An Energy Storage Type Wind Farm Ramp Event Detection Method for Optimizing the Rotating Door Algorithm

Abstract

The present invention discloses a method for detecting ramp events in an energy storage type wind farm by optimizing the swing gate algorithm. It includes the following steps: designing a judgment criterion for ramp events in the energy storage type wind farm; improving the beetle swarm algorithm and using it to search for the optimal gate width of the swing gate algorithm; using the optimized swing gate algorithm to extract the characteristic data points of wind power; adopting the four-point method to process the characteristic data points to remove the "protrusions"; regarding two adjacent characteristic data points processed by the four-point method as a wind power characteristic time period, and classifying, coding and merging it; detecting whether a ramp event occurs in the merged wind power characteristic time period according to the ramp event judgment criterion. The present invention extracts the characteristic data points of wind power, designs a method for merging wind power characteristic time periods, and gives a judgment criterion for ramp events in the energy storage type wind farm, so as to be able to effectively detect the ramp events occurring in the energy storage type wind farm and improve the detection effect of wind power ramp events.

Description

Technical Field

[0001] The present invention relates to the field of electric power systems, and in particular to a method for detecting ramp events in energy storage type wind farms. Technical Background

[0002] The realization of the "dual carbon" goal requires that new energy sources be connected to the power grid on a larger scale. Wind power generation is a mature new energy power generation technology, but its inherent intermittent and volatile nature, especially the wind power ramp events (WPRE) that occur from time to time, poses a great challenge to the safe and stable operation of the power grid. WPRE refers to the phenomenon that wind power output fluctuates greatly in a short period of time due to extreme weather such as strong winds and thunderstorms. When the wind speed increases sharply, the wind power will increase rapidly, and a corresponding ramp event will occur. In severe weather conditions, such as stormy weather, when the wind speed reaches the minimum cut-out wind speed of the wind turbine, the wind turbine will have safety protection for the wind turbine, and will actively control the wind turbine and shut it down. The wind turbine power will drop suddenly, resulting in a wind power ramp-down event.

[0003] WPRE will have a very large impact on the power grid. The rapid change in power will destroy the power supply balance of the power system, which is not conducive to the dispatcher to dispatch other generators to cooperate with wind power output. Wind power down-climbing events will cause the frequency of the power grid to decrease, affecting the safe and stable operation of the power system, and the larger the scale, the more serious the impact. In order to solve the adverse effects of WPRE on the power grid, many studies have used energy storage systems to smooth WPRE to maintain the safe and stable operation of the power grid. In this context, the study of WPRE detection methods suitable for energy storage wind farms is important for the regulation strategy of smoothing WPRE. Summary of the invention

[0004] The purpose of the present invention is to design a reasonable method for detecting ramp events in energy storage wind farms, so as to effectively detect ramp events occurring in energy storage wind farms. The present invention provides a method for detecting ramp events in energy storage wind farms that improves the swingdoor trending (SDT) algorithm optimized by a group of longhorn beetles. The method not only provides a judgment standard for ramp events in energy storage wind farms and extracts characteristic data points of wind power, but also designs a method for merging wind power characteristic time periods based on the above, thereby realizing the detection of ramp events in energy storage wind farms. Finally, the effectiveness of the detection method is verified by simulation.

[0005] The present invention adopts a technical solution: a method for detecting a slope climbing event in an energy storage type wind farm by optimizing a revolving door algorithm, which comprises the following steps:

[0006] (1) Design the judgment criteria for ramping events in energy storage wind farms;

[0007] (2) Improve the beetle swarm optimization algorithm to enhance its optimization speed;

[0008] (3) Use the improved beetle swarm optimization (IBSO) to search for the optimal gate width of the SDT, and based on the optimal gate width, obtain the characteristic data points representing the actual wind power output;

[0009] (4) Use the four-point method to process the characteristic data points extracted by the optimized rotation gate algorithm to eliminate the "bump" problem in the characteristic data points;

[0010] (5) Regard two adjacent characteristic data points after being processed by the four-point method as a wind power characteristic period, then classify and encode it, and then merge the wind power characteristic periods based on the encoding results;

[0011] (6) Detect whether a ramping event occurs in the merged wind power characteristic periods based on the ramping event judgment criterion.

[0012] In the above step (1), the judgment criterion for the ramping event of the energy storage type wind farm is as follows:

[0013]

[0014] In the formula, P fend and P fstart respectively represent the wind power at the end time and the start time of the merged wind power characteristic period, t fend and t fstart respectively represent the end time and the start time of the merged wind power characteristic period, P th is the maximum charge and discharge power of the energy storage system, and Δt is the duration of the energy storage system operating at the maximum charge and discharge power;

[0015] Furthermore, the determination of the ramping direction is as follows:

[0016]

[0017] In the above step (2), the optimization process of the improved beetle swarm optimization algorithm is as follows:

[0018] 1) Initialize the position, velocity and search direction of the beetles:

[0019]

[0020] In the formula, x s k and v s k respectively represent the s-th variable of the position and search velocity of the k-th beetle, S and K are the search space dimension and the number of beetle individuals respectively, u ps 、lps , u vs and l vs are the upper and lower bounds of the s-th variable position and the search speed, respectively;

[0021] 2) Calculate the inertia weight coefficient

[0022]

[0023] In the formula, ω max and ω min represent the maximum and minimum values of the inertia coefficient, respectively, n is the current iteration number, and N is the total number of iterations;

[0024] 3) Calculate the search step size and search distance

[0025]

[0026] In the formula, μ n k is the step size of the n-th iteration, d n k is the search distance of the n-th iteration, and c1, c2, c3, and c4 are all adjustment factors;

[0027] 4) Calculate the left and right antenna positions of the longhorn beetle swarm

[0028]

[0029] In the formula, and are the left and right antenna positions of the k-th longhorn beetle at the n-th iteration.

[0030] 5) Calculate the movement position increment

[0031]

[0032] In the formula, ξ s,n k is the movement position increment of the n-th iteration, sign is the sign function, f is the fitness function, and its calculation formula needs to be set according to the application scenario and will be specially designed later;

[0033] 6) Update the speed

[0034]

[0035] In the formula, β1 and β2 are two positive constant coefficients, r1 and r2 are two random numbers with a value range of [0, 1], I s,n k and G s,n k represent the individual extreme value and the global extreme value, respectively;

[0036] 7) Updating the position of longhorn beetles

[0037]

[0038] Where α is a positive constant.

[0039] 8) Updating the global optimal solution

[0040] Calculate the fitness function of the position of each longhorn beetle, and take the position of the longhorn beetle corresponding to the best fitness function as the global optimal solution for the nth iteration.

[0041] In the step (3), the improved longhorn beetle swarm algorithm is used to search for the optimal gate width E of the SDT algorithm, and the characteristic data points of the wind power are extracted by SDT according to the optimal gate width:

[0042] Calculation steps of the SDT algorithm:

[0043] a) Initialization

[0044]

[0045] Where t0 and x0 are the initial time and the corresponding data value respectively; t1 and x1 are the first time and the corresponding data value respectively; k 1d and k 2d are the initial values of the upper and lower fulcrum gate slopes respectively; E is the compression offset;

[0046] b) Calculate the slope

[0047]

[0048] Where: t j and x j are the jth time and the corresponding data value respectively; t k and x k are the kth time and the corresponding data value respectively;

[0049] c) Slope update

[0050]

[0051] d) Data extraction

[0052] k 2d ≤ k ≤ k 1d (13)

[0053] If the formula (13) is not satisfied, then take the data value x j-1 of the previous time t j-1 as the characteristic data for recording, and return to step b), otherwise return to step c);

[0054] The fitness function for searching the optimal gate width E of the IBSO-SDT algorithm is as follows:

[0055]

[0056] Among them, f E and f C represent the standard deviation and compression ratio of the feature trend respectively. N1 is the number of wind power data samples, P w (t) is the wind power sampling data value at time t, P f (t) is the data at time t after linear interpolation between two adjacent feature data points. N2 is the number of extracted feature data points, and λ1 and λ2 are weights, with their values being 1.25 and 5 respectively.

[0057] In step (4), the four-point method is used to process the feature data points extracted by IBSO-SDT to eliminate the "bulge", and the principle is as follows:

[0058] If the following formula (15) or (16) is satisfied, then the middle 2 points are removed to eliminate the "bulge";

[0059]

[0060] In the formula: x i , x i+1 , x i+2 , x i+3 represent 4 consecutive adjacent feature data points extracted by IBSO-SDT.

[0061] In step (5), the two adjacent feature data points after being processed by the four-point method are regarded as a wind power feature period, and then it is classified and encoded, and then the wind power feature periods are merged based on the encoding results; the classification, encoding, and merging principles of the wind power feature periods are as follows:

[0062] Classification: The line between two adjacent wind power feature data points after being processed by the four-point method is regarded as a wind power feature period; according to formula (1) and the following formula (17), the wind power feature periods are divided into obvious ramp periods, hidden ramp periods, and non-ramp periods; among them, the obvious ramp period refers to the satisfaction of formula (1), the hidden ramp period refers to the non-satisfaction of formula (1) but the satisfaction of formula (17), and the rest of the periods are non-ramp periods;

[0063]

[0064] In the formula, P f (i) and P f (i - 1) represent two adjacent feature data points after being processed by the four-point method;

[0065] Coding: According to the classification result and the ramp direction determined by Equation (2), the explicit ramp period is coded as ±1, the implicit ramp period is coded as ±2, and the non-ramp period is coded as 0, where + indicates an upward ramp and - indicates a downward ramp;

[0066] Merging: Based on the coding result, the wind power characteristic periods are merged. The merging principle is that two adjacent implicit ramp periods with the same ramp direction can be merged, two adjacent non-ramp periods can be merged, and an adjacent explicit ramp period and implicit ramp period with the same ramp direction can be merged.

[0067] In step (6) above, based on the ramp event judgment criterion, it is detected whether a ramp event occurs in the merged wind power characteristic period. The ramp event judgment criterion is shown in Equation (1).

[0068] The beneficial effects of the technical solution provided by the present invention are as follows:

[0069] The judgment criterion applicable to the ramp event of the energy storage type wind farm is given; by improving the beetle antennae search algorithm, its optimization speed can be accelerated; using the optimal gate width of IBSO-SDT to process the wind power data, the characteristic data points that can effectively represent the actual output of the wind power can be obtained; using the four-point method can effectively solve the "protrusion" existing in the characteristic data points; classifying, coding, and merging the wind power characteristic periods can effectively reduce the false negative rate of WPRE and improve the detection effect of WPRE. Description of the Drawings

[0070] The present invention will be further described below with reference to the drawings:

[0071] Figure 1 is the flow chart of the present invention;

[0072] Figure 2 is the optimization process of the improved beetle antennae search and the ordinary beetle antennae search for the optimal gate width;

[0073] Figure 3 is the wind power characteristic data points extracted by IBSO-SDT;

[0074] Figure 4 is the detection result of the wind power ramp event;

[0075] Figure 5 is the comparison of the detection effects of the ramp events under different detection methods. Specific Embodiment

[0076] In order to better understand the purpose, technical solution and technical effect of the present invention, the present invention will be further explained below with reference to the drawings.

[0077] The present invention proposes a method for detecting ramp events in an energy storage type wind farm by optimizing the rotating door algorithm, attached Figure 1This is the flowchart of the present invention, and its implementation process includes the following detailed steps.

[0078] Step 1: Design the judgment criteria for the ramp event of the energy storage type wind farm:

[0079] The judgment criteria for the ramp event of the energy storage type wind farm are as follows:

[0080]

[0081] In the formula, P fend and P fstart respectively represent the wind power at the end time and the start time of the merged wind power characteristic period, t fend and t fstart respectively represent the end time and the start time of the merged wind power characteristic period, P th is the maximum charge-discharge power of the energy storage system, and Δt is the duration of the energy storage system operating at the maximum charge-discharge power;

[0082] Further, the determination of the ramp direction is as follows:

[0083]

[0084] Step 2: Improve the beetle swarm optimization algorithm to improve its optimization speed:

[0085] The detailed optimization process of the improved beetle swarm optimization algorithm is as follows:

[0086] 1) Initialize the position, velocity and search direction of the beetles:

[0087]

[0088] In the formula, x s k and v s k respectively represent the s-th variable of the position and the search velocity of the k-th beetle, S and K are the dimensions of the search space and the number of beetle individuals respectively, u ps , l ps , u vs and l vs are respectively the upper and lower boundaries of the s-th variable position and the search velocity;

[0089] 2) Calculate the inertia weight coefficient

[0090]

[0091] In the formula, ω max and ω min respectively represent the maximum and minimum values of the inertia coefficient, n is the current iteration number, and N is the total iteration number;

[0092] 3) Calculate the search step size and search distance

[0093]

[0094] where μ n k is the step size of the nth iteration, and d n k is the search distance of the nth iteration. c1, c2, c3, and c4 are all adjustment factors;

[0095] 4) Calculate the positions of the left and right antennae of the longhorn beetle swarm

[0096]

[0097] where and are the positions of the left and right antennae of the kth longhorn beetle in the nth iteration;

[0098] 5) Calculate the movement position increment

[0099]

[0100] where ξ s,n k is the movement position increment of the nth iteration, sign is the sign function, and f is the fitness function. Its calculation formula needs to be set according to the application scenario and will be specially designed later;

[0101] 6) Update the velocity

[0102]

[0103] where β1 and β2 are two positive constant coefficients, r1 and r2 are two random numbers with a value range of [0, 1], I s,n k and G s,n k represent the individual extreme value and the global extreme value respectively.

[0104] 7) Update the position of the longhorn beetle

[0105]

[0106] where α is a positive constant.

[0107] 8) Update the global optimal solution

[0108] Calculate the fitness function of the position of each longhorn beetle, and take the position of the longhorn beetle corresponding to the best fitness function as the global optimal solution of the nth iteration.

[0109] Step 3 uses the improved beetle swarm algorithm to search for the optimal gate width E of the SDT algorithm, and extracts the characteristic data points of wind power using SDT based on the optimal gate width:

[0110] Calculation steps of the SDT algorithm:

[0111] a) Initialization

[0112]

[0113] In the formula, t0 and x0 are the initial time and the corresponding data value respectively; t1 and x1 are the first time and the corresponding data value respectively; k 1d and k 2d are the initial values of the upper and lower fulcrum gate slopes respectively; E is the compression offset;

[0114] b) Calculate the slope

[0115]

[0116] In the formula: t j and x j are the jth time and the corresponding data value respectively; t k and x k are the kth time and the corresponding data value respectively;

[0117] c) Slope update

[0118]

[0119] d) Data extraction

[0120] k 2d ≤k≤k 1d (32)

[0121] If the formula (32) is not satisfied, then the data value x j-1 at the previous time t j-1 is recorded as the characteristic data, and step b) is returned; otherwise, step c) is returned;

[0122] The fitness function of the IBSO to search for the optimal gate width E of the SDT algorithm is as follows:

[0123]

[0124] Among them, f E and f C represent the standard deviation and compression ratio of the characteristic trend respectively, N1 is the number of wind power data samples, P w (t) is the wind power sampling data value at time t, P f(t) is the data at time t after linear interpolation between two adjacent characteristic data points, N2 is the number of extracted characteristic data points, and λ1 and λ2 are weights, with values of 1.25 and 5 respectively;

[0125] Step 4 uses the four-point method to process the characteristic data points extracted by IBSO-SDT to eliminate "bulges":

[0126] If the following formula (15) or (16) is satisfied, then the middle 2 points are removed to eliminate the "bulge";

[0127]

[0128] In the formula: x i , x i+1 , x i+2 , x i+3 represent 4 consecutive adjacent characteristic data points extracted by IBSO-SDT.

[0129] Step 5 regards two adjacent characteristic data points after the four-point method as a wind power characteristic period, then classifies and encodes it, and then merges the wind power characteristic periods based on the encoding results; the classification, encoding, and merging principles of the wind power characteristic periods are as follows:

[0130] Classification: The line between two adjacent wind power characteristic data points after the four-point method is regarded as a wind power characteristic period; according to formula (18) and the following formula (36), the wind power characteristic periods are divided into obvious ramp periods, hidden ramp periods, and non-ramp periods; among them, the obvious ramp period refers to the satisfaction of formula (18), the hidden ramp period refers to the non-satisfaction of formula (18) but the satisfaction of formula (36), and the rest of the periods are non-ramp periods;

[0131]

[0132] In the formula, P f (i) and P f (i - 1) represent two adjacent characteristic data points after the four-point method;

[0133] Encoding: According to the classification result and the ramp direction determined by formula (19), the obvious ramp period is encoded as ±1, the hidden ramp period is encoded as ±2, and the non-ramp period is encoded as 0, where + represents the up ramp and - represents the down ramp;

[0134] Merging: Based on the encoding results, the wind power characteristic periods are merged. The merging principle is that two adjacent hidden ramp periods with the same ramp direction can be merged, two adjacent non-ramp periods can be merged, and two adjacent obvious ramp periods and hidden ramp periods with the same ramp direction can be merged, as shown in Table 1 below.

[0135] Table 1 Principles for Combining Wind Power Characteristic Periods

[0136]

[0137]

[0138] Step 6 Detect whether a ramping event occurs in the combined wind power characteristic periods based on the ramping event judgment criteria.

[0139] To further understand the present invention and verify the effectiveness of the proposed wind power ramping event detection method, the actual wind power of a wind farm is used for simulation. The installed capacity of this wind farm is 97.5 MW, and it is equipped with a battery energy storage system with a scale of 10 MW / 10 MWh. Therefore, P th and Δt in Equation (18) are taken as 10 MW and 1 h respectively. The actual wind power output of a typical day is selected for research, and its sampling time is 1 s and the duration is 24 h.

[0140] Firstly, the IBSO is used to search for the optimal gate width of the SDT; to reflect the superiority of the IBSO algorithm, the search process of using the BSO algorithm to search for the optimal gate width is compared. The optimization processes of IBSO and BSO are as shown in the appendix Figure 2 As can be seen, the optimization speed of IBSO is faster. When iterating to 16 times, the IBSO algorithm finds the optimal solution. At this time, the gate width is 0.7154, the corresponding fitness function value is 0.5566, the compression error is 0.2779, and the compression ratio is 0.0418; the wind power characteristic data points extracted by IBSO-SDT are as shown in the appendix Figure 3 As shown.

[0141] To further reflect the effectiveness of IBSO-SDT in extracting wind power characteristic data points, the extraction errors of wind power characteristic data points between it and the ordinary SDT algorithm and the Convolutional Neural Network (CNN) method are compared. The results are shown in Table 2. It can be seen that compared with SDT and CNN, the fitness function value of the characteristic data points extracted by the proposed IBSO-SDT algorithm is the best and the compression error is the smallest.

[0142] Table 2 Wind Power Characteristic Data Points Extracted by Different Methods

[0143]

[0144] The detection results of the ramping events of the typical day by the detection method proposed in the present invention are as shown in the appendix Figure 4 As can be seen, there are more ramping events from 16:00 to 24:00 on this typical day, and the ramping event from 22:00 to 23:00 is relatively serious.

[0145] To further verify the effectiveness of the detection results, they were compared with the SDT algorithm and the sliding window method (SW), and the WPRE detection evaluation index was used to evaluate different detection methods.

[0146] WPRE detection result evaluation index:

[0147] This evaluation index system mainly includes 4 indexes, namely accuracy (ACC), recall (POD), success rate (SR), and critical success index (CSI). Among them, accuracy refers to the proportion of correct judgments on whether the wind power characteristic period belongs to a ramp event or a non-ramp event; recall refers to the proportion of correctly detected wind power ramp events to the actual wind power ramp events; success rate refers to the proportion of actually occurred events among the detected wind power ramp events; the critical success index represents the proportion of correctly detected wind power ramp events. The calculation formulas of the four are as follows:

[0148]

[0149] In the formula, TP refers to the actually occurred ramp event being detected, indicating accurate detection; TN refers to the actually not occurred ramp event not being detected, indicating accurate detection; FN refers to the actually occurred ramp event not being detected, indicating a missed report; FP refers to the actually not occurred ramp event being detected, indicating a false alarm.

[0150] The detection result evaluation of WPRE under different methods is shown in the appendix Figure 5 As shown, it can be seen that in the 4 evaluation indexes, the detection effect of the detection method in this paper is the best, especially ACC and SR are both above 0.83, which indicates that the detection method proposed in the present invention can indeed effectively detect the WPRE of the energy storage type wind farm.

Claims

1. An energy storage type wind farm ramp event detection method for optimizing the swing door algorithm, characterized in that, It includes the following steps: (1) Design the judgment criteria for the ramping events of the energy storage type wind farm as follows: Wherein, P fend and P fstart respectively represent the wind power at the end time and the start time of the combined wind power characteristic period, t fend and t fstart respectively represent the end time and the start time of the combined wind power characteristic period, Pth is the maximum charge and discharge power of the energy storage system, and Δt is the duration of the energy storage system operating at the maximum charge and discharge power; Further determine the ramping direction as follows: (2) Improve the beetle antennae search algorithm to enhance its optimization speed. The step size of the improved beetle antennae algorithm is obtained through the following formula: Where, n is the number of iterations, k represents the k-th beetle, and c1, c2, and c3 are all step size parameters, and their values are 1.5, 0.5, and 8 respectively; (3) Use the improved beetle antennae search algorithm to search for the optimal gate width of the swing gate algorithm, and based on the optimal gate width, obtain the characteristic data points representing the actual wind power output; (4) Use the four-point method to process the characteristic data points extracted by the improved beetle antennae optimization swing gate algorithm to eliminate the "bulges"; The four-point method to remove the "bulges": If the following formula (5) or (6) is satisfied, then the middle two points are removed to eliminate the "bulges"; Where: x i , x i+1 , x i+2 , x i+3 represent 4 consecutive adjacent characteristic data points extracted by IBSO-SDT; (5) Regard two adjacent characteristic data points after being processed by the four-point method as a wind power characteristic time period, then classify and encode it, and then merge the wind power characteristic time periods based on the encoding results; Classification: Regard the connection line between two adjacent wind power characteristic data points after being processed by the four-point method as a wind power characteristic time period; According to formula (1) and formula (7), divide the wind power characteristic time period into an explicit ramping time period, an implicit ramping time period, and a non-ramping time period; The explicit ramping time period refers to the satisfaction of formula (1), the implicit ramping time period refers to the non-satisfaction of formula (1) but the satisfaction of formula (7), and the rest of the time periods are non-ramping time periods; Wherein, P f (i) and P f (i - 1) represent two adjacent characteristic data points after four-point method processing; Encoding: According to the classification results and the ramping direction determined by formula (2), encode the explicit ramping time period as ±1, the implicit ramping time period as ±2, and the non-ramping time period as 0, where, + represents up-ramping, and - represents down-ramping; Merging: Based on the encoding results, merge the wind power characteristic time periods, and the merging principle is shown in the following table; Table 1 Merging principle of wind power characteristic time periods (6) Detect whether a ramping event occurs in the combined wind power characteristic time period based on the ramping event judgment criteria.

2. The energy storage type wind farm ramp event detection method for optimizing the swing door algorithm according to claim 1, characterized in that, In the step (3), use the improved beetle antennae algorithm to search for the optimal gate width of the swing gate algorithm, and extract the characteristic data points representing the actual wind power output by using the swing gate algorithm according to the optimal gate width; The fitness function of the improved beetle antennae search swing gate algorithm is: Among them, f E and f C respectively represent the standard deviation and compression ratio of the feature trend. N1 is the number of wind power data samples, and P w (t) is the wind power sampling data value at time t, P f (t) is the data at time t after linear interpolation between two adjacent feature data points. N2 is the number of extracted feature data points, and λ1 and λ2 are weights, with their values being 1.25 and 5 respectively.

3. The energy storage type wind farm ramp event detection method for optimizing the swing door algorithm according to claim 1, characterized in that, In the step (6), use the ramping event judgment criteria of the energy storage type wind farm to detect whether a ramping event occurs in the merged wind power characteristic time period.

Images