Establishing power systems with a high share of renewable energy sources is a pivotal step toward achieving a globally sustainable transition to green and low-carbon energy. This study focuses on low-output wind power that affects the generation capacity of power systems with a high share of renewable energy sources. Utilizing the Coupled Model Intercomparison Project Phase 6 datasets, a predictive model for low-output wind power was employed to investigate regional trends worldwide. The frequency and duration of low-output wind-power events exhibited increasing trends globally, particularly in East Asia and South America, but not in North America. By 2060, the annual total days with low-output wind power in East Asia and South America could rise to 13 and 5 d, and the maximum continuous duration of low-output wind power could reach 5 and 2 d, respectively. As wind power becomes a primary electricity source, such low output could lead to shortages in energy supply within the power system, triggering large-scale power outages. This issue calls for critical attention when establishing power systems with a high share of renewable energy sources. The conclusions provide a basis for analyzing power supply risks and configuring flexible power sources for scenarios with a high share of renewable energy.
Green and low-carbon energy solutions are being widely adopted around the world [1], [2]. Among the strategies employed to achieve energy conservation and emission reduction, the construction of power systems with a substantial proportion of renewable energy sources stands out as a key approach, replacing conventional energy sources with eco-friendly alternatives [3]. As of December 2023, the global installed capacity of wind power has reached an impressive 1.02 TW [4]—a figure projected to experience exponential growth, surpassing 11 TW by 2030 [5]. The increasing incorporation of renewable energy installations, including wind power plants, amplifies the weather dependency of power systems, particularly when a high share of renewable energy sources is attained [6], [7], [8]. As renewable energy sources such as wind power gradually assume the role of primary electricity providers, the occurrence of low-output wind power, often triggered by weather phenomena such as prolonged calm or low wind speeds, can result in energy shortages within these power systems. This phenomenon has the potential to significantly undermine the power-generation capacity of these systems, leading to incidents akin to the Texas load shedding and price surges witnessed in February 2021 [9]. Such occurrences directly impact the safe and stable operation and dependable power-generation capacity of systems heavily reliant on renewable energy sources [10], [11], [12], posing considerable challenges to the global transition toward green and low-carbon energy. Hence, it has become imperative to comprehensively study the shifting patterns of low-output wind power.
To date, research on low-output wind power has been relatively limited. Early studies can be traced back to the 1970s, when research focused on the reliability of wind-power supply in California [13]. Subsequent studies delved into wind-power analysis in the midwest region of the United States [14]. Following the consensus reached in the Paris Agreement [1], countries worldwide embarked on the development of their renewable energy resources. The German government [15], [16] and media [17], [18] were among the early entities to address the issue of low-output wind power, contributing to the formulation of its initial definition [19], which, in turn, spurred more extensive research on these occurrences. Present research on low-output wind power primarily revolves around the analysis of historical events. These studies can be broadly categorized into three types based on the underlying data source. The first type involves the analyses of observed wind-speed data [20], [21], specifically examining wind-speed sequences within limited temporal and spatial observations using predefined low wind-speed thresholds. However, this approach has limitations stemming from its inadequate spatial coverage, its insufficient temporal duration, and the uncertainty introduced by local climatic effects in translating localized wind speeds into regional wind-power output. The limited spatial coverage restricts these studies from encompassing extensive potential development areas, thereby diminishing the general applicability of their findings. While expanding the modeling of relationships [22] between wind-power development areas and low-output events can partially address the spatial coverage issue, the problem of information loss due to insufficient temporal data remains unresolved.
The second category of research aims to tackle the heightened uncertainty associated with mapping localized wind speeds to regional wind-power output, as observed in the first type of studies. Research within this category relies on sequences of wind-power output data obtained either through the conversion of wind speeds measured by wind turbine units [23], [24], [25], [26] or through direct measurement. By setting a low-output threshold, this approach enables the study of the frequency and duration of low-output wind power [27]. However, it still falls short in addressing the challenges of inadequate spatial coverage and insufficient temporal duration.
The third type of research concentrates on mitigating the spatiotemporal limitations inherent in the study of low-output wind power. Drawing from a reanalysis of historical wind-speed data [28], [29], this approach constructs a conversion model linking reanalyzed wind speeds and wind-power output to generate gridded wind-power output sequences. This methodology [30], [31], [32], [33] facilitates the investigation of the frequency and duration of low-output wind-power occurrences and has revealed valuable insights into the frequency distribution of low-output events resembling a Poisson-like process [34], with heavy-tailed characteristics [35], and with other pertinent patterns [36], [37]. These findings significantly advance our understanding of low-output wind power.
In summary, current research predominantly focuses on historical low-output events, while investigations into future patterns of low output remain scarce. The evolving characteristics of wind-energy resources under the influence of global climate change are reshaping fluctuation patterns, making such patterns increasingly distinct. Therefore, relying solely on historical low-output wind-power patterns is inadequate for guiding the construction of power systems with a substantial share of renewable energy sources.
This study centers on the trends in low-output wind power resulting from climate change and develops a predictive model for low-output wind power based on the Coupled Model Intercomparison Project Phase 6 (CMIP6) datasets [38], [39]. It achieves this by investigating the statistical relationship between wind-power output levels and wind-speed rankings. Using this model, we analyze future patterns of low-output wind power in major wind-power development regions, including East Asia, Western Europe, and North America. The primary objective of this study is to uncover the evolving dynamics of wind-energy resources in the context of climate change, thereby supporting the stable and reliable transition from conventional fossil fuels to clean and sustainable wind power.
2. Modeling and validation
The variability of future wind-energy resources forms the foundation for analyzing low output in wind power. This study primarily utilized CMIP6 data and employed a quantile-based mapping model to establish a statistical relationship between wind-power output and wind-speed levels.
2.1. Introduction to CMIP6 data
The Coupled Model Intercomparison Project (CMIP) model data is the most commonly employed analytical data for researchers and engineers investigating future variations in wind-energy resources. It contains more than 30 sets of data from different climate forecast models. However, research on low-output wind power requires the time resolution of the data to ideally reach the hour level or at least be no lower than days. Therefore, the models available for this study are only nine sets of data from the latest CMIP6 dataset, as illustrated in Table 1. These datasets have a daily temporal resolution, and wind speeds are measured at a height of 10 m. Each dataset underwent analysis within three shared socioeconomic pathway (SSP) scenarios: low emissions (SSP126), medium emissions (SSP245), and high emissions (SSP585).
2.2. Statistical relationship model
The portion of wind-power output sequences falling below Tpower is defined as the low-output process. After organizing a sample of wind-power output data for a given timespan in ascending order, Tpower is taken as the quantile, and the proportion of samples below this quantile is calculated as follows:
where b is the proportion of low-output wind power, Pi is the wind power output at time i, and n is the total sample size. To ensure the stability and accuracy of the results, the sample coverage extended over at least one year.
After obtaining the proportion b of low-output wind power, corresponding regions or points are selected that align with the wind-power sequences, along with the CMIP6 forecast wind-speed sequences vi within the same time period. It is essential that the coverage duration and sample quantity of CMIP6 wind speed align precisely with the wind-power output sequence Pi. Notably, if the wind-power output is an aggregated regional sum, due to the non-additive nature of wind speeds, the forecast wind speed should be obtained by weighting the CMIP wind speeds across various locations:
where vi represents the calculated regional wind speed, Sj is the installed capacity of the jth wind farm, m is the total number of wind farms in the region, and vi,j is the CMIP6 forecast wind speed for the jth wind farm at time i.
Once the wind-power output sequence Pi is aligned with the wind-speed sequence vi, the wind-speed sequence vi is similarly arranged in ascending order. The wind speed at the quantile b is extracted as follows:
where Twind represents the CMIP6 forecast wind-speed threshold reflecting low-output wind power. To capture the variations in the future low-output wind-power process, it is necessary to evaluate wind speeds based on this threshold. It should be noted that and .
Consequently, the future low-output wind-power process detection model can be constructed in the following steps: ① Process and normalize the wind-power output sequence Pi under normal operational conditions, arranging it in ascending order. ② Extract the CMIP6 forecast wind-speed sequence vi corresponding to the wind-power output sequence Pi in terms of time and geographic location, arranging it in ascending order. If Pi is the regional total wind-power output, then the forecast wind-speed sequence vi can be obtained by weighting Eq. (2). ③ Determine the threshold Tpower for low-output wind power rationally based on analytical requirements. ④ According to Eq. (1), coupled with the normalized and ascending-ordered wind-power output sequence Pi, calculate the proportion b of low-output wind power. ⑤ Based on the obtained low-output proportion b, employ Eq. (3) along with the CMIP6 forecast wind-speed sequence vi to compute the wind-speed threshold corresponding to low output Twind. ⑥ Analyze the variations in future low-output wind power based on the wind-speed threshold Twind.
2.3. Model validation
To validate its accuracy, the model was tested using CMIP6 forecast wind-speed data and actual power-generation data from major regions in China. The modeling data spanned from January 1 to December 31, 2020. The wind-power output sequence in the modeling data was averaged daily to align the daily temporal resolution of the CMIP6 data with the 15 min resolution of the wind-power output data. This daily-averaged output sequence—essentially a regional aggregated output—was transformed using Eq. (2) to obtain the regional wind-speed sequence.
In order to analyze the changing characteristics of future low-output wind power as accurately as possible, the simulation accuracy of nine CMIP6 datasets in major regions of China in 2020–2021 was verified. Outputs below 10% of installed capacity (i.e., simultaneous rates below 0.1) were defined as low-output wind power; thus, Tpower = 0.1. Then, Eq. (1) was applied to the regional daily-averaged wind-power output samples in Fig. 1, yielding b = 1.639%. Substituting b = 1.639% into Eq. (3) and introducing the regional wind-speed sequence, the low wind-speed thresholds for different models are calculated as shown in Table 2.
Using the low wind-speed thresholds in Table 2, the low-output wind-power situation for the same region during the period from January 1 to December 31, 2021, was tested. Like the modeling data, the wind-power output data was normalized by installed capacity and averaged daily. The CMIP6 wind-speed data was weighted using Eq. (2). The method for calculating the proportion of low wind speeds based on the low wind-speed threshold is as follows:
Using the above equation, the proportion of low wind speeds from January 1 to December 31, 2021, was calculated as shown in Table 3. If Tpower = 0.1, according to Eq. (1), the actual proportion of low-output wind power in the same region from January 1 to December 31, 2021, is 1.1%. If the difference between the proportion of low wind-speed testing and the actual proportion is within ±0.6%, it can be considered as passing the statistical test. Due to the uncertainty of future carbon-emission scenarios, when verifying the simulation capability of various models for low-output wind power, as long as the simulation of one emission scenario for low-output wind power matches the actual situation, the reliability of the model is considered to satisfy the requirements. The simulation capability conclusions of each model for low-output wind power are shown in Table 3.
From Table 3, it can be seen that the AWI-CM-1-1-MR model and NorESM2-LM model have poor simulation capabilities for low-output wind power and are thus not included in subsequent related analyses. Since this study introduced seven distinct CMIP6 datasets, the forecast wind speeds from the seven models for the same time were averaged to ensure the robustness of the analysis and to account for diverse scenarios. Taking the medium-emissions scenario as an example, by aligning the daily-averaged wind-power output sequence with the regionally weighted wind-speed sequence according to time, a dataset comprising 366 samples was obtained, as depicted in Fig. 1.
In order to ensure the reliability of the research results, the above verification method for a single model was also adopted to verify the simulation ability of the seven models using average wind speed for low-output wind power, and the specific methods will not be described. Similarly, assuming that the process with an output lower than 10% of the installed capacity—that is, a simultaneous rate lower than 0.1—is set as a wind-power low-output process condition, the low wind-speed threshold calculated for the 2020 sample is Twind = 2.74 m·s−1. The proportion of low wind speeds from January 1 to December 31, 2021, was calculated to be 1.1%. This is equal to the calculated actual proportion Tpower = 0.1 of low-output wind power according to Eq. (1), which is 1.1%, supporting the conclusion that the constructed low-output event analysis model is accurate and capable of effectively analyzing future low-output wind power.
3. Future trends in low-output wind power
Future trends in low-output wind power were estimated using a low-output event extraction model, leveraging wind-power output and CMIP6 data from China’s major regions during the period from January 1 to December 31, 2020. Daily average regional low wind-speed thresholds corresponding to low output were extracted based on three emissions scenarios (Table 4).
The low wind-speed thresholds obtained for China’s major regions (Table 2) were applied to Europe, North America, South America, Oceania, and Africa because sufficient wind-power output data were not available for these regions. Furthermore, this study calculated daily average wind speeds by taking the average of all points within each region and assigning uniform weights to all points. This approach was necessary due to a lack of data or uncertainty regarding specific wind-energy resource development sites and potential future development capacity.
Under the low-emissions scenario, East Asia experiences an oscillating upward trend in both the total number of days with low output and the duration of low output (Fig. 2). The total number of days with low output increases from approximately 4 d in 2021 to approximately 13 d by 2060. Similarly, the maximum duration of low output rises from 2 d in 2021 to approximately 5 d by 2060.
In the case of the medium-emissions scenario, the total number of days with low output exhibits an oscillating upward trend (Fig. 3), growing from approximately 4 d in 2021 to around 8 d by 2060, with instances of continuous low output lasting for 6 d in 2051. The maximum duration of low-output events does not change significantly.
Under the high-emissions scenario, the total number of days with low output and the duration of low-output events display an oscillating upward trend (Fig. 4). The total number of days with low output increases from around 10 d in 2021 to approximately 13 d by 2060. The maximum duration of low-output events increases to approximately 5 d by 2060.
A comparative analysis of Fig. 2, Fig. 3, Fig. 4 demonstrates that, except for the maximum duration of low output under the medium-emissions scenario, East Asia is projected to experience an increasing trend in both the annual total number of days with low output and the maximum duration of individual low output. This growth trend is particularly pronounced under the low-emissions scenario.
3.2. Europe
Low-output events are not projected to occur on a substantial scale in Europe by 2060. However, the proportion of relatively low wind speeds is on the rise, as shown when considering a low wind-speed threshold of 3.1 m·s−1 (Fig. 5).
Under the high-emissions scenario, the annual total number of days with daily average wind speeds below 3.1 m·s−1 in Europe grow from around 0 d in 2021 to approximately 1 d by 2060 (Fig. 5). Low-output days are projected to peak in 2047, reaching 2 d. Similar patterns are observed for other emissions scenarios.
3.3. North America
The results indicate that significant low output is not projected to occur by 2060 in North America. In contrast to East Asia and Europe, the proportion of periods with relatively low wind speeds is decreasing in North America, as shown when considering a threshold of 3.0 m·s−1 for low wind speeds (Fig. 6).
Under the low-emissions scenario, the annual total number of days with daily average wind speeds below 3.0 m·s−1 in North America is projected to decrease from around 6 d in 2021 to approximately 1 d by 2060 (Fig. 6). However, this number temporarily increases to 16 d in 2033, with the maximum duration of low wind-speed events decreasing to around 1 d in 2060. Similar patterns are observed for the other emissions scenarios.
Under the low-emissions scenario, South America exhibits a gradual upward trend in both the projected total annual number of days with low-output wind power and the maximum duration of these events (Fig. 7). The total annual number of days with low output increases from approximately 0 d in 2021 to around 4 d by 2060, with the maximum duration expanding from around 0 d in 2021 to approximately 2 d by 2060.
In the medium-emissions scenario, there is relatively little change in the projected total annual number of days with low output or the maximum duration of these events for South America (Fig. 8). More specifically, the annual low-output days mainly vary between 0 and 1 d from 2021 to 2060, with a minor increase to 2 d in 2050.
Under the high-emissions scenario, the projected total annual number of days with low output and the maximum duration of these events also display a gradual upward trend for South America (Fig. 9). More specifically, the projected total annual number of days with low output grows from approximately 0 d in 2021 to approximately 5 d by the year 2060. In addition, the projected maximum duration of these low-output events increases from approximately 0 d in 2021 to approximately 2 d by the year 2060.
To sum up, South America is expected to experience a development of low-output wind-power events, regardless of the emissions scenario, with a more pronounced impact under the high-emissions scenario.
3.5. Oceania
Significant low-output events are not projected to occur in Oceania by the year 2060. However, the proportion of relatively low wind speeds also follows a diminishing trend in this region. This trend is notable when considering a threshold for low wind speed of 3.5 m·s−1, as shown in Fig. 10.
Under the low-emissions scenario, the projected total annual number of days with daily average wind speeds below 3.5 m·s−1 in Oceania grows from 0 d in 2021 to approximately 6 d by the year 2060 (Fig. 10). The peak occurrence of such low-output days is observed in 2043 and 2047, reaching 11 d. Similarly, the projected maximum duration of low wind-speed events increases from around 1 d in 2021 to approximately 3 d in 2060, with the longest durations occurring in 2047 and 2052, lasting 4 d. Similar patterns are observed for the other emissions scenarios.
3.6. Africa
Low-output events are not projected to occur in Africa before the year 2060. Nevertheless, as in the other regions, the projected proportion of relatively low wind speeds in Africa demonstrates an increasing trend. This trend becomes evident when considering a threshold for low wind speed of 3.0 m·s−1, as shown in Fig. 11.
Under the low-emissions scenario, the projected total annual number of days with daily average wind speeds below 3.0 m·s−1 in Africa increases from approximately 0 d in 2021 to approximately 3 d by the year 2060 (Fig. 11). Peak occurrences of such low-output days are projected to occur in 2028, at 6 d. Similarly, the projected maximum duration of low wind-speed events increases from approximately 0 d in 2021 to approximately 2 d in 2060. Similar patterns are observed for the other emissions scenarios.
3.7. Overall situation across the globe
By analyzing the changes in low-output wind power across major regions around the world, we established the variation slope of the projected annual total number of low-output days (i.e., the change in the number of low-output days every ten years; Fig. 12).
Furthermore, the variation slope of the maximum duration of individual low-output events was analyzed on an annual basis, as shown in Fig. 13.
The average variation slope of the annual total number of low-output days and the average variation slope of the annual maximum number of days of individual low-output events in global land regions under different emission scenarios were analyzed, as shown in Table 5.
From Table 5, it can be seen that the number of days with low-output wind power in global land regions is showing an overall increasing trend. Among the scenarios, the growth rate under the low-emissions scenario is the most significant, with an increase of 0.5 d every decade, while the growth rate under the medium-emissions scenario is the slowest, with only 0.04 d of growth every decade. However, from the perspective of the longest durations of a single low-output wind-power event, only the low-emission scenario shows an increase, while the overall longest durations of a single event decrease under the medium- and high-emissions scenarios.
4. Conclusions
Countries worldwide have reached a consensus to realize the sustainable development of green and low-carbon energy sources. In pursuit of this goal, major countries are vigorously promoting the large-scale development of wind and solar power, with the aim of substituting most fossil fuels with renewable energy sources by 2060. By then, wind and solar energy are expected to become the dominant sources of power in energy systems. This study focused on analyzing the patterns of variation in future low-output wind-power events. A mathematical model based on the CMIP6 dataset was constructed to investigate these patterns and provide support for the stable operation of power systems with a high share of renewable energy sources.
The results indicate that, in the coming years, East Asia and South America will witness occurrences of low-output wind power. The projected frequency of these occurrences and the maximum duration of individual low-output events show an increasing trend. By the year 2060, the annual total number of days with low wind-power output may reach a peak of 13 d in East Asia, with the longest continuous durations of low output reaching 5 and 2 d for East Asia and South America, respectively. This poses a significant challenge to ensuring a reliable power supply. In contrast, North America may not encounter low wind-power output before 2060, and the proportion of periods with low wind-power output is projected to decrease. In Europe, Oceania (including Australia), and Africa, although low-output events are not expected to manifest before 2060, the overall proportion of events with reduced wind-power output is projected to continue to grow. Oceania stands out in this respect, with a projected annual total of significant low-output periods potentially reaching up to 11 d, and the maximum duration of a single low-output event extending up to 4 d. Under the high-emissions scenario, the escalation of low-output wind power is anticipated to be even more pronounced.
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work is supported by the Joint Research Fund in Smart Grid (U1966601) under cooperative agreement between the National Natural Science Foundation of China (NSFC) and the State Grid Corporation of China (SGCC).
Note: Please be aware that the following content is generated by artificial intelligence. This website is not responsible for any consequences arising from the use of this content.
PDF
(17826KB)
