Mobile QR Code QR CODE : Journal of the Korean Institute of Illuminating and Electrical Installation Engineers

Journal of the Korean Institute of Illuminating and Electrical Installation Engineers

ISO Journal TitleJ Korean Inst. IIIum. Electr. Install. Eng.

  1. (Sangmyung University, Department of Electrical Engineering, Master's Course)
  2. (Sangmyung University, Department of Electrical Engineering, Associate Professor)



Photovoltaic Power Forecasting, Support Vector Regression(SVR), Day-ahead Forecasting, Short-Term Forecasting, Machine Learning

1. ์„œ๋ก 

์ตœ๊ทผ ์ „ ์„ธ๊ณ„์ ์œผ๋กœ ์‹ ๊ธฐํ›„์ฒด์ œ ์ถœ๋ฒ”์— ๋”ฐ๋ผ ์ „์› ๊ตฌ์„ฑ์€ ๋ณ€ํ™”ํ•˜๊ณ  ์žˆ๋‹ค. 2017๋…„ ์‹ ๊ทœ ์žฌ์ƒ์—๋„ˆ์ง€ ๋ฐœ์ „์šฉ๋Ÿ‰์€ 178GW์ด๋ฉฐ, ๊ทธ์ค‘ ํƒœ์–‘๊ด‘๋ฐœ์ „์€ 97GW๋กœ ์‹ ๊ทœ ๋ฐœ์ „์„ค๋น„ ์ค‘ 1์œ„๋ฅผ ์ฐจ์ง€ํ•˜๋ฉฐ ์žฌ์ƒ์—๋„ˆ์ง€ ๋ณด๊ธ‰์„ ์ฃผ๋„ํ•˜๊ณ  ์žˆ๋‹ค[1]. ํ๋ฆ„์— ๋”ฐ๋ผ ๊ตญ๋‚ด์—์„œ๋„ ์žฌ์ƒ์—๋„ˆ์ง€ 3020 ์ •์ฑ…์„ ํ†ตํ•ด ๋Œ€๊ทœ๋ชจ ๋ณ€๋™์„ฑ ์ „์›์˜ ๊ณ„ํ†ต ์œ ์ž…์ด ๊ณ„ํš๋˜์–ด์žˆ๋‹ค. ์ œ8์ฐจ ์ „๋ ฅ์ˆ˜๊ธ‰๊ธฐ๋ณธ๊ณ„ํš์— ๋”ฐ๋ฅด๋ฉด, ๋ฐœ์ „๋Ÿ‰ ๋น„์ค‘ 20%๋ฅผ ์žฌ์ƒ์—๋„ˆ์ง€๋กœ ๊ณต๊ธ‰ํ•˜๊ธฐ ์œ„ํ•ด 58.5GW์˜ ๋Œ€๊ทœ๋ชจ ์žฌ์ƒ์—๋„ˆ์ง€ ์„ค๋น„๋ฅผ ์ˆ˜์šฉํ•  ๊ฒƒ์œผ๋กœ ๊ณ„ํš๋˜์–ด์žˆ๋‹ค. ์ด ์ค‘ ํƒœ์–‘๊ด‘๋ฐœ์ „์€ ์žฌ์ƒ์—๋„ˆ์ง€ ๊ณต๊ธ‰๋Ÿ‰์˜ 62%์— ํ•ด๋‹นํ•˜๋Š” 36.5GW๊ฐ€ ํ•ด๋‹นํ•œ๋‹ค.

ํƒœ์–‘๊ด‘๋ฐœ์ „์€ ๊ธฐ์กด์˜ ์ผ๋ฐ˜์ ์ธ ๋ฐœ์ „์›๊ณผ๋Š” ๋‹ค๋ฅด๊ฒŒ ์ผ์‚ฌ๋Ÿ‰, ์˜จ๋„, ์šด๋Ÿ‰ ๋“ฑ ๋‹ค์–‘ํ•œ ๊ธฐ์ƒํ•™์ ์ธ ์›์ธ์œผ๋กœ ์ธํ•˜์—ฌ ์ถœ๋ ฅ์ด ๋ณ€๋™ํ•˜๋Š” ํŠน์„ฑ์ด ์žˆ์œผ๋ฏ€๋กœ ๋Œ€๊ทœ๋ชจ ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€๊ฐ€ ๊ณ„ํ†ต์— ์—ฐ๊ณ„๋  ๊ฒฝ์šฐ ํƒœ์–‘๊ด‘๋ฐœ์ „์˜ ๊ฐ€๋ณ€์„ฑ ๋ฐ ๋ถˆํ™•์‹ค์„ฑ์œผ๋กœ ๊ณ„ํ†ต์˜ ๋ถˆ์•ˆ์ •์„ ์œ ๋ฐœํ•  ์ˆ˜ ์žˆ๋‹ค. ๋ณ€๋™์„ฑ ์ „์›์˜ ๊ณ„ํ†ต ์—ฐ๊ณ„์ฆ๊ฐ€์— ๋Œ€์‘ํ•˜์—ฌ ๊ณ„ํ†ต์˜ ์œ ์—ฐ์„ฑ ํ–ฅ์ƒ์„ ์œ„ํ•œ ์กฐ์น˜๊ฐ€ ํ•„์š”ํ•˜๋ฉฐ, ์—ฌ๋Ÿฌ ๋ฐฉ์•ˆ ์ค‘ ์žฌ์ƒ์—๋„ˆ์ง€์˜ ์ •ํ™•ํ•œ ์˜ˆ์ธก์ด ๊ฐ€์žฅ ๋น„์šฉ ํšจ๊ณผ์ ์ธ ๋ฐฉ์•ˆ์œผ๋กœ ๊ณ ๋ ค๋œ๋‹ค[2].

Fig. 1. Cost Comparison for system Flexibilities

../../Resources/kiiee/JIEIE.2019.33.6.042/fig1.png

๋˜ํ•œ ์žฌ์ƒ์—๋„ˆ์ง€์˜ ์˜ˆ์ธก ์ •ํ™•๋„ ํ–ฅ์ƒ์— ๋”ฐ๋ผ ์ „๋ ฅ ๊ณ„ํ†ต์˜ ์šด์˜๋น„ ์‚ญ๊ฐ ๋“ฑ ๊ณ„ํ†ต์˜ ์‹ ๋ขฐ๋„ ์ธก๋ฉด๊ณผ ์•„์šธ๋Ÿฌ ๊ฒฝ์ œ์ ์ธ ์ธก๋ฉด์—์„œ๋„ ํฐ ์ด์ต์ด ๋ฐœ์ƒํ•˜๋Š” ๊ฒƒ์œผ๋กœ ๋ถ„์„๋˜์—ˆ๋‹ค[3]. ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์€ ์ „๋ ฅ ๊ณ„ํ†ต์˜ ๊ณต๊ธ‰ ๋ ˆ๋ฒจ ์ฆ๊ฐ€, ํšจ์œจ์ ์ธ ์ „๋ ฅ์‹œ์Šคํ…œ ์šด์˜์„ ์œ„ํ•ด ํ•„์š”ํ•˜๋‹ค.

ํƒœ์–‘๊ด‘๋ฐœ์ „์˜ˆ์ธก์„ ์œ„ํ•ด ๋ฌผ๋ฆฌ์ , ํ†ต๊ณ„์  ๊ธฐ๋ฐ˜์˜ ์‹œ๊ณ„์—ด ๋ถ„์„ ๋ฐ ํšŒ๊ท€๋ถ„์„์„ ํ†ตํ•œ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์˜ˆ์ธก์— ๊ด€ํ•œ ์—ฐ๊ตฌ๋“ค์ด ์„ ํ–‰๋˜์—ˆ๋‹ค. ์ด๋Ÿฌํ•œ ๊ธฐ๋ฒ•๋“ค์€ ๋น ๋ฅด๊ณ  ๊ฐ„๋‹จํ•˜๊ฒŒ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ์žฅ์ ์ด ์žˆ์ง€๋งŒ, ์˜ˆ์ธก ์‹œ ๋งŽ์€ ๋ณ€์ˆ˜๋ฅผ ์š”๊ตฌํ•˜๋ฉฐ ํƒœ์–‘๊ด‘๋ฐœ์ „์˜ ๋ณ€๋™ ํŠน์„ฑ ๋ฐ ์ถœ๋ ฅ ํŒจํ„ด์„ ๋ฐ˜์˜ํ•˜๋Š” ๋ฐ ํ•œ๊ณ„๊ฐ€ ์กด์žฌํ•œ๋‹ค[4].

๋”ฐ๋ผ์„œ ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ๊ธฐ๊ณ„ํ•™์Šต ๊ธฐ๋ฒ•์˜ ํ•˜๋‚˜์ธ SVR(Support Vector Regression)์„ ์ด์šฉํ•˜์—ฌ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•˜์˜€๋‹ค. SVR ๊ธฐ๋ฒ•์€ -band๋ฅผ ๋„์ž…ํ•˜์—ฌ ํšŒ๊ท€ํ•™์Šต์„ ์ˆ˜ํ–‰ํ•˜๊ธฐ ๋•Œ๋ฌธ์— ์ด์ƒ์ (Outlier)์— ๋Œ€ํ•œ ์˜ํ–ฅ์ด ์ ๊ณ , ์ปค๋„ ํ•จ์ˆ˜๋ฅผ ๋„์ž…ํ•˜์—ฌ ๋น„์„ ํ˜• ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•ด ํšŒ๊ท€๋ถ„์„์ด ๊ฐ€๋Šฅํ•˜๋‹ค. SVR์„ ์ด์šฉํ•˜์—ฌ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•  ๊ฒฝ์šฐ ๋†’์€ ์ •ํ™•๋„๋ฅผ ๊ฐ€์งˆ ์ˆ˜ ์žˆ๋‹ค. ๋˜ํ•œ ๊ธฐ์ƒ์ฒญ์—์„œ ๊ฒฉ์ž๋ฌด๋Šฌ๋กœ ์ฃผ์–ด์ง€๋Š” ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋ฅผ ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ์ผ์‚ฌ๋Ÿ‰์œผ๋กœ ๋ถ„์„ํ•˜๊ธฐ ์œ„ํ•ด ํฌ๋ฆฌ๊น…(Kriging)์„ ์ด์šฉํ•˜์—ฌ ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ์ผ์‚ฌ๋Ÿ‰์„ ์˜ˆ์ธกํ•˜์—ฌ ์ ์šฉํ•˜์˜€๋‹ค. ๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ์ „๋ผ๋„ A ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ์ถœ๋ ฅ ๋ฐ์ดํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•˜์˜€์œผ๋ฉฐ, ARIMAX (Autoregression Integrated Moving Average with Exogenous Variables), ์ง€์†์„ฑ(Persisterce) ๊ธฐ๋ฒ•๊ณผ ๋น„๊ต ๋ถ„์„ํ•˜์˜€๋‹ค. ์˜ˆ์ธก ์ •ํ™•๋„ ๊ฒ€์ฆ์„ ์œ„ํ•˜์—ฌ ์‹ค์ธก ๋ฐ์ดํ„ฐ์™€ MAPE(Mean Absolute Percentage Error), RMSE(Root Mean Square Error)๋ฅผ ๋ถ„์„ํ•˜์˜€๋‹ค.

2. SVR์˜ ์ด๋ก ์  ๋ฐฐ๊ฒฝ

SVR์€ ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ\left\{\left(x_{1}, y_{1}\right), \ldots\left(x_{\ell,} y_{\ell}\right)\right\} \subset x \times R์— ๋Œ€ํ•ด์„œ ์‹ค์ œ ๊ฐ’ $y_{i}$๋กœ๋ถ€ํ„ฐ ์ตœ๊ณ  $\epsilon $(๋ฌด๊ฐ๊ฐ ๋ชจ์ˆ˜, Insensitive Parameter)๋งŒํผ์˜ ํŽธ์ฐจ ๋‚ด์— ์žˆ์œผ๋ฉฐ ์ตœ์†Œํ™”๋œ $w$๋ฅผ ๊ฐ–๋Š” ํ•จ์ˆ˜ $f(x)$๋ฅผ ์ฐพ๋Š” ๊ฒƒ์ด๋‹ค. ์—ฌ๊ธฐ์„œ $x$๋Š” ์ž…๋ ฅ ๋ฒกํ„ฐ, $y$๋Š” ์ถœ๋ ฅ ๋ฒกํ„ฐ, $R ^{m}$์€ ์ž…๋ ฅ๊ณต๊ฐ„์„ ๋‚˜ํƒ€๋‚ธ๋‹ค. ์œ„์˜ ์กฐ๊ฑด์„ ๋งŒ์กฑํ•˜๋Š” ์„ ํ˜•ํ•จ์ˆ˜ $f(x)$๋Š” ๋‹ค์Œ ์‹ (1)๊ณผ ๊ฐ™๋‹ค. ํ•ด๋‹น ๊ธฐ๋ฒ•์€ ์—ฌ์œ  ๊ฐ„๊ฒฉ(Margin)์„ ์„ค์ •ํ•˜์—ฌ ์ตœ์ ์˜ ํšŒ๊ท€ ํ•จ์ˆ˜๋ฅผ ๊ตฌํ•˜๋Š” ๊ฒƒ์„ ๊ธฐ๋ณธ์œผ๋กœ ํ•œ๋‹ค[5-7].

(1)
$f(x)=w \cdot x+b i a s$

์ตœ์†Œํ™”๋œ $x$๋ฅผ ์ฐพ๊ธฐ ์œ„ํ•ด ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋ณผ๋ก ์ตœ์ ํ™” ๋ฌธ์ œ(Convex Optimization Problem)๋ฅผ ๊ตฌ์„ฑํ•  ์ˆ˜ ์žˆ๋‹ค.

(2)
$\min _{w} \frac{1}{2}\|w\|^{2} \quad s \cdot t\left\{\begin{array}{l}{y_{i}-f(x) \leq \epsilon} \\ {f(x)-y_{i} \leq \epsilon}\end{array}\right.$

๊ทธ๋Ÿฌ๋‚˜ ๊ทธ๋ฆผ. 2์™€ ๊ฐ™์ด $\epsilon $-tube ๋ฐ”๊นฅ์— ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ๊ฐ€ ์กด์žฌํ•œ๋‹ค๋ฉด ์‹ (2)๋Š” ์„ฑ๋ฆฝํ•˜์ง€ ์•Š๋Š”๋‹ค. $\epsilon $-tube ๋ฐ”๊นฅ์— ์กด์žฌํ•˜๋Š” ๋ฐ์ดํ„ฐ๋“ค์„ ํฌํ•จํ•˜์—ฌ ๋ณผ๋ก ์ตœ์ ํ™”์˜ ๋ฌธ์ œ๊ฐ€ ์„ฑ๋ฆฝํ•˜๋„๋ก ์—ฌ์œ  ๋ณ€์ˆ˜(Slack Variable) $\xi _{i} , \xi _{i} ^{*}$์™€ ๋น„์šฉ ํ•จ์ˆ˜(Cost) $C$๋ฅผ ๋„์ž…ํ•˜๊ณ , ์‹ (3)๊ณผ ๊ฐ™์ด ๋ณผ๋ก ์ตœ์ ํ™” ๋ฌธ์ œ๋ฅผ ์ƒˆ๋กœ ๊ตฌ์„ฑํ•œ๋‹ค.

Fig. 2. The graph of $\epsilon $-insensitive loss

../../Resources/kiiee/JIEIE.2019.33.6.042/fig1.png

(3)
$\min _{w} \frac{1}{2}\|w\|^{2}+C \sum_{i=1}^{N}\left(\xi_{i}+\xi_{i}^{*}\right)$ s.t$\left\{\begin{array}{l}{y_{i}-f(x) \leq \epsilon+\xi_{i}} \\ {f(x)-y_{i} \leq \epsilon+\xi_{i}^{*}}\end{array}, \xi_{i}, \xi_{i}^{*} \geq 0\right.$

์‹ (3)์˜ ์ตœ์ ํ™” ๋ฌธ์ œ๋Š” ๋ผ๊ทธ๋ž‘์ฃผ ์Šน์ˆ˜๋ฒ•(Lagrange Multiplier Method)์„ ๋„์ž…ํ•˜์—ฌ ํ•ด๋ฅผ ๊ตฌํ•œ๋‹ค. ์—ฌ๊ธฐ์„œ $\alpha _{i} , \alpha _{i} ^{*} , \eta _{i} , \eta _{i} ^{*}$๋Š” ๋ผ๊ทธ๋ž‘์ฃผ ์Šน์ˆ˜์ด๋ฉฐ, < , > ๋Š” ๋‚ด์ ์„ ์˜๋ฏธํ•œ๋‹ค.

(4)
$\begin{aligned} L := & \frac{1}{2}\|w\|^{2}+C \sum_{i=1}^{\ell}\left(\xi_{i}+\xi_{i}^{*}\right) \\ & -\sum_{i=1}^{l}\left(\eta_{i} \xi_{i}+\eta_{i}^{*} \xi_{i}^{*}\right) \\ & -\sum_{i=1}^{\ell} \alpha_{i}\left(\epsilon+\xi_{i}-y_{i}+ < w, x_{i} > +b\right) \\ & -\sum_{i=1}^{\ell} \alpha_{i}^{*}\left(\varepsilon+\xi_{i}^{*}+ < w, x_{i} > -b\right) \end{aligned}$

(5)
$w=\sum_{i}^{\ell}\left(\alpha_{i}-\alpha_{i}^{*}\right) x_{i},$ $f(x)=\sum_{i=1}^{\ell}\left(\alpha_{i}-\alpha_{i}^{*}\right) < x_{i}, x > +$bias

์‹ (4)์˜ ์„ ํ˜• SVR ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ์ปค๋„(Kernel) ํ•จ์ˆ˜๋ฅผ ์ด์šฉํ•˜์—ฌ ๊ทธ๋ฆผ. 3๊ณผ ๊ฐ™์ด ๋น„์„ ํ˜•์œผ๋กœ ํ™•์žฅ ์‹œํ‚ฌ ์ˆ˜ ์žˆ๋‹ค. ์ž…๋ ฅ๊ณต๊ฐ„์— ์žˆ๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ๋น„์„ ํ˜• ์‚ฌ์ƒํ•จ์ˆ˜๋ฅผ ์ด์šฉํ•˜์—ฌ ์ฐจ์› ๋†’์€ ๊ณต๊ฐ„์œผ๋กœ ์‚ฌ์ƒ์‹œํ‚จ ํ›„ ๋น„์„ ํ˜• ์‚ฌ์ƒ์„ ํ•˜๋ฉด ๋น„์„ ํ˜• ํ•จ์ˆ˜๋ฅผ ๊ทผ์‚ฌํ™”ํ•  ์ˆ˜ ์žˆ๋‹ค. ์—ฌ๊ธฐ์„œ ๋น„์„ ํ˜• ์‚ฌ์ƒํ•จ์ˆ˜๋ฅผ ์ปค๋„ ํ•จ์ˆ˜๋ผ ํ•˜๋ฉฐ, $k\left(x, x^{\prime}\right) := < \Phi(x), \Phi\left(x^{\prime}\right) > $์ด๋‹ค. ์ปค๋„ ํ•จ์ˆ˜์˜ ๋„์ž…์„ ํ†ตํ•ด ์‹ (5)์˜ ํšŒ๊ท€ ํ•จ์ˆ˜๋Š” ์‹ (6)๊ณผ ๊ฐ™์ด ๋‚˜ํƒ€๋‚œ๋‹ค.

Fig. 3. Nonlinear Mapping of Linear Functions

../../Resources/kiiee/JIEIE.2019.33.6.042/fig3.png

(6)
$w=\sum_{i}^{\ell}\left(\alpha_{i}-\alpha_{i}^{*}\right) \Phi\left(x_{i}\right)$ $f(x)=\sum_{i=1}^{\ell}\left(\alpha_{i}-\alpha_{i}^{*}\right) k\left(x_{i}, x\right)+$bias

๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ๊ฐ€์šฐ์‹œ์•ˆ ์ปค๋„์„ ์ด์šฉํ•˜์˜€์œผ๋ฉฐ, ๊ฐ€์šฐ์‹œ์•ˆ ์ปค๋„์— ๋Œ€ํ•œ ์‹ ๋‹ค์Œ ์‹ (7)๊ณผ ๊ฐ™๋‹ค. $x, x ^{'}$๋Š” ๋ฐ์ดํ„ฐ ํฌ์ธํ„ฐ๋ฅผ ์˜๋ฏธํ•˜๋ฉฐ, $\gamma $๋Š” ๊ฐ€์šฐ์‹œ์•ˆ ์ปค๋„์˜ ํญ์„ ์ œ์–ดํ•˜๋Š” ๋งค๊ฐœ๋ณ€์ˆ˜์ด๋‹ค.

(7)
$k_{r b f}\left(x, x^{\prime}\right)=\exp \left(-\gamma\left\|x_{1}-x_{2}\right\|^{2}\right)$

3. ํƒœ์–‘๊ด‘๋ฐœ์ „ ์˜ˆ์ธก์„ ์œ„ํ•œ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค ๊ตฌ์ถ•

๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ํƒœ์–‘๊ด‘์ถœ๋ ฅ ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•˜๊ธฐ ์œ„ํ•˜์—ฌ, 2016๋…„ ์ „๋ผ๋„ A ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€๋กœ๋ถ€ํ„ฐ ์ทจ๋“ํ•œ 1์‹œ๊ฐ„ ๋‹จ์œ„ ํƒœ์–‘๊ด‘๋ฐœ์ „์ถœ๋ ฅ๋ฐ์ดํ„ฐ์™€ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋กœ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค ๋ชจ๋ธ๋ง์„ ์ˆ˜ํ–‰ํ•˜์˜€๋‹ค. ํ•ด๋‹น ๋ฐ์ดํ„ฐ๋Š” ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์œ„ํ•œ ๋ชจ๋ธ์˜ ํ›ˆ๋ จ ๋ฐ์ดํ„ฐ๋กœ ์ด์šฉํ•˜์˜€์œผ๋ฉฐ, ํฌ๋ฆฌ๊น… ๊ธฐ๋ฒ•์„ ํ†ตํ•œ ์ผ์‚ฌ๋Ÿ‰ ์˜ˆ์ธก ๊ฐ’์€ ์ž…๋ ฅ๋ฐ์ดํ„ฐ๋กœ ์‚ฌ์šฉํ•˜์˜€๋‹ค.

3.1 ํฌ๋ฆฌ๊น…๊ธฐ๋ฒ•์„ ์ด์šฉํ•œ ์ผ์‚ฌ๋Ÿ‰ ์˜ˆ์ธก

ํฌ๋ฆฌ๊น…์€ ๊ณต๊ฐ„๋ชจ๋ธ๋ง ๊ธฐ๋ฒ•์˜ ํ•˜๋‚˜๋กœ, ์ด๋ฏธ ์•Œ๊ณ  ์žˆ๋Š” ์ฃผ๋ณ€์˜ ๊ฐ’๋“ค์„ ์„ ํ˜•์œผ๋กœ ์กฐํ•ฉํ•จ์œผ๋กœ ๊ด€์‹ฌ ์ง€์ ์˜ ํŠน์„ฑ๊ฐ’ ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•œ๋‹ค[8]. ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ํฌ๋ฆฌ๊น…๊ธฐ๋ฒ• ์ค‘ ์ •๊ทœ ํฌ๋ฆฌ๊น…์„ ์ด์šฉํ•˜์—ฌ ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ์ผ์‚ฌ๋Ÿ‰ ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•˜์˜€๋‹ค. ์ •๊ทœ ํฌ๋ฆฌ๊น…์€ ์‹ (8)๊ณผ ๊ฐ™์ด i๊ฐœ์˜ ์ธ๊ทผ ๋ฐ์ดํ„ฐ $\alpha $์— ๋Œ€ํ•ด ๊ฐ๊ฐ ๊ฐ€์ค‘์น˜ $\lambda $๋ฅผ ๋ถ€์—ฌํ•˜์—ฌ, ์ธ๊ทผ ๋ฐ์ดํ„ฐ์˜ ํ•ฉ์„ ํ†ตํ•ด ์˜ˆ์ธก ์ง€์ ์˜ ๊ฐ’ $\alpha *$์„ ๋„์ถœ ํ•œ๋‹ค[7]. ๊ฐ ์ง€์ ์— ๋Œ€ํ•œ ์˜ํ–ฅ์ด ํŽธํ–ฅ๋˜๋Š” ๊ฒƒ์„ ๋ฐฉ์ง€ํ•˜๊ธฐ ์œ„ํ•ด, ์ •๊ทœ ํฌ๋ฆฌ๊น…์˜ ๊ฐ€์ค‘์น˜์˜ ํ•ฉ์€ 1์ด ๋˜๋„๋ก ์œ ์ง€ํ•œ๋‹ค. ์ง€์ ๋ณ„ ๊ฐ€์ค‘์น˜๋Š” ์ธ๊ทผ ์ง€์  ๊ฐ„ ๊ณต๋ถ„์‚ฐ ๊ฐ’์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ณ„์‚ฐ ๋œ๋‹ค[9].

(8)
$\alpha^{*}=\sum_{i=1}^{n} \lambda_{i} \alpha_{i} \quad\left(s . t \sum_{i=1}^{n} \lambda_{i}=1\right)$

ํ‘œ 2์˜ 30๊ฐœ ๊ธฐ์ƒ ํƒ‘์˜ ์œ„์น˜ ๋ฐ์ดํ„ฐ์™€ ๊ธฐ์ƒ์ฒญ์˜ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ์ •๊ทœ ํฌ๋ฆฌ๊น…์„ ์ ์šฉํ•˜์˜€๊ณ , ํฌ๋ฆฌ๊น… ๊ธฐ๋ฒ•์„ ํ†ตํ•ด ์˜ˆ์ธก ์ง€์ ์— ๋Œ€ํ•œ ์ผ์‚ฌ๋Ÿ‰์„ ์˜ˆ์ธกํ•˜์˜€๋‹ค. ์˜ˆ์ธก ์ง€์ ์— ๋Œ€ํ•œ ์ •๋ณด๋Š” ํ‘œ 1๊ณผ ๊ฐ™์œผ๋ฉฐ, 2016๋…„ 5์›” 16์ผ์— ๋Œ€ํ•œ ์ผ์‚ฌ๋Ÿ‰ ์˜ˆ์ธก๊ฐ’์€ ํ‘œ 3๊ณผ ๊ฐ™๋‹ค.

Table 1. About the Location of the Solar Farm A

์˜ˆ์ธก์ง€์ 

์œ„๋„

๊ฒฝ๋„

์„ค๋น„์šฉ๋Ÿ‰

A ํƒœ์–‘๊ด‘ ๋ฐœ์ „๋‹จ์ง€

35.2950

126.3855

11MW

Table 2. Location Data for Met Tower

MET

Longitude

Latitude

MET 1

126.9658

37.5714

$\vdots$

$\vdots$

$\vdots$

MET 10

127.4407

36.6392

MET 11

127.3721

36.3720

MET 12

126.3812

34.8169

MET 13

126.4778

35.2837

MET 14

127.1190

35.8408

MET 15

127.1286

35.3714

MET 16

126.8916

35.1729

MET 17

127.6914

34.9434

MET 18

126.7689

34.6261

MET 19

127.2123

34.7633

MET 20

126.6970

35.4266

$\vdots$

$\vdots$

$\vdots$

MET 30

128.8930

35.2267

Table 3. Estimation of Solar Radiation Using Krigging Method

์‹œ๊ฐ„(Hour)

์ผ์‚ฌ๋Ÿ‰( $MJ/M ^{2}$ )

6

0.012

7

0.127

8

0.466

9

1.295

10

2.236

11

2.901

12

3.044

13

3.575

14

3.462

15

3.030

16

2.328

17

2.023

18

1.190

19

0.426

20

0.027

3.2 ARIMAX, ์ง€์†์„ฑ ๋ชจ๋ธ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก

ARIMAX๋Š” ๋‹จ๊ธฐ ์‹œ๊ณ„์—ด ์˜ˆ์ธก ๋ชจ๋ธ๋กœ, ARIMA (Autoregression Integrated Moving Average) ๋ชจ๋ธ์— ์™ธ๋ถ€๋ณ€์ˆ˜๋ฅผ ๊ฒฐํ•ฉํ•œ ๋ชจ๋ธ์ด๋‹ค[10]. ์™ธ๋ถ€๋ณ€์ˆ˜๋กœ ํฌ๋ฆฌ๊น…์„ ํ†ตํ•œ ์˜ˆ์ธก ์ผ์‚ฌ๋Ÿ‰์„ ์ด์šฉํ•˜์—ฌ ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ–ˆ๋‹ค. ์ฃผ๋ณ€์ˆ˜๋กœ๋Š” A ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ์ถœ๋ ฅ ๋ฐ์ดํ„ฐ๋ฅผ ์‚ฌ์šฉํ•˜๊ณ  ์™ธ๋ถ€๋ณ€์ˆ˜๋กœ ํฌ๋ฆฌ๊น…์œผ๋กœ ์˜ˆ์ธกํ•œ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ–ˆ๋‹ค. ์ง€์†์„ฑ ๋ชจ๋ธ์€ ์˜ˆ์ธก ์‹œ์ ๊ณผ ๊ณผ๊ฑฐ ์‹œ์ ์˜ ์ถœ๋ ฅ์„ ๋™์ผํ•˜๋‹ค ๊ฐ€์ •ํ•˜๋Š” ๊ฐ€์žฅ ๊ธฐ๋ณธ์ ์ธ ์˜ˆ์ธก ๋ฐฉ๋ฒ•์ด๋‹ค[11]. ์ง€์†์„ฑ ๋ชจ๋ธ์€ ์‹ (9)๋กœ ํ‘œํ˜„๋œ๋‹ค. $p _{t+k}$๋Š” ์˜ˆ์ธก ์‹œ์ ์˜ ์ถœ๋ ฅ์„ ์˜๋ฏธํ•˜๋ฉฐ $p _{t}$๋Š” ๊ณผ๊ฑฐ ์‹œ์ ์˜ ์ถœ๋ ฅ์„ ์˜๋ฏธํ•œ๋‹ค. k๋Š” ๊ณผ๊ฑฐ ์‹œ์ ๊ณผ ํ˜„์žฌ ์‹œ์ ์˜ ์ฐจ์ด๋ฅผ ์˜๋ฏธํ•œ๋‹ค.

(9)
$p_{t+k}=p_{t}$

3.3 SVR ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ด์šฉํ•œ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก

๋ณธ ์—ฐ๊ตฌ์—์„œ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ ์˜ˆ์ธก์„ ์œ„ํ•œ ์ž…๋ ฅ ๋ฐ ์ถœ๋ ฅ ๋ฐ์ดํ„ฐ๋Š” 1์‹œ๊ฐ„ ๋‹จ์œ„ ๋ฐ์ดํ„ฐ๋กœ ๊ตฌ์„ฑ๋˜์–ด ์žˆ์œผ๋ฉฐ, ์ด๋Š” ํ˜„์žฌ ๊ตญ๋‚ด ๊ณ„ํ†ต์„ ์šด์˜ ๋ฐ ๊ณ„ํšํ•˜๋Š”๋ฐ 1์‹œ๊ฐ„ ๋‹จ์œ„๋ฅผ ๊ธฐ๋ณธ์œผ๋กœ ์ ์šฉํ•˜๊ณ  ์žˆ๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค.

๋ชจ๋ธ ํ•™์Šต์„ ์œ„ํ•œ ๋ฐ์ดํ„ฐ๋กœ๋Š” ์‚ฌ๊ณ„์ ˆ์„ ๊ฐ€์ง€๊ณ  ์žˆ๋Š” ์šฐ๋ฆฌ๋‚˜๋ผ์˜ ๊ธฐํ›„์  ํŠน์„ฑ์„ ๊ณ ๋ คํ•˜๊ณ  ๋น ๋ฅธ ๊ณ„์‚ฐ์†๋„๋ฅผ ํ†ตํ•ด ์‹ค์‹œ๊ฐ„ ๊ณ„ํ†ต์— ์ ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ๋ฐฉ์•ˆ์„ ๊ณ ๋ คํ•˜๊ธฐ ์œ„ํ•ด ์˜ˆ์ธก์ผ ์ „ 720์‹œ๊ฐ„(30์ผ) ์ถœ๋ ฅ ๋ฐ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ๋ชจ๋ธ ํ•™์Šต์„ ์ง„ํ–‰ํ•˜์˜€๋‹ค. ์ด๋ฅผ ์ •๋ฆฌํ•˜๋ฉด ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

โ€ค ๋ชจ๋ธ ํ•™์Šต ๊ธฐ๊ฐ„: 2016๋…„ 4์›” 1์ผ(ํ•˜๋ฃจ์”ฉ ์ด๋™) ~ ์˜ˆ์ธก ์‹œ์  ํ•˜๋ฃจ ์ „

โ€ค ๋ชจ๋ธ ํ‰๊ฐ€ ๊ธฐ๊ฐ„: 2016๋…„ 5์›”

SVR์„ ์ด์šฉํ•œ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์˜ˆ์ธก ์•Œ๊ณ ๋ฆฌ์ฆ˜์€ ๊ทธ๋ฆผ. 4์™€ ๊ฐ™๋‹ค.

Fig. 4. Photovoltaic Power Forecasting Algorithm using SVR

../../Resources/kiiee/JIEIE.2019.33.6.042/fig4.png

์ˆ˜๋ฆฝํ•œ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•ด ์ผ์‚ฌ๋Ÿ‰๊ณผ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ ์ƒ๊ด€๊ด€๊ณ„์— ๋Œ€ํ•ด ๋ถ„์„ํ•˜์˜€๋‹ค. ํ•™์Šต๋ฐ์ดํ„ฐ์— ๋Œ€ํ•œ ํ‰๊ท  ์ƒ๊ด€๊ด€๊ณ„๋Š” 0.9107๋กœ ๋†’์€ ์–‘์˜ ์ƒ๊ด€๊ด€๊ณ„๋ฅผ ๊ฐ€์ง€๋Š” ๊ฒƒ์œผ๋กœ ๋‚˜ํƒ€๋‚ฌ๋‹ค. ๋ฐ์ดํ„ฐ ์ „์ฒ˜๋ฆฌ ๊ณผ์ •์—์„œ๋Š” ๊ด‘์ „ํšจ๊ณผ๋ฅผ ์ด์šฉํ•˜์—ฌ ๋ฐœ์ „ํ•˜๋Š” ํƒœ์–‘๊ด‘๋ฐœ์ „์˜ ํŠน์„ฑ์„ ๊ณ ๋ คํ•˜์—ฌ ์ผ์‚ฌ๋Ÿ‰์ด 0์ธ ๊ฒฝ์šฐ์— ์ถœ๋ ฅ์ด ๋ฐœ์ƒํ•˜๋Š” ๊ฒฝ์šฐ ํ•ด๋‹น ๋ฐ์ดํ„ฐ๋ฅผ ์ œ์™ธํ•˜๊ณ  ๋ชจ๋ธ ํ•™์Šต์„ ์ˆ˜ํ–‰ํ–ˆ๋‹ค. ๋ณธ SVR ์˜ˆ์ธก ๋ชจ๋ธ์—์„œ ์‚ฌ์šฉํ•œ ํŒŒ๋ผ๋ฏธํ„ฐ๋Š” ์ตœ์ดˆ$\epsilon=0.01 \sim 2$, $C=2 \sim 256$, $\gamma=0.01 \sim 1$์˜ ๋ฒ”์œ„์ด๋‹ค. R ํ”„๋กœ๊ทธ๋žจ์„ ์ด์šฉํ•˜์—ฌ ๋ชจ๋ธ์˜ RMSE๊ฐ€ ์ตœ์†Œํ™”๋˜๋Š” ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ์„ ์ •ํ•˜์˜€์œผ๋ฉฐ, $\epsilon=0.01$, $\quad C=2$, $\quad \gamma=0.01$๋กœ ์„ ์ •๋˜์—ˆ๋‹ค. ํ•™์Šต๋ชจ๋ธ์— ํฌ๋ฆฌ๊น… ๊ธฐ๋ฒ•์„ ์ด์šฉํ•œ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋ฅผ ์ž…๋ ฅํ•˜์—ฌ ํƒœ์–‘๊ด‘ ๋ฐœ์ „ ์ถœ๋ ฅ ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ–ˆ๋‹ค.

4. ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก ๊ฒฐ๊ณผ

ํƒœ์–‘๊ด‘ ๋ฐœ์ „ ์ถœ๋ ฅ ์˜ˆ์ธก ํ‰๊ฐ€๋ฅผ ์œ„ํ•ด MAPE, RMSE์„ ์ด์šฉํ•˜์—ฌ ๊ฒฐ๊ณผ๋ฅผ ํ‰๊ฐ€ํ–ˆ์œผ๋ฉฐ, ๋‹ค์Œ์˜ ์‹ (10), ์‹ (11)๊ณผ ๊ฐ™๋‹ค[12]. N์€ ์‹œ๊ฐ„ ๋ฒ”์œ„๋ฅผ ์˜๋ฏธํ•˜๋ฉฐ, ๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ํƒœ์–‘๊ด‘ ์ถœ๋ ฅ์ด ๋ฐœ์ƒํ•˜๋Š” 8์‹œ๊ฐ„์œผ๋กœ ์„ค์ •ํ–ˆ๋‹ค. $\hat{p}_{i}$๋Š” ์˜ˆ์ธก๊ฐ’, $p_{i}$๋Š” ์‹ค์ œ ์ถœ๋ ฅ์„ ์˜๋ฏธํ•œ๋‹ค.

(10)
$M A P E=\frac{1}{N} \sum_{i=1}^{N}\left|\frac{\hat{p}_{i}-p_{i}}{\text { capacity }}\right|$

(11)
$\operatorname{RMSE}=\sqrt{\frac{1}{\mathrm{N}} \sum_{\mathrm{i}=1}^{\mathrm{N}}\left(\hat{\mathrm{p}}_{\mathrm{i}}-\mathrm{p}_{\mathrm{i}}\right)^{2}}$

2016๋…„ 5์›” 7์ผ~10์„ ์˜ˆ์ธกํ•œ ๊ฒฐ๊ณผ๋Š” ๊ทธ๋ฆผ. 5์™€ ๊ฐ™๋‹ค. ๊ทธ๋ฆผ. 5์—์„œ ๋ณด๋“ฏ์ด SVR์˜ ์˜ˆ์ธก์ด ์‹ค์ œ ์ถœ๋ ฅ๊ณผ ์œ ์‚ฌํ•˜๊ฒŒ ์˜ˆ์ธก๋˜๋Š” ๊ฒƒ์„ ํ™•์ธํ•  ์ˆ˜ ์žˆ๋‹ค.

Fig. 5. Results of Forecasting of May 7-10, 2016

../../Resources/kiiee/JIEIE.2019.33.6.042/fig5.png

5์›” 1์ผ์—์„œ 5์›” 31์ผ๊นŒ์ง€์˜ ํƒœ์–‘๊ด‘๋ฐœ์ „ ๋‹จ์ง€์˜ ์˜ˆ์ธก ๊ฒฐ๊ณผ์— ๋Œ€ํ•œ ํ‰๊ฐ€๋Š” ๊ทธ๋ฆผ. 6, ๊ทธ๋ฆผ. 7๊ณผ ๊ฐ™๋‹ค. 5์›”์˜ ARIMAX ๋ชจ๋ธ์— ๋Œ€ํ•œ ํ‰๊ท  MAPE ๋ฐ RMSE๋Š” 21.67%, 2.56MW๋ฅผ ๋‚˜ํƒ€๋ƒˆ๋‹ค. ์ง€์†์„ฑ ๊ธฐ๋ฒ•์— ๋Œ€ํ•œ ํ‰๊ท  MAPE ๋ฐ RMSE๋Š” 22.87%, 2.52MW๋ฅผ ๋‚˜ํƒ€๋ƒˆ๋‹ค. SVR ๋ชจ๋ธ์— ๋Œ€ํ•œ ํ‰๊ท  MAPE ๋ฐ RMSE๋Š” 7.44%, 0.86MW๋ฅผ ๋‚˜ํƒ€๋ƒˆ๋‹ค. ์˜ˆ์ธก ์ผ์‚ฌ๋Ÿ‰์— ๋Œ€ํ•œ ์˜ค์ฐจ๊ฐ€ ๋†’์€ ๊ฒฝ์šฐ MAPE ๋˜ํ•œ ๋†’๊ฒŒ ๋‚˜ํƒ€๋‚˜๋Š” ๊ฒƒ์„ ์•Œ ์ˆ˜ ์žˆ์œผ๋ฉฐ, ํ–ฅํ›„ ์ •ํ™•ํ•œ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ•˜๋ฉด ์˜ˆ์ธก์˜ ์ •ํ™•๋„ ๋˜ํ•œ ํ–ฅ์ƒ๋  ๊ฒƒ์œผ๋กœ ์ƒ๊ฐ๋œ๋‹ค.

Fig. 6. MAPE for May 2016

../../Resources/kiiee/JIEIE.2019.33.6.042/fig6.png

Fig. 7. RMSE for May 2016

../../Resources/kiiee/JIEIE.2019.33.6.042/fig7.png

5. ๊ฒฐ ๋ก 

๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ํƒœ์–‘๊ด‘๋ฐœ์ „์˜ ์ถœ๋ ฅ์„ ์˜ˆ์ธกํ•˜์—ฌ ๊ณ„ํ†ต์˜ ์•ˆ์ •๋„๋ฅผ ํ–ฅ์ƒํ•˜๊ธฐ ์œ„ํ•ด ํƒœ์–‘๊ด‘๋ฐœ์ „ ์˜ˆ์ธก ๋ฐฉ์•ˆ์„ ์ œ์‹œํ•˜์˜€๋‹ค. ํฌ๋ฆฌ๊น… ๊ธฐ๋ฒ•์„ ์ ์šฉํ•˜์—ฌ ์ผ์‚ฌ๋Ÿ‰ ๋ฐ์ดํ„ฐ๊ฐ€ ์—†๋Š” ํƒœ์–‘๊ด‘๋ฐœ์ „๋‹จ์ง€์˜ ์ผ์‚ฌ๋Ÿ‰์„ ์˜ˆ์ธกํ•˜๊ณ , ๊ณผ๊ฑฐ ์ถœ๋ ฅ ๋ฐ์ดํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ์˜ˆ์ธก ๋ชจ๋ธ์„ ํ•™์Šตํ•˜์˜€๋‹ค. ์‹ค์‹œ๊ฐ„ ๊ณ„ํ†ต์— ์ ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ๋น ๋ฅธ ๊ณ„์‚ฐ์†๋„๋ฅผ ๊ฐ€์ง„ SVR ๋ชจ๋ธ์„ ์ด์šฉํ•˜์—ฌ ํƒœ์–‘๊ด‘๋ฐœ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ–ˆ๋‹ค. ์˜ˆ์ธก์ผ ์ „ 720์‹œ๊ฐ„์— ๋Œ€ํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์ด์šฉํ•˜์—ฌ ๋ชจ๋ธ์˜ ํ›ˆ๋ จ์„ ์ˆ˜ํ–‰ํ•˜๊ณ , ํ•˜๋ฃจ ์ „ ์ถœ๋ ฅ์˜ˆ์ธก์„ ์‹œํ–‰ํ•˜์˜€๋‹ค. ์ˆ˜ํ–‰๊ฒฐ๊ณผ SVR ๋ชจ๋ธ์˜ 5์›” ํ‰๊ท  MAPE๋Š” 7.44%, ํ‰๊ท  RMSE๋Š” 0.86MW์˜ ๊ฒฐ๊ณผ๋ฅผ ํ™•์ธํ•˜์˜€์œผ๋ฉฐ, ํƒ€ ๋ชจ๋ธ์— ๋น„ํ•ด ์•ฝ 14% ์ •๋„ ๊ฐ์†Œํ•˜์˜€๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ๋ณธ ์—ฐ๊ตฌ์—์„œ๋Š” ํƒœ์–‘๊ด‘ ๋ฐœ์ „ ์˜ˆ์ธก ์‹œ SVR ๋ชจ๋ธ์„ ์ด์šฉํ•˜๋Š” ๊ฒฝ์šฐ ์ •ํ™•๋„๊ฐ€ ๋†’์•„์ง์„ ์•Œ ์ˆ˜ ์žˆ์—ˆ๋‹ค.

ํ–ฅํ›„ ์—ฐ๊ตฌ์—์„œ๋Š” ๊ธฐ์กด ๊ฒฐ์ •๋ก ์  ์˜ˆ์ธก์—์„œ ๋ณ€๋™์„ฑ ๋ถ„์„์„ ๋„์ž…ํ•˜์—ฌ ํ™•๋ฅ ๋ก ์  ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•  ์˜ˆ์ •์ด๋ฉฐ, ํ™•๋ฅ ๋ก ์  ์˜ˆ์ธก์„ ํ†ตํ•ด ๊ณ„ํ†ต ์˜์‚ฌ ๊ฒฐ์ •์ž์—๊ฒŒ ์ถ”๊ฐ€์ ์ธ ์ง€ํ‘œ๋ฅผ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ๋Š” ์˜ˆ์ธก ๋ชจ๋ธ ์—ฐ๊ตฌ ๋˜ํ•œ ์ˆ˜ํ–‰๋˜์–ด์•ผ ํ•  ๊ฒƒ์œผ๋กœ ์ƒ๊ฐ๋œ๋‹ค.

Acknowledgements

๋ณธ ์—ฐ๊ตฌ๋Š” ์‚ฐ์—…ํ†ต์ƒ์ž์›๋ถ€(MOTIE)์™€ ํ•œ๊ตญ์—๋„ˆ์ง€๊ธฐ์ˆ ํ‰๊ฐ€์›(KETEP)์˜ ์ง€์›์„ ๋ฐ›์•„ ์ˆ˜ํ–‰ํ•œ ์—ฐ๊ตฌ ๊ณผ์ œ์ž…๋‹ˆ๋‹ค.(No. 20161210200560)

This work was supported by the Korea Institute of Energy Technology Evaluation and Planning(KETEP) and the Ministry of Trade, Industry & Energy(MOTIE) of the Republic of Korea (No. 20161210200560).

References

1 
International Energy Agency , 2018, Renewable 2018Google Search
2 
Oโ€™Neill Barbara, 2016, Forecasting Wind, NRELGoogle Search
3 
Vermont T., 2014, New Forecasting Tool Enhance Wind Energy Integration In Idaho and Oregon, U.S Department of EnergyGoogle Search
4 
Antonanzas J., Osorio N., Escobar R., Urrarca R., Martinez-de-Prison F. J., et al , 2016, Review of Photovoltaic Power Forecasting, Solar Energy, Vol. 136, pp. 78-111DOI
5 
Smola A. J., Schรถlkopf B., 2004, A Tutorial on Support Vector Regression, Statistics and Computing, Vol. 12, No. 5, pp. 199-222DOI
6 
Drucker H., Burges C. J., Kaufman L., Smola A. J., et al , 1997, Support Vector Regression Machines, In Advances in neural information processing systems, pp. 155-161Google Search
7 
Hearst M. A., Dumais S. T., Osuna E., Platt J., Scholkopf B., 1998, Support vector machines, IEEE Intelligent Systems and their applications, Vol. 13, No. 4, pp. 18-28Google Search
8 
Stein M. L., 2012, Interpolation of spatial data: some theory for kriging, Springer Science & Business MediaGoogle Search
9 
Montero J. M., Mateu J., 2015, Spatial and Spatio-Temporal Geostatistical Modeling and Kriging, John Wiley & SonsGoogle Search
10 
Kim D. H., Hur J., 2018, Short-term probabilistic forecasting of wind energy resources using the enhanced ensemble method, Vol. 157, pp. 211-226DOI
11 
Nielsen T. S., Joensen A., Landberg L., Giebel G., 1998, A New Reference for Wind Power Forecasting, Wind Energy, Vol. 1, No. 1, pp. 29-34DOI
12 
Zhang J., Hodge B. M., et al , 2013, Metrics for Evaluating the Accuracy of Solar Power Forecasting, National Renewable Energy LaboratoryGoogle Search

Biography

Ki-Han Kim
../../Resources/kiiee/JIEIE.2019.33.6.042/au1.png

KiHan Kim has B.S. degree in Electrical Engineering from Sangmyung University, Seoul, Korea, in 2018.

Jin Hur
../../Resources/kiiee/JIEIE.2019.33.6.042/au2.png

Jin Hur received his B.S., M.S. degrees in Electrical Engineering from Korea University, Seoul, Korea, in 1997 and 1999, respectively, and a Ph.D. degree in Electrical and Computer Engineering from the University of Texas at Austin in 2012.