New Wind Power Data Cleaning Method and System Combining Dynamic Programming and Dichotomy

Abstract

The present invention discloses a novel wind power data cleaning method and system combining dynamic programming and dichotomy. The method includes obtaining a wind power data sample; preprocessing the wind power data sample to obtain a corresponding serial number list for each day; establishing a state list corresponding to the serial number list, where each subsequence records the longest decreasing serial number list segment reaching the corresponding index position in the serial number list; establishing a sorted list corresponding to the state list, and each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence in the state list; constructing a state transition equation through dynamic programming and using dichotomy to solve the length of the longest decreasing subsequence in each sorted list; deleting the abnormal wind power data points corresponding to the range between the shortest decreasing subsequence length and the longest decreasing subsequence length according to a preset shortest decreasing subsequence length; and the system is adapted to be used with the above method. The present invention can quickly and effectively clean wind power abnormal data in special cases.

Description

Technical Field

[0001] The present invention relates to a new method and system for cleaning wind power data by combining dynamic programming and dichotomy, belonging to the technical field of wind power data cleaning. Background Art

[0002] At present, there are five main methods for cleaning abnormal wind power data at home and abroad:

[0003] One is to calculate the Z-Score score of each sample and compare it with the threshold θ to determine whether it is an abnormal point or a normal point. When the data volume is large or there are many abnormal points, the time complexity of this algorithm is relatively high and it is difficult to determine the value of the threshold θ;

[0004] The second is to use the LOF algorithm to eliminate outliers according to the density difference between clusters, which can effectively eliminate discrete outliers, but the recognition effect of this algorithm on stacked outliers is poor;

[0005] The third is to use the optimal variance within the group. This algorithm can effectively eliminate the abnormal data below the wind power data curve, but the elimination effect on the abnormal data above is not good;

[0006] The fourth is to use quartiles as a method for identifying abnormal data. However, when there are many abnormal data, the recognition effect of quartiles is poor and using quartiles will cause misdeletion of data;

[0007] The fifth is to use statistical learning models, machine learning models or deep learning methods to capture features such as linear, non-linear, short-term or long-term patterns in the sequence for prediction, and then compare the predicted value with the true value to delete abnormal sequences.

[0008] However, these models require a large amount of data for training. It is difficult to capture the complex non-linear relationships, periodic relationships, etc. of wind power data. Moreover, if there are many abnormal sequences in the wind power data itself, it will lead to a worse model effect, resulting in a worse cleaning effect.

[0009] In summary, the existing wind power data cleaning technologies cannot quickly and effectively clean the abnormal sequences in wind power data. In order to solve the above problems, the present application proposes a new method and system for cleaning wind power data by combining dynamic programming and dichotomy. Summary of the Invention

[0010] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a new method and system for cleaning wind power data by combining dynamic programming and dichotomy. The present invention can quickly and effectively clean abnormal wind power data in special cases.

[0011] To achieve the above object, the present invention is implemented by the following technical solutions:

[0012] On the one hand, the present invention provides a novel wind power data cleaning method combining dynamic programming and dichotomy, comprising the following steps:

[0013] Obtain wind power data samples, where the wind power data includes historical measured wind speed data and measured power data;

[0014] Preprocess the wind power data samples to obtain a serial number list corresponding to each day, and the serial number list records the serial numbers representing the changing trend relationship between the wind speed and power at each moment;

[0015] Establish a state list corresponding to the serial number list, where the state list includes multiple subsequences, and each subsequence records the longest decreasing serial number list segment in the serial number list reaching the corresponding index position;

[0016] Establish a sorted list corresponding to the state list, and each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence of the state list;

[0017] Construct a state transition equation through dynamic programming, and use dichotomy to solve the length of the longest decreasing subsequence in each sorted list;

[0018] Delete the abnormal points of the wind power data corresponding to the range between the shortest decreasing subsequence length and the longest decreasing subsequence length according to the preset shortest decreasing subsequence length.

[0019] Further, the preprocessing of the wind power data samples to obtain a serial number list corresponding to each day, where the serial number list records the changing trend relationship between the wind speed and power at each moment includes the following steps:

[0020] Divide the wind power data samples by day, and reorganize the wind power data of each day to obtain a reorganized sample;

[0021] Initialize the reorganized samples of each day one by one, and calibrate the serial number of the initial moment of each day as 0;

[0022] Calculate the product of the measured wind speed difference and the measured power difference between the t-th moment and the (t - 1)-th moment of the initialized reorganized sample one by one: if the product of the measured wind speed difference and the measured power difference is greater than 0, the serial number at the t-th moment is t - 1 plus 1; if the product of the measured wind speed difference and the measured power difference is less than 0, the serial number at the t-th moment is t - 1 minus 1; if the product of the measured wind speed difference and the measured power difference is equal to 0, the serial number at the t-th moment is t - 1;

[0023] Establish a serial number list corresponding to the reorganized samples of each day one by one, and each element in the serial number list is the serial number at the corresponding moment.

[0024] Further, the establishment of a state list corresponding to the serial number list, where the state list includes multiple subsequences, and each subsequence records the longest decreasing serial number list segment in the serial number list reaching the corresponding index position includes equation (1), specifically as follows:

[0025] dp[i] = max(dp[i], dp[j] + 1) (1)

[0026] Where dp is the state list, dp[i] is the subsequence corresponding to index i of the state list, recording the longest decreasing sequence list segment in the index of the sequence number list from 0 to i. The value of dp[i] is the length of the longest decreasing sequence list in the index of the sequence number list from 0 to i. dp[j] is the subsequence corresponding to index j of the state list, recording the decreasing sequence list segment in the index of the sequence number list from 0 to j. The value of dp[j] is the length of the decreasing sequence list in the index of the sequence number list from 0 to j. j ∈ [0, i), i = len(A), and A is the sequence number list.

[0027] Furthermore, the state transition equation includes the following formula:

[0028] dichotomy[b] = max(dichotomy[b], A[b]) for b in [0, i)

[0029] Where b = len(dichotomy), dichotomy is the sorted list, and the sorted list is dichotomy[0...len(A)].

[0030] Furthermore, using the binary search method to solve the length of the longest decreasing subsequence in each sorted list includes the following steps: calculating each dichotomy[b] using the state transition equation and updating the value of the tail element of the subsequence with a length of [1, b] according to the preset rule to obtain the length len(dichotomy) of the global longest decreasing subsequence.

[0031] Furthermore, the updating of the value of the tail element of the subsequence with a length of [1, b] according to the preset rule includes:

[0032] If there exists dichotomy[b] < A[k] in the interval [0, len(dichotomy)), then replace the value of the first dichotomy[b] that satisfies this condition with A[k];

[0033] If there does not exist dichotomy[b] < A[k] in the interval [0, len(dichotomy)), then arrange A[k] after all subsequences, that is, the length of the sorted list will increase by 1;

[0034] Where K ∈ [0, len(A)).

[0035] On the other hand, the present invention provides a novel wind power data cleaning system combining dynamic programming and binary search method, including the following modules:

[0036] A sample acquisition module for acquiring wind power data samples, where the wind power data includes historical measured wind speed data and measured power data;

[0037] A sample preprocessing module for preprocessing the wind power data samples to obtain a serial number list for each day, where the serial number list records the serial numbers representing the changing trend relationship between the wind speed and power at each moment;

[0038] A status list module for establishing a status list corresponding to the serial number list, where the status list includes multiple subsequences, and each subsequence records the longest decreasing serial number list segment reaching the corresponding index position in the serial number list;

[0039] A sorted list module for establishing a sorted list corresponding to the status list, where each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence of the status list;

[0040] A solution module for constructing a state transition equation through dynamic programming and using the binary search method to solve the length of the longest decreasing subsequence in each sorted list;

[0041] An abnormal data deletion module for deleting abnormal wind power data points corresponding to the range between the shortest decreasing subsequence length and the longest decreasing subsequence length according to a preset shortest decreasing subsequence length.

[0042] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0043] The present invention constructs a serial number list of the product of the difference between the wind speed and power by using the first-order difference operation, sequentially constructs the corresponding status list and sorted list, constructs a state transition equation using dynamic programming, and uses the binary search method to traverse the sorted list of the state transition equation to solve the length of the longest decreasing subsequence, so as to delete the corresponding abnormal wind power data points; the present invention optimizes the time complexity in the solution process, reducing the overall time complexity from O(N2) to O(NlogN), thereby achieving fast deletion of abnormal wind power data with opposite wind speed and measured power conditions. Description of the Drawings

[0044] Figure 1 Shown is a flowchart of an embodiment of a novel wind power data cleaning method combining dynamic programming and binary search according to the present invention;

[0045] Figure 2 Shown is a curve comparison diagram of measured wind speed - measured power - abnormal measured power in the time period from 20:00:00 on July 24, 2021 to 22:30:00 on July 24, 2021 according to the present invention;

[0046] Figure 3The figure shows a curve comparison diagram of the measured wind speed - measured power - abnormal measured power of the present invention during the time period from 11:10:00 on September 21, 2021 to 16:10:00 on September 21, 2021. Detailed implementation mode

[0047] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention and cannot be used to limit the protection scope of the present invention.

[0048] Embodiment 1

[0049] This embodiment provides a new wind power data cleaning method combining dynamic programming and dichotomy. Refer to Figure 1 , including the following steps:

[0050] Obtain wind power data samples, where the wind power data includes historical measured wind speed data and measured power data;

[0051] Preprocess the wind power data samples to obtain a serial number list for each day. The serial number list records the serial numbers representing the change trend relationship between the wind speed and power at each moment;

[0052] Establish a state list corresponding to the serial number list. The state list includes multiple subsequences, and each subsequence records the longest decreasing serial number list segment in the serial number list reaching the corresponding index position;

[0053] Establish a sorted list corresponding to the state list. Each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence of the state list;

[0054] Construct a state transition equation through dynamic programming, and use dichotomy to solve the length of the longest decreasing subsequence in each sorted list;

[0055] According to the preset shortest decreasing subsequence length, delete the abnormal points of the wind power data corresponding to the range between the shortest decreasing subsequence length and the longest decreasing subsequence length.

[0056] The present invention constructs a serial number list of the product of the wind speed and power difference through the first-order difference operation, constructs the corresponding state list and sorted list in sequence, constructs a state transition equation using dynamic programming, and uses dichotomy to traverse the sorted list to solve the length of the longest decreasing subsequence to delete the corresponding abnormal points of the wind power data; the present invention optimizes the time complexity in the solution process, reducing the entire time complexity from O(N 2 ) to O(NlogN), so as to quickly delete the abnormal wind power data with the opposite situation between the wind speed and the measured power, which has practical engineering value.

[0057] Embodiment 2

[0058] Based on Embodiment 1, the present invention specifically introduces a method for preprocessing wind power data samples, a state list, a state transition equation, and a method for solving the length of the longest decreasing subsequence in each sorted list using the bisection method.

[0059] (I) Preprocessing wind power data samples

[0060] Preprocess the wind power data samples to obtain a serial number list corresponding to each day. The serial number list records the relationship between the wind speed and power change trends at each moment, including the following steps:

[0061] Divide the wind power data samples by day and reorganize the wind power data of each day to obtain a reorganized sample;

[0062] Initialize the reorganized sample of each day one by one, and mark the serial number of the initial moment of each day as 0;

[0063] Calculate the product of the measured wind speed difference and the measured power difference between the t-th moment and the (t - 1)-th moment of the initialized reorganized sample one by one: if the product of the measured wind speed difference and the measured power difference is greater than 0, the serial number at the t-th moment is t - 1 plus 1; if the product of the measured wind speed difference and the measured power difference is less than 0, the serial number at the t-th moment is t - 1 minus 1; if the product of the measured wind speed difference and the measured power difference is equal to 0, the serial number at the t-th moment is t - 1;

[0064] Establish a serial number list corresponding to the reorganized sample of each day. Each element in the serial number list is the serial number corresponding to the corresponding moment.

[0065] (II) State list

[0066] Establish a state list dp corresponding to the serial number list. The state list includes multiple subsequences. Each subsequence records the longest decreasing serial number list segment reaching the corresponding index position in the serial number list, including Equation (1), specifically as follows:

[0067] dp[i] = max(dp[i], dp[j] + 1) (1)

[0068] Where dp[i] is the subsequence corresponding to the index i of the state list, recording the longest decreasing serial number list segment in the serial number list from index 0 to i. The value of dp[i] is the length of the longest decreasing serial number list in the serial number list from index 0 to i. dp[j] is the subsequence corresponding to the index j of the state list, recording the decreasing serial number list segment in the serial number list from index 0 to j. The value of dp[j] is the length of the decreasing serial number list in the serial number list from index 0 to j. j ∈ [0, i), i = len(A), and A is the serial number list.

[0069] (III) State transition equation

[0070] The state transition equation includes the following formula:

[0071] dichotomy[b] = max(dichotomy[b], A[b]) for b in [0, i)

[0072] where b = len(dichotomy), dichotomy is a sorted list, and the sorted list is dichotomy[0...len(A)].

[0073] (4) Using the dichotomy method to solve the length of the longest decreasing subsequence in each sorted list

[0074] Using the dichotomy method to solve the length of the longest decreasing subsequence in each sorted list includes the following steps:

[0075] Calculate each dichotomy[b] using the state transition equation, and update the value of the tail element of the subsequence with length [1, b] according to the preset rules to obtain the length len(dichotomy) of the global longest decreasing subsequence.

[0076] When applying this embodiment, updating the value of the tail element of the subsequence with length [1, b] according to the preset rules includes:

[0077] If there exists dichotomy[b] < A[k] in the interval [0, len(dichotomy)), then replace the value of the first dichotomy[b] that satisfies this condition with A[k], because a larger A[k] is more likely to be followed by a number smaller than it;

[0078] If there does not exist dichotomy[b] < A[k] in the interval [0, len(dichotomy)), then arrange A[k] after all subsequences, that is, the length of the sorted list will increase by 1;

[0079] where K ∈ [0, len(A)).

[0080] Embodiment 3

[0081] Based on Embodiment 1 or 2, this embodiment provides a wind power data cleaning method for a certain time segment by combining dynamic programming and dichotomy, including the following steps:

[0082] S11 Collect the historical measured wind speed data and measured power data of each flat ground wind farm for a certain time segment.

[0083] where the measured wind speed list is WS: [2, 5, 4, 7, 10, 9, 6, 3, 5, 6, 1, 7, 3, 8, 5, 2], with the unit of m / s;

[0084] The corresponding measured power list is P: [16, 11, 15, 10, 9, 50, 55, 57, 30, 29, 5, 60, 30, 20, 22, 25], with the unit of MW;

[0085] S12 performs a first-order difference operation on the WS list and the P list respectively, and the WS_1 list and the P_1 list can be obtained as follows:

[0086] WS_1: [3, -1, 3, 3, -1, -3, -3, 2, 1, -5, 6, -4, 5, -3, -3], with the unit of m / s;

[0087] P_1: [-5, 4, -5, -1, 41, 5, 2, -27, -1, -24, 55, -30, -10, 2, 3], with the unit of MW.

[0088] S13 multiplies the WS_1 list and the P_1 list to obtain the WSP list as follows:

[0089] WSP: [-15, -4, -15, -3, -41, -15, -6, -54, -1, 120, 330, 120, -50, -6, -9].

[0090] According to the WSP list in S14, if the product after differentiation at time t is negative, it means that at time t, the wind speed and the power trend are opposite. According to the wind power formula:

[0091]

[0092] Among them, P represents wind power, ρ represents air density, γ represents swept area, and v represents wind speed;

[0093] Then the power is proportional to the wind speed. When the wind speed and the power trend are opposite, it belongs to an abnormal point with an opposite trend.

[0094] According to the WSP list, judge the relationship between the product of the first-order difference of the WS list and the P list and 0, and create the sequence number list A. The sequence number list A can be obtained: [-1, -2, -3, -4, -5, -6, -7, -8, -9, -8, -7, -6, -7, -8, -9]; Combining the WSP list and the sequence number list A, it can be seen that the decreasing segments in the sequence number list A are the segments where the wind speed and the power trend are opposite, that is, the abnormal sequences.

[0095] S21 solves the state list dp corresponding to the sequence number list A.

[0096] Each state list includes: the value of index j in the state list represents the length of the decreasing sequence of the sequence number list before reaching the current sequence number list index j;

[0097] The state list dp is dp[1…len(A)]. Initialize all elements of dp[1…len(A)] to 1, which means that each element can at least form a subsequence alone, and at this time the length is 1;

[0098] In each round of calculating the new dp[i], when j ∈ [0, i), i = len(A), when traversing the list interval [0, i), the following judgment is made:

[0099] When A[i] < A[j], that is, A[j], A[i] is strictly decreasing, denoted as a strictly decreasing subsequence;

[0100] When A[i] ≥ A[j], that is, A[j], A[i] is non-decreasing, in this case the decreasing subsequence does not hold, skip;

[0101] In summary, the state list dp can be obtained: [1, 2, 3, 4, 5, 6, 7, 8, 9, 8, 7, 6, 7, 8, 9]

[0102] Among them, dp[0] represents the length of the longest decreasing sequence in the subsequence composed of A[0] in the sequence list A. Since A[0] = -1, and the length of the longest decreasing sequence in the subsequence composed of A[0] is the length of A[0] itself, that is, dp[0] = 1;

[0103] dp[6] represents the length of the longest decreasing sequence in the subsequence composed of A[0]-A[6] in the sequence list A. Since A[6] = -7, and the length of the longest decreasing sequence in the subsequence composed of A[0]-A[6] is the length of this subsequence from A[0] to A[6], that is, dp[6] = 7;

[0104] dp

[13] represents the length of the longest decreasing sequence in the subsequence composed of A[0]-A

[13] in the sequence list A. Since A

[13] = -8 < A[6] = -7, and A[7], A[8], A[9] are all less than or equal to A

[13] , therefore, from dp[i] = max(dp[i], dp[j] + 1), j belongs to [0, i), it can be known that dp

[13] = dp[6] + 1, that is, dp

[13] = 8.

[0105] S31 Solve the sorted list dichotomy corresponding to the state list dp

[0106] Set the sorted list dichotomy[0…len(A)], where the value of dichotomy[b] represents the tail element value of the subsequence with length b + 1 in the state list dp, that is, represents the value of the last element in the decreasing subsequence of this time segment;

[0107] S41 Construct the state transition equation

[0108] The state transition equation includes: dichotomy[b] = max(dichotomy[b], A[b]) for b in [0, i)

[0109] where b = len(dichotomy).

[0110] S42 uses the dichotomy method to solve the length of the longest decreasing subsequence in each sorted list

[0111] When traversing and calculating each dichotomy[j] in the state transition equation, it is necessary to continuously update the tail element value of the subsequence with length [1, j], always keeping each tail element value the largest. In this way, it is ensured that the dichotomy list is strictly decreasing. Therefore, each time the sequence number list A is traversed, when calculating each dichotomy[b] using the dichotomy method with the state transition equation, the sorted list dichotomy can be updated, reducing the original algorithm time complexity O(N 2 ) to O(NlogN). The update steps are as follows:

[0112] If there exists dichotomy[b] < A[k] in the interval [0, len(dichotomy)), then replace the value of the first dichotomy[b] that satisfies this condition with A[k], because a larger A[k] is more likely to be followed by a number smaller than it;

[0113] If there does not exist dichotomy[b] < A[k] in the interval [0, len(dichotomy)), then arrange A[k] after all subsequences, that is, the length of the sorted list will increase by 1;

[0114] where K ∈ [0, len(A)).

[0115] Then the length of the longest decreasing subsequence Maxres can be obtained, that is, len(dichotomy).

[0116] In the application of this embodiment, the sorted list dichotomy is [-1, -2, -3, -4, -5, -6, -7, -8, -9],

[0117] where dichotomy[4] being -5 represents that the tail element value of the subsequence with length 5 is -5, that is, it represents that the value of the last decreasing element in the local decreasing subsequence [-1, -2, -3, -4, -5] is -5.

[0118] In the application, the time complexity of traversing the A list is O(N), and the time complexity of updating the dichotomy list using the dichotomy method each time A is traversed is O(logN). Therefore, the overall time complexity is O(NlogN).

[0119] S5 Deleting Abnormal Points in Wind Power Data

[0120] According to the preset minimum decreasing subsequence length Minres, delete the increasing sequence before the index that matches the range of the minimum decreasing subsequence length Minres and the maximum decreasing subsequence length Maxres in the state list dp.

[0121] In the application, refer to Figure 2 and 3 It can be known that in the time interval when the measured wind speed and the measured power trends are opposite in this application, the abnormal measured power sequence is effectively deleted; among them, although the trends of the measured power curve and the measured wind speed curve are the same around the time of 2021-09-21 12:25:00, it is not misdeleted, indicating the practicability and reliability of this application.

[0122] Embodiment 4

[0123] This embodiment provides a new type of wind power data cleaning system combining dynamic programming and dichotomy, including the following modules:

[0124] A sample acquisition module for acquiring wind power data samples, where the wind power data includes historical measured wind speed data and measured power data;

[0125] A sample preprocessing module for preprocessing the wind power data samples to obtain a serial number list for each day, and the serial number list records the serial numbers representing the trend relationship between the wind speed and power at each moment;

[0126] A state list module for establishing a state list corresponding to the serial number list, where the state list includes multiple subsequences, and each subsequence records the longest decreasing serial number list segment reaching the corresponding index position in the serial number list;

[0127] A sorted list module for establishing a sorted list corresponding to the state list, and each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence of the state list;

[0128] A solution module for constructing a state transition equation through dynamic programming and using dichotomy to solve the length of the longest decreasing subsequence in each sorted list;

[0129] A deletion abnormal module for deleting abnormal points of wind power data corresponding to the range between the minimum decreasing subsequence length and the maximum decreasing subsequence length according to the preset minimum decreasing subsequence length.

[0130] For the specific function implementation of the above each functional module, refer to the relevant content in the method of Embodiment 2 or 3.

[0131] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0132] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0133] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0134] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0135] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit and scope protected by the present invention and the claims. These all fall within the protection scope of the present invention.

Claims

1. A novel wind power data cleaning method combining dynamic programming and dichotomy, characterized in that Including the following steps: Obtain wind power data samples, where the wind power data includes historical measured wind speed data and measured power data; Preprocess the wind power data samples to obtain a sequence number list corresponding to each day. The sequence number list records the sequence numbers representing the change trend relationship between the wind speed and power at each moment; Establish a state list corresponding to the sequence number list. The state list includes multiple subsequences, and each subsequence records the longest decreasing sequence number list segment reaching the corresponding index position in the sequence number list; Establish a sorted list corresponding to the state list. Each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence in the state list; Construct a state transition equation through dynamic programming and use the binary search method to solve the length of the longest decreasing subsequence in each sorted list; According to the preset shortest decreasing subsequence length, delete the abnormal points of the wind power data corresponding to the range between the shortest decreasing subsequence length and the longest decreasing subsequence length; The step of preprocessing the wind power data samples to obtain a sequence number list corresponding to each day, where the sequence number list records the change trend relationship between the wind speed and power at each moment includes the following steps: Divide the wind power data samples by day and reorganize the wind power data of each day to obtain a reorganized sample; Initialize the reorganized samples of each day one by one, and mark the sequence number of the initial moment of each day as 0; Calculate the product of the measured wind speed difference and the measured power difference between the t-th moment and the (t - 1)-th moment of the initialized reorganized sample one by one: If the product of the measured wind speed difference and the measured power difference is greater than 0, the sequence number at the t-th moment is (t - 1) + 1; If the product of the measured wind speed difference and the measured power difference is less than 0, the sequence number at the t-th moment is (t - 1) - 1; If the product of the measured wind speed difference and the measured power difference is equal to 0, the sequence number at the t-th moment is (t - 1); Establish a sequence number list corresponding to the reorganized samples of each day one by one. Each element in the sequence number list is the sequence number corresponding to the corresponding moment; Establish a state list corresponding to the sequence number list. The state list includes multiple subsequences, and each subsequence records the longest decreasing sequence number list segment reaching the corresponding index position in the sequence number list, including formula (1), specifically as follows: （1） Among them, is the status list, is the subsequence corresponding to the index i of the status list, recording the longest decreasing sequence list segment in the index of the sequence number list from 0 to i, the value of is the length of the longest decreasing sequence list in the index of the sequence number list from 0 to i, is the subsequence corresponding to the index j of the status list, recording the decreasing sequence list segment in the index of the sequence number list from 0 to j, the value of is the length of the decreasing sequence list in the index of the sequence number list from 0 to j, , , A is the sequence number list , , A is the sequence number list; The state transition equation includes the following formula: ， Among them, , is an ordered list, and the ordered list is ; Using the dichotomy method to solve the length of the longest decreasing subsequence in each sorted list includes the following steps: calculating each using the state transition equation , and updating the tail element value of the subsequence with length according to the preset rules to obtain the length of the global longest decreasing subsequence ; The updating of the value of the tail element of the subsequence with length according to the preset rule includes: If there is in the interval there is , then replace the value of the first one that meets this condition with ; ; If the interval does not contain , then move to the end of all subsequences, i.e., the length of the sorted list will increase by 1; Among them, .

2. A novel wind power data cleaning system combining dynamic programming and dichotomy, characterized in that, Including the following modules: A sample acquisition module for obtaining wind power data samples, where the wind power data includes historical measured wind speed data and measured power data; A sample preprocessing module for preprocessing the wind power data samples to obtain a sequence number list corresponding to each day. The sequence number list records the sequence numbers representing the change trend relationship between the wind speed and power at each moment; A state list module for establishing a state list corresponding to the sequence number list. The state list includes multiple subsequences, and each subsequence records the longest decreasing sequence number list segment reaching the corresponding index position in the sequence number list; A sorted list module for establishing a sorted list corresponding to the state list. Each element in the sorted list records the maximum tail element value of each strictly decreasing subsequence in the state list; A solution module for constructing a state transition equation through dynamic programming and using the binary search method to solve the length of the longest decreasing subsequence in each sorted list; An abnormal point deletion module for deleting the abnormal points of the wind power data corresponding to the range between the shortest decreasing subsequence length and the longest decreasing subsequence length according to the preset shortest decreasing subsequence length; Preprocess the wind power data samples to obtain a sequence number list for each day. The sequence number list records the relationship between the wind speed and power change trends at each moment, including the following steps: Divide the wind power data samples by day and reorganize the wind power data for each day to obtain reorganized samples; Initialize the reorganized samples for each day one by one, and calibrate the initial moment sequence number of each day as 0; Calculate the product of the measured wind speed difference and the measured power difference between the t-th moment and the (t - 1)-th moment of the initialized reorganized samples one by one: if the product of the measured wind speed difference and the measured power difference is greater than 0, the sequence number at the t-th moment is (t - 1) plus 1; if the product of the measured wind speed difference and the measured power difference is less than 0, the sequence number at the t-th moment is (t - 1) minus 1; if the product of the measured wind speed difference and the measured power difference is equal to 0, the sequence number at the t-th moment is (t - 1); Establish a sequence number list corresponding to the reorganized samples for each day. Each element in the sequence number list is the sequence number at the corresponding moment; Establish a state list corresponding to the sequence number list. The state list includes multiple subsequences. Each subsequence records the longest decreasing sequence number list segment that reaches the corresponding index position in the sequence number list, including equation (1), specifically as follows: （1） Among them, is the status list, is the subsequence corresponding to the index i of the status list, recording the longest decreasing sequence list segment in the index of the sequence number list from 0 to i, The value of is the length of the longest decreasing sequence list in the index of the sequence number list from 0 to i, is the subsequence corresponding to the index j of the status list, recording the decreasing sequence list segment in the index of the sequence number list from 0 to j, The value of is the length of the decreasing sequence list in the index of the sequence number list from 0 to j, , , A is the sequence number list , , A is the sequence number list; The state transition equation includes the following formula: ， Among them, , is an ordered list, and the ordered list is ; Using the dichotomy method to solve the length of the longest decreasing subsequence in each sorted list includes the following steps: Calculate each using the state transition equation , and update the value of the tail element of the subsequence with a length of according to the preset rule to obtain the length of the global longest decreasing subsequence ; The method for updating the value of the tail element of the subsequence with a length of according to a preset rule includes: If the interval contains , then replace the value of the first that satisfies this condition with ; If the interval does not contain , then is placed after all subsequences, that is, the length of the sorted list will increase by 1; Among them, 。

Images