Offshore Wind ... - IEEE Xplore
Cross-correlation Study of Onshore/Offshore Wind Generation and Load in Texas Shijia Zhao, Student Member, IEEE, Le Xie, Member, IEEE, Chanan Singh, Fellow, IEEE Department of Electrical and Computer Engineering Texas A&M University, College Station, 77843, TX USA Email: [email protected], [email protected], [email protected] Abstract—In this paper, cross-correlation study between wind generation and load in Texas has been conducted. Through comparison between onshore and potential offshore wind farm sites, offshore wind farm proved to have slightly higher correlation with load than onshore wind farm. The results also show significant seasonal changes, which could be of importance in making operational and planning decisions.
I.
INTRODUCTION
With higher emphasis on reducing carbon emissions and promoting sustainable economic development, more and more renewable energy resources are being integrated into electric power system throughout the world. Among several renewable energy options, wind has proved to be a promising technology. It is rapidly becoming generation technology of great significance around the world [1]. Despite the rapid development of wind energy, there are still major challenges in integrating them into the electric grid. Due to high variation and limited predictability of wind energy, it is difficult to integrate wind energy into power systems efficiently [1]. Most existing wind generation sites nowadays are in land. However, due to the transmission requirement and limited land resources, offshore wind generation has begun to penetrate the market with both opportunities and challenges. The first offshore wind farm was installed in Denmark in 1991. United States has been relatively slower in adopting offshore wind, but as indicated in recent reports of Department of Energy [2, 3], United States is on its way to build the world-class offshore wind farm industry. One of the key barriers to wind integration to the power grid is that intermittent wind and electric loads are both considered as “must-take” components. Therefore, it would be desirable to have strong correlation between available wind generation and load. Along this line of research to study the correlation between wind generation and load, Carlin introduced a theoretical model in 1982 for estimating capacity credit of wind power with correlation between wind speed and electricity demand in [4]. In 1983, empirical study in [5] proved the basic validity of the model, but also pointed out that wind power generation in
978-1-4799-1255-1/13/$31.00 ©2013 IEEE
certain period of time with high load had fundamental effect on the capacity credit and need to be carefully investigated separately. In light of this observation, results in this paper are shown in monthly view to display how the dramatic seasonal changes of wind power generation adapt to the yearly portfolio of load. In recent years, several independent system operators have been conducting research and testing in correlation study, but different from this paper, they were focused on the auto and cross correlation of wind forecast errors like data analysis conducted by California ISO in [6, 7]. Correlation studies between wind and load have been used in several research areas such as capacity credit of wind power [4, 5], power spectrum density of wind power [8, 9], transmission and interconnection effect on wind integration [10, 11] and reliability implications of integrating these sources [12]. All of this research mainly relied on theoretical models instead of actual data analysis. Data analysis in this paper, thus could serve as the first hand study results of Texas areas. In this paper, two sets of data analysis were conducted focusing on the cross-correlation between load and both onshore and offshore wind generation in Texas, which have not been thoroughly studied before. In [13], Baldick suggested that due to the limited predictability of wind and demand correlation, it was hard to introduce this relationship into ERCOT operation. Correlation patterns in certain areas found in this paper could be used as guidelines of wind generation companies and demand response programs. As the state in the U.S. with highest wind penetration capacity, this study could help decision makers to better understand the correlation between loads and onshore/offshore wind farms in Texas. This paper is organized as follows. Section II introduces the data source and assumptions used in data analysis. Section III describes the definition of cross-correlation and its actual representation. In section IV and V, two sets of data analysis regarding both onshore and offshore wind farm are conducted. Conclusion and future directions of works are discussed in section VI.
II.
DATA SOURCE
A. Load Data The load data were collected from Electric Reliability Council of Texas (ERCOT), ERCOT has historical data of load since year 1995 [14]. The load sampling rate is 60 minutes. In this paper, data of year 2010 and year 2012 were used, and ERCOT “total load” were chosen in the correlation study. B. Wind Turbine Model In order to find the cross-correlation between load and wind generation, both data are necessary, however, neither onshore nor offshore wind farm in Texas provided any historical data that were sufficient to conduct data analysis. Thus, in this paper, wind power generation was calculated from the wind speed with the help of wind turbine model. In this paper, wind turbine “GE-1.5-77” was chosen for the wind turbine model. According to General Electric (GE), it is currently the most widely used wind turbine in the wind generation industry [15]. With the data provided by GE, polynomial regression was used to build the wind turbine model. As shown in Fig. 1, the wind turbine model represents the relationship between wind speed and wind power generation in polynomial. Equation (1) is the regression formula of wind power calculation 0 0.1276v5 -5.5925v4 P= +90.012v3 -652.15v2 +2251.5v-2977.5 1500
(v
I.
INTRODUCTION
With higher emphasis on reducing carbon emissions and promoting sustainable economic development, more and more renewable energy resources are being integrated into electric power system throughout the world. Among several renewable energy options, wind has proved to be a promising technology. It is rapidly becoming generation technology of great significance around the world [1]. Despite the rapid development of wind energy, there are still major challenges in integrating them into the electric grid. Due to high variation and limited predictability of wind energy, it is difficult to integrate wind energy into power systems efficiently [1]. Most existing wind generation sites nowadays are in land. However, due to the transmission requirement and limited land resources, offshore wind generation has begun to penetrate the market with both opportunities and challenges. The first offshore wind farm was installed in Denmark in 1991. United States has been relatively slower in adopting offshore wind, but as indicated in recent reports of Department of Energy [2, 3], United States is on its way to build the world-class offshore wind farm industry. One of the key barriers to wind integration to the power grid is that intermittent wind and electric loads are both considered as “must-take” components. Therefore, it would be desirable to have strong correlation between available wind generation and load. Along this line of research to study the correlation between wind generation and load, Carlin introduced a theoretical model in 1982 for estimating capacity credit of wind power with correlation between wind speed and electricity demand in [4]. In 1983, empirical study in [5] proved the basic validity of the model, but also pointed out that wind power generation in
978-1-4799-1255-1/13/$31.00 ©2013 IEEE
certain period of time with high load had fundamental effect on the capacity credit and need to be carefully investigated separately. In light of this observation, results in this paper are shown in monthly view to display how the dramatic seasonal changes of wind power generation adapt to the yearly portfolio of load. In recent years, several independent system operators have been conducting research and testing in correlation study, but different from this paper, they were focused on the auto and cross correlation of wind forecast errors like data analysis conducted by California ISO in [6, 7]. Correlation studies between wind and load have been used in several research areas such as capacity credit of wind power [4, 5], power spectrum density of wind power [8, 9], transmission and interconnection effect on wind integration [10, 11] and reliability implications of integrating these sources [12]. All of this research mainly relied on theoretical models instead of actual data analysis. Data analysis in this paper, thus could serve as the first hand study results of Texas areas. In this paper, two sets of data analysis were conducted focusing on the cross-correlation between load and both onshore and offshore wind generation in Texas, which have not been thoroughly studied before. In [13], Baldick suggested that due to the limited predictability of wind and demand correlation, it was hard to introduce this relationship into ERCOT operation. Correlation patterns in certain areas found in this paper could be used as guidelines of wind generation companies and demand response programs. As the state in the U.S. with highest wind penetration capacity, this study could help decision makers to better understand the correlation between loads and onshore/offshore wind farms in Texas. This paper is organized as follows. Section II introduces the data source and assumptions used in data analysis. Section III describes the definition of cross-correlation and its actual representation. In section IV and V, two sets of data analysis regarding both onshore and offshore wind farm are conducted. Conclusion and future directions of works are discussed in section VI.
II.
DATA SOURCE
A. Load Data The load data were collected from Electric Reliability Council of Texas (ERCOT), ERCOT has historical data of load since year 1995 [14]. The load sampling rate is 60 minutes. In this paper, data of year 2010 and year 2012 were used, and ERCOT “total load” were chosen in the correlation study. B. Wind Turbine Model In order to find the cross-correlation between load and wind generation, both data are necessary, however, neither onshore nor offshore wind farm in Texas provided any historical data that were sufficient to conduct data analysis. Thus, in this paper, wind power generation was calculated from the wind speed with the help of wind turbine model. In this paper, wind turbine “GE-1.5-77” was chosen for the wind turbine model. According to General Electric (GE), it is currently the most widely used wind turbine in the wind generation industry [15]. With the data provided by GE, polynomial regression was used to build the wind turbine model. As shown in Fig. 1, the wind turbine model represents the relationship between wind speed and wind power generation in polynomial. Equation (1) is the regression formula of wind power calculation 0 0.1276v5 -5.5925v4 P= +90.012v3 -652.15v2 +2251.5v-2977.5 1500
(v
