A time series clustering method based on extreme value priority clustering strategy

Abstract

The present invention discloses a time series clustering method based on an extreme value priority clustering strategy, comprising the following steps: 1. Counting the distribution characteristics of power extreme value points of each cluster segment based on an initial power sequence; 2. Calculating the cluster center vector of the cluster segment based on the number of power extreme values ​​contained in the cluster segment; 3. Calculating the similarity of all adjacent cluster segments in chronological order based on the cluster center vectors of the cluster segment; 4. Merging two consecutive cluster segments with the smallest similarity based on whether adjacent cluster segments contain extreme values; 5. Repeating steps 1-4 until the number of remaining cluster segments meets the requirement, thereby obtaining a clustered power sequence. By increasing the merging priority of cluster segments containing extreme values, the present invention enables the clustering results to effectively reflect the extreme value distribution information of the initial power sequence, thereby reducing clustering errors.

Description

Technical Field

[0001] The present invention relates to the field of power system power curve clustering reduction, and in particular to a time series clustering method based on an extreme value priority clustering strategy. Background Art

[0002] As the proportion of renewable energy continues to increase, the long-term fluctuations and network congestion associated with high renewable energy integration are placing higher demands on the power system in terms of energy storage duration and grid capacity. However, considering both long timescales and complex grid structures significantly increases the computational complexity of the planning problem. Therefore, it is necessary to adopt a cluster reduction strategy to reduce the time scale of the power series and thus the computational complexity of the planning problem.

[0003] Currently, the most commonly used power system power sequence reduction method is the continuous time segment merging method based on hierarchical clustering. However, this method uses an average approach to calculate the cluster centers after clustering the initial power sequence, which causes the clustering results to smooth the initial power sequence, resulting in the loss of extreme value information of the power sequence, which can introduce large errors in the planning results. Therefore, it is necessary to study a power sequence clustering reduction method that can reflect the extreme value information of the power sequence and reduce the clustering error. Summary of the Invention

[0004] In order to address the shortcomings of the above-mentioned prior art, the present invention proposes a time series clustering method based on the extreme value priority clustering strategy, in order to correct the clustering strategy of the time period according to the extreme value distribution information of the power series, thereby improving the accuracy of the clustering results and improving the computational efficiency of the power system planning problem.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] The time series clustering method based on the extreme value priority clustering strategy of the present invention is characterized in that it includes the following steps:

[0007] Step 1: Use formula (1) to construct the initial power sequence P, and define the initial number of cluster segments as O, initialize O=S, and define the initial cluster segment set as I P ;

[0008]

[0009] In formula (1): P n is the power matrix on the nth bus, and are the power sequences of load, wind power and photovoltaic power on the nth bus, N is the total number of buses, and S represents the time length of the initial power sequence;

[0010] Step 2: Use formula (2) to determine the time set T corresponding to the maximum and minimum points of the initial power sequence P ex ;

[0011]

[0012] In formula (2): T ex,n represents the time set corresponding to the extreme point of the initial power sequence on the nth bus, and are the extreme moment sets of the load, wind power and photovoltaic power sequences on the nth bus respectively;

[0013] Step 3: Determine the number of remaining cluster segments Q at the end of clustering;

[0014] Step 4: According to the starting time t of the I-th cluster segment I,S and end time t I,E , calculate the cluster center of each cluster segment; I∈I P ;

[0015] Step 5: According to the cluster segment number set I P The cluster centers corresponding to all cluster segments in the cluster are combined using formula (3) to obtain the total cluster center vector

[0016]

[0017] In formula (3): represents the vector of cluster centers of the load, wind power and photovoltaic data of each node in the I-th cluster segment, represents the vector of cluster centers of the load, wind power and photovoltaic data of each node in the Jth cluster segment adjacent to the Ith cluster segment, R 1,I Indicates the initial power sequence P at T I The set of all columns containing extreme values ​​in the period, Represents R in the initial power sequence P 1,I The cluster center in the I-th cluster segment, R 2,I Indicates the initial power sequence P at T I The set of all columns that do not contain extreme values ​​in the period, Represents R in the initial power sequence P 2,I The cluster center in the I-th cluster segment;

[0018] Step 6: Calculate using formula (4) and The similarity between them is DIS(I,J);

[0019]

[0020] In formula (4), t J,S and t J,E Respectively represent the start time and end time of the J-th cluster segment;

[0021] Step 7: Determine the sequence numbers of the two adjacent cluster segments I′ and J′ with the smallest similarity according to formula (5) and use them as candidate merged cluster segments;

[0022]

[0023] Step 8: Determine the cluster segment merging method based on the extreme value distribution of the I′th and J′th cluster segments, and merge the I′th and J′th cluster segments;

[0024] Step 9: Output the cluster centers of each cluster segment And calculate the cluster set I according to formula (6) P The duration τ of the I-th cluster segment in I , thus obtaining the cluster set I P The duration τ of each cluster segment in :

[0025] τ I =t I,E -t I,S +1 I∈I P (6).

[0026] The time series clustering method based on the extreme value priority clustering strategy described in the present invention is also characterized in that the step four includes:

[0027] Step 4.1: According to T I The number of extreme values ​​in the segment is calculated as follows: 1,I Cluster centers in the Ith cluster segment:

[0028] When T I When the time period contains only one extreme value, use formula (7) to calculate the R in the initial power sequence P 1,I Cluster center in the Ith cluster segment

[0029]

[0030] In formula (7): Indicates that the rth column in the initial power sequence P corresponds to T in the Ith cluster segment I Power extremes during the time period;

[0031] When T I When the number of extreme values ​​in a time period is greater than 1, use formula (8) to calculate the R 1,I The cluster center of the I-th cluster segment

[0032]

[0033] In formula (8): Indicates that the rth column in the initial power sequence P corresponds to the time period T in the I cluster segment I The kth power extreme value in the initial power sequence P, K represents R 1,I In the time period T corresponding to the I-th cluster segment I The total number of extreme values ​​in ;

[0034] Step 4.2: Use formula (9) to calculate R in the initial power sequence P 2,I Cluster center in the Ith cluster segment

[0035]

[0036] In formula (9): Indicates that the rth column in the initial power sequence P corresponds to the time period T in the I cluster segment I The power value at time u in .

[0037] The eighth step comprises:

[0038] Step 8.1: If Then merge I′ and J′ into U and execute step 8.3;

[0039] Step 8.2: If Then let DIS(I′,J′) be a positive number M, and determine whether the similarities between adjacent cluster segments are all M. If not, return to step 7. If they are all M, execute step 8.3.

[0040] Step 8.3: Merge I′ and J′ according to formula (10) and update the cluster segment set I P Assign the sum of the number of cluster segments O-1 to O; if O≤Q, execute step 9, otherwise, return to step 4;

[0041]

[0042] In formula (10), t I′,S Indicates the starting time of the I′th cluster segment, t J′,E Indicates the end time of the J′th cluster segment.

[0043] The electronic device of the present invention includes a memory and a processor, and is characterized in that the memory is used to store a program that supports the processor to execute the time series clustering method, and the processor is configured to execute the program stored in the memory.

[0044] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program executes the steps of the time series clustering method when executed by a processor.

[0045] Compared with the prior art, the beneficial effects of the present invention are embodied in:

[0046] 1. The present invention overcomes the problem that traditional time series clustering produces a smoothing effect on the power series clustering results, and proposes an extreme value priority clustering strategy, which effectively retains the extreme value distribution information of the power series by increasing the merging priority of cluster segments containing extreme values.

[0047] 2. The present invention modifies the cluster center calculation method of the traditional time series clustering method according to the number of extreme values ​​contained in two adjacent cluster segments in the time series, so that the cluster center calculation result is closer to the extreme value size of the power sequence, thereby improving the accuracy of the clustering result.

[0048] 3. The present invention proposes a time series clustering method based on an extreme value priority clustering strategy, which solves the problem of loss of extreme value information of the power series caused by calculating the cluster center based on the average idea compared with the existing technology. This method corrects the cluster center according to the distribution of extreme value points of adjacent cluster segments, and at the same time improves the merging priority of cluster segments containing extreme values, which can effectively retain the extreme value distribution of the power series and reduce the clustering error. The present invention proposes a time series clustering method based on an extreme value priority clustering strategy that can reduce the data dimension of the initial input data before power system planning and scheduling, thereby reducing the running time of the planning program while ensuring the accuracy of power system planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The figure is a flow chart of a time series clustering method based on an extreme value priority clustering strategy of the present invention. DETAILED DESCRIPTION

[0050] In order to have a further understanding and recognition of the structural features and the effects achieved by the present invention, a detailed description is provided with reference to preferred embodiments and accompanying drawings as follows:

[0051] like Figure 1 As shown in FIG, a time series clustering method based on the extreme value priority clustering strategy includes the following steps:

[0052] Step 1: Use formula (1) to construct the initial power sequence P, define the initial number of cluster segments as O, and initialize O=S, define the initial cluster segment set as I P ;

[0053]

[0054] In formula (1): Pn is the power matrix on the nth bus, and are the power sequences of load, wind power and photovoltaic power on the nth bus, N is the total number of buses, and S represents the time length of the initial power sequence.

[0055] Step 2: Use formula (2) to determine the time set T corresponding to the maximum and minimum points of the initial power sequence P ex ;

[0056]

[0057] In formula (2): T ex,n represents the time set corresponding to the extreme point of the initial power sequence on the nth bus, and are the extreme moment sets of the load, wind power and photovoltaic power sequences on the nth bus respectively.

[0058] Step 3: Determine the number of remaining cluster segments Q at the end of clustering;

[0059] Step 4: According to the starting time t of the I-th cluster segment I,S and end time t I,E , calculate the cluster center of each cluster segment; I∈I P ;

[0060] Step 4.1: According to T I The number of extreme values ​​in the segment is calculated as follows: 1,I Cluster centers in the Ith cluster segment:

[0061] When T I When the time period contains only one extreme value, use formula (3) to calculate the R in the initial power sequence P 1,I Cluster center in the Ith cluster segment

[0062]

[0063] In formula (3): Indicates that the rth column in the initial power sequence P corresponds to T in the Ith cluster segment I Power extremes during the time period;

[0064] When T I When the number of extreme values ​​in a time period is greater than 1, use formula (8) to calculate the R 1,I The cluster center of the I-th cluster segment

[0065]

[0066] In formula (4): Indicates that the rth column in the initial power sequence P corresponds to the time period T in the I cluster segment I The kth power extreme value in the initial power sequence P, K represents R 1,I In the time period T corresponding to the I-th cluster segment I The total number of extreme values ​​in .

[0067] Step 4.2: Use formula (5) to calculate R in the initial power sequence P 2,I Cluster center in the Ith cluster segment

[0068]

[0069] In formula (5): Indicates that the rth column in the initial power sequence P corresponds to the time period T in the I cluster segment I The power value at time u in .

[0070] Step 4 further explanation:

[0071] The calculation idea of ​​the cluster center of each cluster segment is as follows: if the cluster segment contains only one extreme value, then the extreme value can reflect the maximum value of the power sequence in the cluster segment, and the extreme value of the cluster segment should be retained in the clustering result. Therefore, the extreme value size of the cluster segment is used as the cluster center; if the cluster segment contains more than one extreme value, it means that the power sequence in the cluster segment presents a fluctuating characteristic. At this time, the average value of each extreme value is calculated as the cluster center of the cluster segment to approximately reflect the extreme value distribution information in the cluster segment.

[0072] Step 5: According to the cluster segment number set I P The cluster centers corresponding to all cluster segments in the cluster are combined using formula (3) to obtain the total cluster center vector

[0073]

[0074] In formula (6): represents the vector of cluster centers of the load, wind power and photovoltaic data of each node in the I-th cluster segment; represents the vector of cluster centers of the load, wind power and photovoltaic data of each node in the Jth cluster segment adjacent to the Ith cluster segment, R 1,I Indicates the initial power sequence P at T I The set of all columns containing extreme values ​​in the period, Represents R in the initial power sequence P 1,I The cluster center in the I-th cluster segment, R 2,I Indicates the initial power sequence P at T IThe set of all columns that do not contain extreme values ​​in the period, Represents R in the initial power sequence P 2,I The cluster center in the I-th cluster segment.

[0075] Step 6: Calculate using formula (7) and The similarity between them is DIS(I,J);

[0076]

[0077] In formula (7), t J,S and t J,E Respectively represent the start time and end time of the J-th cluster segment;

[0078] Step 7: Determine the sequence numbers of the two adjacent cluster segments I′ and J′ with the smallest similarity according to formula (8) and use them as candidate merged cluster segments;

[0079]

[0080] Step 8: Determine the cluster segment merging method based on the extreme value distribution of the I′th and J′th cluster segments, and merge the I′th and J′th cluster segments;

[0081] Step 8.1: If At this point, it means that neither cluster segment I′ nor J′ contains extreme values, so I′ and J′ are merged into U and step 8.3 is executed;

[0082] Step 8.2: If This means that at least one segment in clusters I′ and J′ contains an extreme value. Let DIS(I′, J′) be a positive number M, and determine whether the similarities between adjacent cluster segments are all M. If not, return to step 7. If all are M, execute step 8.3.

[0083] Step 8.3: Merge I′ and J′ according to formula (9) to update the cluster segment set I P , use formula (6) to determine the updated cluster segment set I P The corresponding cluster center And update the number of cluster segments to O-1; if O≤Q, execute step 9, otherwise, return to step 4;

[0084] T U ={t|t I′,S ≤t≤t J′,E} (9)

[0085] In formula (9), t I′,S Indicates the starting time of the I′th cluster segment, t J′,E Indicates the end time of the J′th cluster segment.

[0086] Step 9: Output the cluster centers of each cluster segment And calculate the cluster set I according to formula (10) P The duration τ of the I-th cluster segment in I , thus obtaining the cluster set I P The duration τ of each cluster segment in :

[0087] τ I =t I,E -t I,S +1 I∈I P (10)

[0088] Although the specific implementation methods of the present invention are described above, those skilled in the art should understand that these are merely examples and that various changes or modifications may be made to these implementation methods without departing from the principles and implementations of the present invention. Therefore, the scope of protection of the present invention is limited by the appended claims.

Claims

1. A time series clustering method based on extreme value priority clustering strategy, characterized in that: The following steps are involved: Step 1: Use formula (1) to construct the initial power sequence P, and define the initial number of cluster segments as O, initialize O=S, and define the initial cluster segment set as I P ; In formula (1): P n is the power matrix on the nth bus, and are the power sequences of load, wind power and photovoltaic power on the nth bus, N is the total number of buses, and S represents the time length of the initial power sequence; Step 2: Use formula (2) to determine the time set T corresponding to the maximum and minimum points of the initial power sequence P ex ; In formula (2): T ex,n represents the time set corresponding to the extreme point of the initial power sequence on the nth bus, and are the extreme moment sets of the load, wind power and photovoltaic power sequences on the nth bus respectively; Step 3: Determine the number of remaining cluster segments Q at the end of clustering; Step 4: According to the starting time t of the I-th cluster segment I,S and end time t I,E , calculate the cluster center of each cluster segment; I∈I P ; Step 5: According to the cluster segment number set I P The cluster centers corresponding to all cluster segments in the cluster are combined using formula (3) to obtain the total cluster center vector In formula (3): represents the vector of cluster centers of the load, wind power and photovoltaic data of each node in the I-th cluster segment, represents the vector of cluster centers of the load, wind power and photovoltaic data of each node in the Jth cluster segment adjacent to the Ith cluster segment, R 1,I Indicates the initial power sequence P at T I The set of all columns containing extreme values ​​in the period, Represents R in the initial power sequence P 1,I The cluster center in the I-th cluster segment, R 2,I Indicates the initial power sequence P at T I The set of all columns that do not contain extreme values ​​in the period, Represents R in the initial power sequence P 2,I The cluster center in the I-th cluster segment; Step 6: Calculate using formula (4) and The similarity between them is DIS(I,J); In formula (4), t J,S and t J,E Respectively represent the start time and end time of the J-th cluster segment; Step 7: Determine the sequence numbers of the two adjacent cluster segments I′ and J′ with the smallest similarity according to formula (5) and use them as candidate merged cluster segments; Step 8: Determine the cluster segment merging method based on the extreme value distribution of the I′th and J′th cluster segments, and merge the I′th and J′th cluster segments; Step 9: Output the cluster centers of each cluster segment And calculate the cluster set I according to formula (6) P The duration τ of the I-th cluster segment in I , thus obtaining the cluster set I P The duration τ of each cluster segment in : τ I =t I,E -t I,S +1 I∈I P (6)。 2. The time series clustering method based on the extreme value priority clustering strategy according to claim 1, characterized in that: The fourth step includes: Step 4.1: According to T I The number of extreme values ​​in the segment is calculated as follows: 1,I Cluster centers in the Ith cluster segment: When T I When the time period contains only one extreme value, use formula (7) to calculate the R in the initial power sequence P 1,I Cluster center in the Ith cluster segment In formula (7): Indicates that the rth column in the initial power sequence P corresponds to T in the Ith cluster segment I Power extremes during the time period; When T I When the number of extreme values ​​in a time period is greater than 1, use formula (8) to calculate the R 1,I The cluster center of the I-th cluster segment In formula (8): Indicates that the rth column in the initial power sequence P corresponds to the time period T in the I cluster segment I The kth power extreme value in the initial power sequence P, K represents R 1,I In the time period T corresponding to the I-th cluster segment I The total number of extreme values ​​in ; Step 4.2: Use formula (9) to calculate R in the initial power sequence P 2,I Cluster center in the Ith cluster segment In formula (9): Indicates that the rth column in the initial power sequence P corresponds to the time period T in the I cluster segment I The power value at time u in .

3. The time series clustering method based on extreme value priority clustering strategy according to claim 2, characterized in that: The eighth step comprises: Step 8.1: If Then merge I′ and J′ into U and execute step 8.3; Step 8.2: If Then let DIS(I′,J′) be a positive number M, and determine whether the similarities between adjacent cluster segments are all M. If not, return to step 7. If they are all M, execute step 8.

3. Step 8.3: Merge I′ and J′ according to formula (10) and update the cluster segment set I P Assign the sum of the number of cluster segments O-1 to O; if O≤Q, execute step 9, otherwise, return to step 4; T U ={t|t I′,S ≤t≤t J′,E } (10) In formula (10), t I′,S Indicates the starting time of the I′th cluster segment, t J′,E Indicates the end time of the J′th cluster segment.

4. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store a program that supports the processor to execute the time series clustering method according to any one of claims 1 to 3, and the processor is configured to execute the program stored in the memory.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the time series clustering method according to any one of claims 1 to 3 are performed.

Images