A wind power ramp event detection method based on eigenvalue correction and trend integration

Wind power ramp events have the characteristic of small probability and big hazard, so it is significant to improve the recognition rate and accuracy of ramp events for the safe and stable operation of the power grid. In order to improve the efficiency of ramp event detection, a wind power ramp event detection method based on eigenvalue evaluation correction and trend integration is proposed by combining the feature information of ramp events. The original wind power data is extracted with extreme value feature points and corrected with eigenvalue evaluation to achieve the trend feature extraction effect. In order to avoid being affected by small power jitter events, a trend labeling method is used to integrate the correction sequences. The actual wind power data of a wind farm in Xinjiang is used as an example for ramp event detection. The results of the case show that compared with the original swinging door algorithm, the proposed method has both better trend extraction effect and can avoid the influence of power small jitter events, and can more accurately and more completely identify ramp events.


Introduction
With the rapid growth of the global economy and the increasing demand for energy, the development of renewable energy is becoming more and more critical.wind power is a common green energy source with many significant advantages, and has become one of the first choices for governments and enterprises in the energy transition [1].However, wind power also faces a series of technical challenges.The randomness and uncertainty of wind power can have an impact on the safe and stable operation of the power grid [2].Wind power ramp events can be triggered when extreme weather event occurs.Wind power ramp events means that a large unidirectional change in wind power over a short period of time, which can seriously threaten the safe operation of the grid and even lead to frequency instability, load shedding and other problems [3].
Therefore, the ability to solve the wind power ramp problem is crucial for the sustainable development of the wind power industry as well as the safe operation of the power grid.Qu et al. [4] [4] proposed an adaptive algorithm based on parameters and resolution, which is robust and solves the impact of data resolution on the performance of the climbing detection algorithm, but does not take into account the problem of the accuracy rate of the climbing point.Zhang et al. [5] mentions the use of improved revolving door algorithm (SDT) for trend extraction of wind power data and combined with trend labeling, which is suitable for the identification of ramp events in the wind energy field.However, the SDT algorithm is too dependent on the selection of parameters and has some limitations.Kuang et al. [6] proposes a new climbing segment identification method based on the extreme value extraction technique.The method can compress the data effectively, but there are problems of omission of climbing characterizations and inaccuracy of climbing points.
Combined with the above analysis, the problems are as follows: 1) The existing algorithms are mostly based on the SDT algorithm, which optimizes the parameter-seeking process to play the role of trend extraction.These algorithms rely too much on the selection of parameters and needs to be re-parameterized for different scenarios, which has certain limitations [8].2) In the process of trend extraction of the original wind power data, there are problems such as the omission of climbing characterizations and inaccurate climbing points, which lead to the reduction in recognition accuracy.
Aiming at the above problems, this paper proposes a detection method based on eigenvalue correction and trend integration.Combined with the existing definitions, ramp events are detected and analyzed with an example of a wind farm in Xinjiang to verify the effectiveness of the proposed method.

Data compression method based on extreme value extraction and eigenvalue correction
The extreme value extraction method achieves the purpose of trend extraction and data compression by extracting the extreme value feature points from the original wind power data.However, not all the trend features between the compressed points can be included, and there is the problem of climbing characterization omission.
Aiming at the above problems, this paper proposes the method of eigenvalue correction.The correction process equations are as follows: 1) Calculate the rate of change of power at the next moment at each point in the original wind power sequence X and store it in the sequence of eigenvalues Z .
where , 1 R  represents the power change rate from moment i to moment 1 i  ; Z represents the original sequence eigenvalue.
2) Calculate the rate of change of power at the next moment at each extreme point in the extreme value sequence and store it in the sequence of eigenvalues R S . 1,2 where , 1 S j j R  represents the power change rate from moment j to moment 1 j  ; R S represents the sequence of extreme eigenvalues.
3) Conditions for determining eigenvalue corrections. , where represents the rate of power change between sequences of extreme eigenvalues contemporaneous with , 1 If the climbing characterization between the compressed points cannot be covered by the climbing characterization of the simultaneous extreme points, the compressed points are corrected by incorporating them into the extreme value sequence S .The corrected sequence is called the correction sequence Y .

Processing of data sets based on trend integration
There are some power small jitter events in the wind power data, which are called bump events and are characterized by different climbing directions and small climbing amplitudes in a certain time period, these events will lead to many ramp events not being detected completely, so it is necessary to process the data set for trend integration.
In this paper, 2% of the rated installed capacity is used as the trend judgment criterion [9].Since the capacity of the selected wind farm is 200 MW, 4 MW is selected as the criterion for trend integration processing.The judgment equation is as follows:

General framework
This paper proposes a detection method based on eigenvalue correction and trend integration, referred to as EC-TI algorithm.The algorithm carries out extreme value extraction, eigenvalue correction and trend judgment processing on the basis of wind power data, so as to achieve the effect of trend feature extraction, data compression and trend integration.The specific detection steps are shown below.
Step 1: Data compression based on extreme value extraction.Trend feature extraction of wind power data by searching for local extreme value points of numerical sequences.
Step 2: Correction of extreme value sequences.The rate of change of power between points is used for climbing characterization, and compression points that cannot be encompassed by the climbing characterization of simultaneous extreme points are included in the correction sequence.
Step 3: Trend judgment consolidation processing of datasets.Trend segmentation of datasets is performed to identify ramp events only in datasets with an upward or downward trend, thus achieving the effect of merging adjacent climbing segments.
Step 4: Combine existing definitions to detect ramp events on the dataset.Based on the EC-TI detection algorithm proposed in this paper, the detection process of ramp events can be accomplished more efficiently.

Case analysis
This paper takes a wind farm in Xinjiang as an example to verify the effectiveness of the proposed wind power ramp event detection method, which has an installed capacity of 200 MW, and the data sampling time is from January 1 to December 31, 2019, which contains a total of 35040 data with a sampling interval of 15 minutes.

Effectiveness analysis of trend extraction
Aiming at the omission of climbing characterizations and the inaccuracy of climbing points in the extreme value extraction method, this paper proposes the method of eigenvalue correction.In this section, the compression rate R C and correction rate a C are used as the evaluation indexes of the trend extraction effect.The calculation formulas are as follows: where b represents polar series data before correction; a represents new data added after correction.
By proposing a method based on climbing characterization correction, data compression is carried out under the premise of guaranteeing the effect of trend feature extraction, which further improves the detection efficiency.The effects of compression and correction are shown in Table 1

Effectiveness analysis of trend integration
In the detection process, bump events will be detected and integrated into the neighboring ramp event identification, thus affecting the effectiveness of detection, as can be seen in Figure 1, there are some bump events, and the ramp event between points 155 and 165 will be split into two sub-ramp events, which obviously destroys the integrity of the detection process.By introducing the trend integration process, the bump events are incorporated into the overall ramp events, which avoids the interruption of the ramp event recognition and effectively improves the integrity of the ramp event detection.

Effectiveness analysis of wind power ramp event detection
In this section, the EC-TI algorithm is compared with the traditional SDT algorithm for the effect of wind power ramp event detection.To do the analysis with real data from a wind farm in Xinjiang, the existing definition is used as the judgment condition, and a time interval of 1 hour is selected as the basis for detecting the occurrence of wind power ramp events when the change in wind power is greater than 20% of the rated power [10].Get the ramp event detection results graph shown in Figure 2. The comparison results of the ramp event detection results are obtained as shown in Table 2.In Figure 2, we select the power points in the time interval from 2530 to 2680 in the dataset to demonstrate the effectiveness of the algorithm proposed in this paper in recognizing ramp events.As can be seen from Table 2, compared with the traditional SDT algorithm, the EC-TI algorithm is able to detect more ramp events and can avoid the effect of power jitter events, and after the trend integration, the small fluctuation events are incorporated into the overall ramp events, avoiding the interruption of ramp event recognition.Compared with the SDT algorithm, the EC-TI algorithm is more effective in extracting the trend characteristics of wind power data, and the climbing points are more accurate, which can detect the ramp events more comprehensively and accurately, and has a more excellent detection effect.

Conclusions
This paper proposes a new method for wind power ramp event detection that combines the eigenvalue correction method and the trend integration method.After comparative analysis before and after correction and comparison with other algorithms.The main conclusions are obtained as follows.

1)
Incorporating the eigenvalue correction algorithm into the extreme value extraction algorithm significantly optimizes the trend extraction.Compared with the pure extreme value extraction method, the correction series derived from the incorporation of the eigenvalue correction algorithm maintains the optimal power fluctuation trend while achieving the data compression effect.In the dataset of a wind farm in Xinjiang discussed in this paper, the compression rate reaches 2.085-2.833and the correction rate reaches 31.5%-55.1%. 2) The application of the trend integration step can enhance the completeness of the identification of ramp events.Combining the trend extraction algorithm mentioned in this paper with the trend integration method can lead to more accurate detection of ramp events. 3) The algorithm combining eigenvalue correction with the trend integration can provide a better identification effect of ramp events.The results show that compared with the traditional SDT algorithm, the EC-TI algorithm has a better trend extraction effect, avoids the influence of bump events, and can detect the occurrence of ramp events more efficiently and completely.

P
 represents the wind power at point 1 n  ; n P represents the wind power at point n ; trend +1 T  represents an upward trend, trend -1 T  represents a downward trend, trend 0 T  represents no trend.
2023 5th International Conference on Energy, Power and Grid (ICEPG 2023) Journal of Physics: Conference Series 2703 n represents original data; m represents compressed data.

Figure 1 .
Figure 1.The results of trend integration.

Figure 2 .
Figure 2. Results of wind power ramp event detection.

Table 1 .
. Compression effect comparison before and after correction.

Table 2 .
Comparison of ramp event detection results.