A robust probabilistic wind power forecasting method considering wind scenarios

Jie Yan, Yongqian Liu, Shuang Han, Chenghong Gu, Furong Li

Research output: Contribution to conferencePaper

6 Citations (Scopus)

Abstract

Wind power forecasting is one of the cheapest and direct methods to alleviate negative impacts on power system reliability and stability from intermittent wind generation. Compared with deterministic forecasts, probabilistic forecasts can provide additional information concerning wind uncertainty for economic operation and efficient trading. However, it is far from ideal with respect to the accuracy, reliability and sharpness, since the wind shows strong variable property. In this paper, a robust probabilistic wind power forecasting method is proposed as (RPWPF) that can reflect the variability of wind generation under different wind conditions. The wind scenarios are identified concerning wind generation process and dominance of wind direction in a wind farm. And then, forecasting models for each scenario can be established and executed separately so that model parameters, such as kernel function and kernel width will be adjusted with the changing external wind conditions, wind speed and direction, in a real time operation. In this way, the forecasting model will provide more fined information on power outputs and their variabilities under different wind conditions. This proposed model is validated through comparison between the simulated power outputs and their variabilities under differing wind speeds and directions with the actual outputs on a practical 183 MW wind farm in northwest China. The results show that RPWPF achieves lower root mean square error comparing with artificial neural network model, while higher skill score for forecasting interval comparing with quantile regression. Finally, a sensitivity analysis is carried out to investigate the contribution of individual input (NWP variables) to help optimize the dimension of forecasting model.

Conference

Conference3rd Renewable Power Generation Conference, RPG 2014
CountryUK United Kingdom
CityNaples
Period24/09/1425/09/14

Fingerprint

Wind power
Farms
Mean square error
Sensitivity analysis
Neural networks

Keywords

  • Power curve
  • Power generation process
  • Probabilistic forecast
  • Robust
  • Wind power

Cite this

Yan, J., Liu, Y., Han, S., Gu, C., & Li, F. (2014). A robust probabilistic wind power forecasting method considering wind scenarios. Paper presented at 3rd Renewable Power Generation Conference, RPG 2014, Naples, UK United Kingdom. https://doi.org/10.1049/cp.2014.0828

A robust probabilistic wind power forecasting method considering wind scenarios. / Yan, Jie; Liu, Yongqian; Han, Shuang; Gu, Chenghong; Li, Furong.

2014. Paper presented at 3rd Renewable Power Generation Conference, RPG 2014, Naples, UK United Kingdom.

Research output: Contribution to conferencePaper

Yan, J, Liu, Y, Han, S, Gu, C & Li, F 2014, 'A robust probabilistic wind power forecasting method considering wind scenarios' Paper presented at 3rd Renewable Power Generation Conference, RPG 2014, Naples, UK United Kingdom, 24/09/14 - 25/09/14, . https://doi.org/10.1049/cp.2014.0828
Yan J, Liu Y, Han S, Gu C, Li F. A robust probabilistic wind power forecasting method considering wind scenarios. 2014. Paper presented at 3rd Renewable Power Generation Conference, RPG 2014, Naples, UK United Kingdom. https://doi.org/10.1049/cp.2014.0828
Yan, Jie ; Liu, Yongqian ; Han, Shuang ; Gu, Chenghong ; Li, Furong. / A robust probabilistic wind power forecasting method considering wind scenarios. Paper presented at 3rd Renewable Power Generation Conference, RPG 2014, Naples, UK United Kingdom.
@conference{1f98ccfeb27645f7ac26ea96fe9dd74b,
title = "A robust probabilistic wind power forecasting method considering wind scenarios",
abstract = "Wind power forecasting is one of the cheapest and direct methods to alleviate negative impacts on power system reliability and stability from intermittent wind generation. Compared with deterministic forecasts, probabilistic forecasts can provide additional information concerning wind uncertainty for economic operation and efficient trading. However, it is far from ideal with respect to the accuracy, reliability and sharpness, since the wind shows strong variable property. In this paper, a robust probabilistic wind power forecasting method is proposed as (RPWPF) that can reflect the variability of wind generation under different wind conditions. The wind scenarios are identified concerning wind generation process and dominance of wind direction in a wind farm. And then, forecasting models for each scenario can be established and executed separately so that model parameters, such as kernel function and kernel width will be adjusted with the changing external wind conditions, wind speed and direction, in a real time operation. In this way, the forecasting model will provide more fined information on power outputs and their variabilities under different wind conditions. This proposed model is validated through comparison between the simulated power outputs and their variabilities under differing wind speeds and directions with the actual outputs on a practical 183 MW wind farm in northwest China. The results show that RPWPF achieves lower root mean square error comparing with artificial neural network model, while higher skill score for forecasting interval comparing with quantile regression. Finally, a sensitivity analysis is carried out to investigate the contribution of individual input (NWP variables) to help optimize the dimension of forecasting model.",
keywords = "Power curve, Power generation process, Probabilistic forecast, Robust, Wind power",
author = "Jie Yan and Yongqian Liu and Shuang Han and Chenghong Gu and Furong Li",
year = "2014",
doi = "10.1049/cp.2014.0828",
language = "English",
note = "3rd Renewable Power Generation Conference, RPG 2014 ; Conference date: 24-09-2014 Through 25-09-2014",

}

TY - CONF

T1 - A robust probabilistic wind power forecasting method considering wind scenarios

AU - Yan, Jie

AU - Liu, Yongqian

AU - Han, Shuang

AU - Gu, Chenghong

AU - Li, Furong

PY - 2014

Y1 - 2014

N2 - Wind power forecasting is one of the cheapest and direct methods to alleviate negative impacts on power system reliability and stability from intermittent wind generation. Compared with deterministic forecasts, probabilistic forecasts can provide additional information concerning wind uncertainty for economic operation and efficient trading. However, it is far from ideal with respect to the accuracy, reliability and sharpness, since the wind shows strong variable property. In this paper, a robust probabilistic wind power forecasting method is proposed as (RPWPF) that can reflect the variability of wind generation under different wind conditions. The wind scenarios are identified concerning wind generation process and dominance of wind direction in a wind farm. And then, forecasting models for each scenario can be established and executed separately so that model parameters, such as kernel function and kernel width will be adjusted with the changing external wind conditions, wind speed and direction, in a real time operation. In this way, the forecasting model will provide more fined information on power outputs and their variabilities under different wind conditions. This proposed model is validated through comparison between the simulated power outputs and their variabilities under differing wind speeds and directions with the actual outputs on a practical 183 MW wind farm in northwest China. The results show that RPWPF achieves lower root mean square error comparing with artificial neural network model, while higher skill score for forecasting interval comparing with quantile regression. Finally, a sensitivity analysis is carried out to investigate the contribution of individual input (NWP variables) to help optimize the dimension of forecasting model.

AB - Wind power forecasting is one of the cheapest and direct methods to alleviate negative impacts on power system reliability and stability from intermittent wind generation. Compared with deterministic forecasts, probabilistic forecasts can provide additional information concerning wind uncertainty for economic operation and efficient trading. However, it is far from ideal with respect to the accuracy, reliability and sharpness, since the wind shows strong variable property. In this paper, a robust probabilistic wind power forecasting method is proposed as (RPWPF) that can reflect the variability of wind generation under different wind conditions. The wind scenarios are identified concerning wind generation process and dominance of wind direction in a wind farm. And then, forecasting models for each scenario can be established and executed separately so that model parameters, such as kernel function and kernel width will be adjusted with the changing external wind conditions, wind speed and direction, in a real time operation. In this way, the forecasting model will provide more fined information on power outputs and their variabilities under different wind conditions. This proposed model is validated through comparison between the simulated power outputs and their variabilities under differing wind speeds and directions with the actual outputs on a practical 183 MW wind farm in northwest China. The results show that RPWPF achieves lower root mean square error comparing with artificial neural network model, while higher skill score for forecasting interval comparing with quantile regression. Finally, a sensitivity analysis is carried out to investigate the contribution of individual input (NWP variables) to help optimize the dimension of forecasting model.

KW - Power curve

KW - Power generation process

KW - Probabilistic forecast

KW - Robust

KW - Wind power

UR - http://www.scopus.com/inward/record.url?scp=84912096322&partnerID=8YFLogxK

UR - http://dx.doi.org/10.1049/cp.2014.0828

U2 - 10.1049/cp.2014.0828

DO - 10.1049/cp.2014.0828

M3 - Paper

ER -