

http://www.data.jma.go.jp/gmd/risk/obsdl/index.php.

  • NHK Broadcasting Culture Research Institute: Databook・Citizen Activity

Hourly Survey, NHK Publishing, 2015. (In Japanese)

  • James, Gareth, Witten, Daniela, Hastie, Trevor, Tibshirani, R. J.: An

introduction to statistical learning : with applications in R (Springer texts in statistics), Springer, 2013.

  • Shimokawa, Toshio, Sugimoto, Tomoyuki, Goto, Musashi: Tree structure

approach method, Tokyo, Kyoritsu Publishing, 2013. (In Japanese)

  • Sigauke, C., Chikobvu, D.: Daily peak electricity load forecasting in South

Africa using a multivariate non‐parametric regression approach, ORiON,

  • Vol. 26, No. 2 (2010), 2010.
  • Al‐Musaylh, Mohanad S., Deo, Ravinesh C., Adamowski, Jan F., Li, Yan.:

Short‐term electricity demand forecasting with MARS, SVR and ARIMA models using aggregated demand data in Queensland, Australia, Advanced Engineering Informatics, vol. 35, pp. 1–16, 2018.

  • Japan Ministry of the Environment: Actual analyses and countermeasures

for the heat island phenomenon (Heat Island Condition Analysis, Survey, and Investigation Committee Report), 2001. (In Japanese)

slide-20
SLIDE 20

Thank you

I always welcome your critical comments, suggestions, and corrections.