Flexible load identification method and device

Abstract

This invention discloses a method and apparatus for flexible load identification. The method includes: acquiring target load data corresponding to multiple electricity accounts in a target power grid during a first sampling period; performing a first classification process on the target load data corresponding to the multiple electricity accounts to obtain a first load classification result, wherein the first load classification result includes: a first category of load data composed of rigid loads, and a second category of load data composed of composite loads, wherein the composite loads include both rigid and flexible loads; and using the K-nearest neighbor algorithm based on the second category of load data to obtain the flexible load identification result corresponding to the target power grid. This invention solves the technical problem in related technologies where the accuracy of user-side flexible load identification is low, leading to poor distribution network control balance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of flexible load identification, in particular to a flexible load identification method and device. BACKGROUND

[0002] With the continuous advancement and in-depth development of the smart grid, combined with the improvement of technologies such as demand response and energy efficiency management, the inherent characteristics of user-side load have changed, and have certain adjustment ability, resulting in a new dispatchable load resource-flexible load (FL). Flexible load mainly includes interruptible load, adjustable load and transferable load, etc. Compared with rigid load, flexible load can participate in grid dispatching. With the development of distributed energy, the load adjustment ability of the user side is further enhanced. However, distributed energy includes photovoltaic power generation, wind power generation, new energy electric vehicles, etc. Distributed energy can participate in grid dispatching on the one hand, which can maintain the stability and safety of the distribution system, and on the other hand, due to its environmental properties, it cannot achieve accurate identification of user-side flexible load, thereby causing the output of distributed energy to be unstable and unable to achieve the balance of distribution network regulation and control.

[0003] At present, no effective solution has been proposed for the above problems. SUMMARY

[0004] The embodiments of the present application provide a flexible load identification method and device, to at least solve the technical problem of low accuracy of user-side flexible load identification in related technologies, thereby causing poor balance of distribution network regulation and control.

[0005] According to an aspect of an embodiment of the present application, a flexible load identification method is provided, comprising: obtaining target load data corresponding to a plurality of electricity accounts in a target power grid in a first sampling period; performing first classification processing on the target load data corresponding to the plurality of electricity accounts to obtain a first load classification result, wherein the first load classification result includes: first type load data composed of rigid load, and second type load data composed of composite load, and the composite load includes rigid load and flexible load; based on the second type load data, using a K-nearest neighbor algorithm to obtain a flexible load identification result corresponding to the target power grid.

[0006] According to another aspect of the embodiments of the present application, a flexible load identification device is also provided, comprising: an acquisition module configured to acquire target load data corresponding to a plurality of power consumption accounts in a target power grid in a first sampling period; a classification module configured to perform first classification processing on the target load data corresponding to the plurality of power consumption accounts to obtain a first load classification result, wherein the first load classification result comprises: first load data composed of rigid load, and second load data composed of composite load, and the composite load comprises rigid load and flexible load; and an identification module configured to obtain a flexible load identification result corresponding to the target power grid based on the second load data and using a K-nearest neighbor algorithm.

[0007] In the embodiments of the present application, by acquiring target load data corresponding to a plurality of power consumption accounts in a target power grid in a first sampling period, performing first classification processing on the target load data corresponding to the plurality of power consumption accounts to obtain a first load classification result, wherein the first load classification result comprises: first load data composed of rigid load, and second load data composed of composite load, and the composite load comprises rigid load and flexible load, and obtaining a flexible load identification result corresponding to the target power grid based on the second load data and using a K-nearest neighbor algorithm, the purpose of accurately identifying flexible load on the user side is achieved, thereby realizing the technical effect of improving the accuracy of flexible load identification on the user side and realizing the balanced and stable operation of the power grid system, and further solving the technical problem of low accuracy of flexible load identification on the user side in the related art, which further leads to poor balance of power distribution network regulation. BRIEF DESCRIPTION OF DRAWINGS

[0008] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0009] Figure 1 FIG. 1 is a schematic diagram of a flexible load identification method according to an embodiment of the present application;

[0010] Figure 2 FIG. 2 is a schematic diagram of an alternative flexible load identification method according to an embodiment of the present application;

[0011] Figure 3 FIG. 3 is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0012] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application in order to make the technical personnel in the technical field better understand the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by the ordinary technical personnel in the technical field without creative labor should belong to the scope of protection of the present application.

[0013] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device that includes a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0014] First, for the convenience of understanding the embodiments of the present application, the following will explain some terms or nouns involved in the present application:

[0015] The K-nearest neighbor algorithm (i.e. K-means algorithm) is used to judge the similarity between data by using a certain mathematical formula to calculate the distance between data, the more similar the distance is, the higher the degree of similarity is, and the data sets with high similarity are merged into the same category, and the data sets between different categories have a large distance, so that the data sets are formed with compact within the class and distant between the classes. The specific process of the algorithm is as follows:

[0016] The K-means algorithm uses geometric distance as an evaluation index of whether the data points have similarity, if the geometric distance of two data points is smaller, the degree of similarity is higher, that is, the two data points can be divided into a unified category, and the flow steps of the algorithm are as follows:

[0017] It is known that the selected data set contains N sample data, each sample data has n-dimensional feature dimension, that is, each data is expressed in the following form: Zi=(zi1,zi2,...,zin), i=1,2,...,N; Set the required number of clusters K, and randomly extract the same number of data vectors as the set cluster number from the selected data set as the initial cluster center, and the expression form of the cluster center is as follows: Cj=(cj1,cj2,...,cjn), j=1,2,...,K; Calculate the geometric distance of each data point in the data set to the cluster center, use the Euclidean distance calculation, and put the data point into the class with the smallest distance, and the calculation formula is as follows:

[0018]

[0019] The algorithm uses the sum of the squared errors (SSE) as the clustering criterion function, and the sum of the squared errors (SSE) is defined as:

[0020]

[0021] In the formula, K is the number of clusters; Nj is the number of samples in the jth class, and aj is the mean of the jth class sample, which represents the cluster center of the data sample category, that is:

[0022] The core idea of the traditional K-means algorithm is: through multiple iteration calculations to achieve the expected target, in each iteration process, the sample center of each class sample set needs to be updated, that is, by calculating the data mean (geometric center) of each class sample set, and replacing the original sample center with the calculation result of each iteration as the clustering reference center for the next iteration. After several iterations of calculation, the final sample center can be considered as the true cluster centroid. This method pre-sets the cluster number K, and randomly extracts K data from the target data set as the initial sample center, and aggregates the data close to the center by distance calculation, and then calculates the new sample center, and iterates until the calculated center no longer changes, that is, Jw converges, and the clustering is completed.

[0023] Although the K-means algorithm has the advantages of easy understanding, convenient use, and fast and efficient calculation of large data sets, it still has defects:

[0024] First, the number of classes to be divided in the data set must be set in advance, that is, the value of K, and the final quality of the clustering will be affected by the value of K. If the specific dispersion state of the data sample is unknown, it will be very difficult to set the number of classes to be divided, which is a big problem to be solved.

[0025] Second, the selected sample center initial value in iteration has an important influence on the subsequent results. The randomly selected initial centroid has differences, and the final clustering results will also be different, so the initial center affects the stability of the final result, and if the selected initial sample center deviates from the global optimal exploration field, the final iteration result will lead to the appearance of a local optimal solution.

[0026] Based on the above problems, the embodiment of the present application provides a flexible load identification method embodiment. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in a different order.

[0027] Figure 1 The flowchart of the flexible load identification method according to the embodiment of the present application is shown in FIG. 1, which comprises the following steps: Figure 1

[0028] In step S102, the target load data corresponding to a plurality of electricity accounts in the target power grid in the first sampling period is obtained.

[0029] Optionally, the target power grid is a distribution network system containing a distributed micro-grid. The target load data corresponding to a plurality of electricity accounts is load data in vector form, and the target load data corresponding to each electricity account corresponds to a load vector. The first sampling period can be, but is not limited to, one day. The dimension of the load vector corresponds to the number of segments divided in one day, that is, the target load data corresponding to each electricity account can be regarded as an n-dimensional vector data, and n is the number of time segments divided in one day. For example, if sampling is performed once an hour, n = 24, then the initial load data of one electricity account is a 24-dimensional vector, and n can be flexibly selected according to the actual load change of the distribution network. For example, the greater the load level fluctuation, the greater the value of n, and the greater the identification accuracy. In necessary cases, n = 24*60, that is, minute-level identification can be realized.

[0030] In an optional embodiment, the step of obtaining the target load data corresponding to a plurality of electricity accounts in the target power grid in the first sampling period comprises the following steps: obtaining the initial load data corresponding to a plurality of electricity accounts in the first sampling period; determining whether there is abnormal data in the initial load data corresponding to a plurality of electricity accounts; and in the case where there is no abnormal data in the initial load data corresponding to a plurality of electricity accounts, taking the initial load data corresponding to a plurality of electricity accounts as the target load data corresponding to a plurality of electricity accounts. Through the above method, the load data on the user side is preprocessed, which can effectively remove abnormal data and improve the accuracy level of subsequent load identification.​

[0031] Optionally, each user in the power grid corresponds to an electricity account, and initial load data corresponding to each electricity account is user initial load data. The user initial load data is calculated according to the power demand data of the electricity account at different time points in a day from the power system, that is, the user initial load data can be regarded as an n-dimensional vector data, and n is the number of time periods divided in a day. In the power distribution network, the user side includes N electricity accounts, and therefore it is necessary to collect initial load data of the N electricity accounts. The initial load of each electricity account is a sample data Xi=(xi1, xi2,..., xin), and the initial load data corresponding to the multiple electricity accounts of the user side respectively form a sample set A={X1, X2,..., XN}. Considering that in the actual power grid, environmental factors such as weather, human factors of abnormal user-side electricity behaviors, and system factors of communication quality changes when the power communication device acquires data will affect the acquired load data, the collected data needs to be preprocessed to improve the data acquisition quality and provide data support for subsequent load identification. The specific processing method is divided according to the type of abnormal data: specifically including first-class abnormal data and second-class abnormal data. The first-class abnormal data includes blank data, repeated data and impact data, and the second-class abnormal data includes invalid data.

[0032] In an optional embodiment, in the case where the above abnormal data includes first-class abnormal data, the above method further includes: in the case where the first-class abnormal data exists in the initial load data corresponding to any one of the multiple electricity accounts, acquiring first historical sampling data of the any one electricity account corresponding to the first-class abnormal data; based on the first historical sampling data, using the K-nearest neighbor algorithm to obtain first supplementary data corresponding to the first-class abnormal data; replacing the first-class abnormal data with the first supplementary data to obtain target load data corresponding to the any one electricity account.

[0033] Optionally, the first historical sampling data corresponds to the sampling time of the first-class abnormal data, and the first-class abnormal data is any one of blank data, repeated data and impact data.

[0034] Optionally, for the blank data, the generation reasons include database storage space shortage, communication impairment, and acquisition device damage, etc. Such data is represented as part of the continuous data (not all data) in the vector being 0. For such data, the historical load data of the power account on different dates is obtained, and the remaining normal data in the vector is extracted. The KNN algorithm is used to calculate the closest load data, and the blank data is supplemented according to the closest load data. A specific example is as follows: assuming that the initial load data Xi=(xi1, xi2,..., xin) is obtained, where xi2-xi12 are all 0, the load data Yi=(yi1, yi2,..., yin) of the corresponding user on different dates is obtained, where i is the date identifier corresponding to different dates, and the Euclidean distance is calculated According to the ascending order, the smallest L is obtained, and the corresponding Yi data is taken as the first supplementary data to fill the blank data of Xi.

[0035] Optionally, for the repeated data, the generation reasons include database impairment and communication anomaly. Part of the continuous data (not including the case of the whole data) in such data appears repeatedly. For such data, the repeatedly appearing data is first removed, and then the same processing method as the blank data is used to determine the corresponding first supplementary data for filling processing.

[0036] Optionally, for the impact data, the generation reasons include abnormal power use of the user, abnormal communication, abnormal acquisition device, and abnormal power use behavior due to regional and temporal influence, etc. Such data is represented as local adjacent time period load sharp rise and load sharp drop, and local load data is too large or too small. For such data, it is first judged whether the data anomaly is caused by the user's own power use. The judgment method is to judge the rationality of the load change. A system load change threshold Pl is set, and the sum of the two change values of the continuous three elements in the load data vector max{ |xi(b-1)-xib| + |xib-xi(b+1)|} is calculated. When max{ |xi(b-1)-xib| + |xib-xi(b+1)|}≥Pl, the data is not caused by the user's own power use anomaly. At this time, the abnormal point is removed, and the same processing method as the blank data is used to determine the corresponding first supplementary data for filling processing.

[0037] In an optional embodiment, in the case that the abnormal data includes the second type of abnormal data, the method further includes: in the case that the second type of abnormal data exists in the initial load data corresponding to any one of the plurality of electricity accounts, obtaining second historical sampling data of the any one of the electricity accounts corresponding to the second type of abnormal data; based on the second historical sampling data, using the K-nearest neighbor algorithm to obtain second supplementary data corresponding to the second type of abnormal data; and taking the second supplementary data as the target load data corresponding to the any one of the electricity accounts.

[0038] Optionally, the sampling time of the second historical sampling data corresponds to the first sampling period, and the second type of abnormal data at least includes invalid data. For invalid data, it is manifested as that the entire load data is too large or too small (or all empty data), or the entire data is repeated, and the entire property is steeply rising or steeply falling. Such data cannot be repaired, can only be removed, and the average value of the load data of other dates of the corresponding user is used as the second supplementary data to replace the original data.

[0039] Step S104, performing first classification processing on the target load data corresponding to the plurality of electricity accounts respectively to obtain first load classification results, wherein the first load classification results include: first type of load data composed of rigid load, and second type of load data composed of composite load, and the composite load includes rigid load and flexible load.

[0040] In an optional embodiment, the first classification processing on the target load data corresponding to the plurality of electricity accounts respectively to obtain the first load classification results includes: obtaining load modes corresponding to the plurality of electricity accounts respectively, wherein the load mode is a rigid load mode or a composite load mode; taking the target load data corresponding to a first electricity account with the rigid load mode as the first type of load data among the plurality of electricity accounts; taking the target load data corresponding to a second electricity account with the composite load mode as the second type of load data among the plurality of electricity accounts; and obtaining the first load classification results based on the first type of load data and the second type of load data. Through the above manner, first, the user-side load data is initially classified according to the load mode corresponding to the plurality of electricity accounts, that is, whether the electricity account is rigid load or composite load (that is, both rigid load and flexible load exist), and the first type of load data composed of rigid load and the second type of load data composed of composite load obtained by classification are used for further flexible load identification.

[0041] Step S106, based on the second type of load data, using the K-nearest neighbor algorithm to obtain the flexible load identification result corresponding to the target power grid.

[0042] In an optional embodiment, the flexible load identification result corresponding to the target power grid is obtained based on the second type of load data by using the K-nearest neighbor algorithm, including: obtaining a plurality of flexible load forms corresponding to the plurality of electricity accounts, and a first combination number of flexible load combinations composed of one or more of the plurality of flexible loads, wherein the plurality of flexible load forms at least include: air conditioner cluster, water heater cluster, electric vehicle charging pile, air source heat pump; based on the first combination number, the target clustering number of clustering the second type of load data is determined by using dichotomy; based on the target clustering number and the second type of load data, the K-nearest neighbor algorithm is used to obtain the flexible load identification result.

[0043] In the above manner, first, the sample center initial value is determined, since the user-side load is mainly of two modes: rigid load, rigid load + flexible load, and the flexible load is mainly interruptible load, and the regulation and control of the flexible load is unified and coordinated according to the agreement signed by the power grid system, therefore, the user who has not signed the agreement is the user who does not contain flexible load, that is, the initial sample center of the user data containing only rigid load is selected, the sample as a whole (i.e., the target load data corresponding to a plurality of electricity accounts) is divided into two categories according to the load mode, including: the first type of load data composed of rigid load, and the second type of load data composed of composite load (i.e., rigid load and flexible load), the first type of sample clustering is filtered by using the existing first type of load data corresponding to the rigid load, and the remaining second type of load data is clustered again by using the K-means algorithm; when the K-means algorithm is used for the second type of load data, the number of categories (i.e., K value) that the data set should be divided into should be determined, so as to avoid local optimal solution caused by improper selection of K value, since the flexible load on the user side is currently mainly of the following forms: air conditioner cluster, water heater cluster, electric vehicle pile and air source heat pump, which are for individual users and industrial users, the proportions of these types of flexible load forms are different, by combination, there are combinations, that is, the combination interval is By using the concept of dichotomy, the target clustering number of clustering the second type of load data is determined, and on this basis, the K-nearest neighbor algorithm is used to obtain the corresponding flexible load identification result based on the target clustering number and the second type of load data, thereby improving the accuracy of clustering number acquisition and realizing accurate identification of flexible load.

[0044] In an optional embodiment, the target cluster number of the second type of load data is determined by using the dichotomy method based on the first combination number, including: determining a first combination interval corresponding to the first combination number; using the clustering quality as reference index data, and using the dichotomy method to dichotomize the first combination interval to determine the target cluster number of the second type of load data.

[0045] Optionally, the clustering quality can be represented by the Davison-Botting index.

[0046] Optionally, the target cluster number of the second type of load data is determined by using the dichotomy method based on the first combination number, including: determining a first combination interval corresponding to the first combination number; using the clustering quality as reference index data, and using the dichotomy method to dichotomize the first combination interval to determine the target cluster number of the second type of load data.

[0047] Step S1: using the dichotomy method to dichotomize the first combination interval to obtain a first median corresponding to the first combination interval, and using the first median as a first cluster number;

[0048] Step S2: obtaining a first clustering quality of the second type of load data based on the first cluster number, and using the first clustering quality as a reference clustering quality;

[0049] Step S3: determining a second combination interval composed of a lower limit value of the first combination interval and the first median, and a third combination interval composed of the first median and an upper limit value of the first combination interval;

[0050] Step S4: using the dichotomy method to dichotomize the second combination interval to obtain a second median corresponding to the second combination interval, using the second median as a second cluster number, and using the dichotomy method to dichotomize the third combination interval to obtain a third median corresponding to the third combination interval, using the third median as a third cluster number;

[0051] Step S5: obtaining a second clustering quality of the second type of load data based on the second cluster number, and a third clustering quality of the second type of load data based on the third cluster number;

[0052] Step S6: using the minimum value of the second clustering quality and the third clustering quality as a new reference clustering quality, and using the combination interval corresponding to the minimum value as a new first combination interval, repeating the operations of steps S3 to S5 until the new second clustering quality and the new third clustering quality based on the new first combination interval are both less than the corresponding new reference quality in two consecutive cycles.

[0053] Step S7, taking the median of the new first combination interval corresponding to the new reference quality as the target cluster number.

[0054] Optionally, the above first combination number is In the combination case, the first combination interval is In the case, the cluster quality is taken as the reference index data, and the above dichotomy is used to dichotomize the first combination interval to determine the target cluster number for clustering the second type of load data, and specifically includes the following sub-steps:

[0055] Step S11, first selecting i.e. the median value of the combination number interval, to calculate the cluster quality after clustering, and the cluster quality is evaluated by the Davies-Bouldin index DBI index, and the calculation formula is as follows: In the formula, Si and Sj represent the dispersion degree of the i-th and j-th class object; K represents the cluster number; di,j represents the inter-class distance of the i-th and j-th class. The obtained DBI is taken as the reference index data, and then the median values of and are selected respectively, and the DBI is calculated respectively, and the calculated DBI is compared with the reference index data (i.e. the reference cluster quality). Step S12, if the DBI is less than the reference index data, the smaller DBI is used as the new reference index data, and the interval corresponding to the median value is selected, and the interval is dichotomized again, and the operation of step S11 is repeated.

[0056] Step S13, when the DBI is greater than the reference index data, the interval corresponding to the median value of the two median values producing the smaller DBI is selected, the interval is dichotomized again, and the operation of step S11 is repeated.

[0057] Step S14, when the DBI is greater than the reference index data in the continuous two cycles, it is confirmed that the median value producing the reference index data is taken as the cluster K value (i.e. the target cluster number).

[0058] In an optional embodiment, the flexible load identification result corresponding to the target power grid is obtained based on the second type of load data by using the K-nearest neighbor algorithm, including: dividing the first sampling period into a plurality of sub-periods, and determining the period types corresponding to the plurality of sub-periods, wherein the period types are non-regulation period and regulation period; determining third type of load data corresponding to the non-regulation period in the second type of load data, and fourth type of load data corresponding to the regulation period in the second type of load data; determining fifth type of load data corresponding to the non-regulation period in the first type of load data, load type corresponding to the fifth type of load data, and sixth type of load data corresponding to the regulation period in the first type of load data; based on the fifth type of load data and the load type corresponding to the fifth type of load data, the K-nearest neighbor algorithm is used to determine the target load type corresponding to the third type of load data; determining seventh type of load data corresponding to the target load type from the sixth type of load data; and determining the flexible load identification result based on the fourth type of load data and the seventh type of load data.

[0059] It can be understood that the third type of load data is the load data of the non-regulation period in the composite load data, the fourth type of load data is the load data of the non-regulation period in the composite load data; the fifth type of load data is the load data of the non-regulation period in the rigid load data, and the sixth type of load data is the load data of the regulation period in the rigid load data. Through the above method, first, the K-nearest neighbor algorithm is used, and the load type corresponding to the load data of the non-regulation period in the rigid load is determined to determine the target load type of the composite load in the non-regulation period; and on this basis, the load data of the rigid load in the regulation period similar to the target load type (i.e. the seventh type of load data) is selected; by comparing and analyzing the load data of the composite load in the regulation period with the load data of the rigid load in the regulation period of the same load type, the flexible load of the composite load in the regulation period can be obtained. It should be noted that through the above bisection method, the combination form closest to the actual load situation can be deduced.

[0060] In the above steps, the load classification on the user side has been obtained, including pure rigid load and composite load with rigid and flexible load. In order to more accurately obtain the flexible load on the user side, the flexible load on the user side needs to be further screened based on the classified basis. In actual application, the flexible load on the user side is generally aggregated and uniformly regulated, and the time period of the flexible load regulation is relatively fixed, generally in the peak and valley periods of electricity consumption, and the regulation frequency of the flexible load within a day has an upper limit, so the real-time data of the flexible load needs to be correctly mastered to improve the regulation accuracy. The embodiment of the present application further analyzes the composite load on the basis of the user side load identification to obtain the flexible load corresponding to the user, specifically including:

[0061] Based on electricity consumption control periods, a day is divided into multiple time periods, which are divided into non-control periods and control periods. The selection of time periods can be based on the actual control time or peak and off-peak electricity consumption times. Simultaneously, the aforementioned classification samples (i.e., the target load data corresponding to multiple electricity accounts) are also divided into multiple sub-vectors according to this time period division method. For example, a day is divided into the following four time periods: 0-7, 7-11, 11-19, and 20-24, where 7-11 and 20-24 are control periods, and 0-7 and 11-19 are non-control periods. Then, multiple electricity... The target load data corresponding to each account is decomposed into four sub-vectors: Xi0-7, Xi7-11, Xi11-19, and Xi20-24. The sub-vectors corresponding to the non-control period are selected, namely Xi0-7 and Xi11-19, and concatenated in order to form a new vector, namely Xi0-7+Xi11-19. The classification samples of rigid load (i.e., the fifth type of load data) are used as known training samples, and the classification samples of composite load (i.e., the third type of load data) are used as samples to be classified. The closest classification is obtained by using the KNN algorithm.

[0062] Since the sample data has already undergone one round of clustering and classification, the KNN algorithm can quickly yield results. The KNN algorithm uses Euclidean distance to calculate the dissimilarity between data samples, and its mathematical principle is expressed as follows:

[0063] Assume the sample to be classified (i.e., the third type of loading data) is x i (i = 1, 2, 3, ..., m1), where n is the number of electricity accounts corresponding to the fifth type of load data, and c (c = 1, 2, 3, ..., m2) is the category corresponding to the second load classification result, where m2 is the total number of categories. The basic idea of ​​the KNN algorithm is: first, calculate the samples to be classified according to the magnitude of the Euclidean distance value. The dissimilarity between the sample and the fifth category of load data is calculated, where the load data corresponding to each electricity account in the fifth category of load data is used as a training sample. Then, the k training samples with the smallest dissimilarity to the sample to be classified are selected as the final training samples. Find the k nearest neighbors. Finally, based on x... i Determine the k nearest neighbors of x i The category.

[0064] The algorithm starts by constructing a priority queue with a capacity of k, arranged in descending order of distance, to store the nearest neighbors of the test sample. Then, k training samples are randomly selected from the fifth type of load data as the initial nearest neighbor samples. The distances between the k training samples and the test sample are calculated, and the class labels and distances of the k training samples are stored in the priority queue. Then, the training data is traversed, and the distance between the current training sample and the test sample is calculated. If the distance L is greater than or equal to the maximum distance L in the priority queue, the training sample is discarded and the next one is traversed. If the distance L is less than the maximum distance L in the priority queue, the training sample with the maximum distance L is deleted from the priority queue, and the current training sample is stored in the priority queue. When the fifth type of load data is completely traversed, the most common class label among the k training samples in the priority queue is calculated, and the class label with the most common number is taken as the class label of the test sample. To achieve high accuracy, different values of k are set to retrain, and the value of k with the highest accuracy is finally taken. max max max max

[0065] Since the composite load account is considered to have only rigid load demand when unregulated, the fourth type of load data of the composite load account during the regulation period and the seventh type of load data corresponding to the rigid load of the target composite type closest to the composite load account are found. Based on the fourth type of load data and the seventh type of load data, the flexible load identification result corresponding to the composite load account in the target power grid is determined.

[0066] In an optional embodiment, the determination of the flexible load identification result based on the fourth type of load data and the seventh type of load data includes: determining the average difference between the fourth type of load data and the seventh type of load data; and determining the flexible load identification result based on the average difference.

[0067] In the above manner, the average difference between the fourth type of load data and the seventh type of load data is used to determine the flexible load identification result corresponding to the composite load account in the target power grid. The above manner can be achieved by searching for the corresponding load curve, but is not limited thereto. That is, after finding the load curve of the rigid load closest to the composite load account during the unregulated period, the average difference between the composite load curve and the corresponding rigid load curve during the regulation period is compared, and the flexible load value of the composite load account is calculated as the flexible load identification result corresponding to the composite load account. The average difference calculation formula is: ​​​​In the formula, X is the combination of the load sample sub-vector of the composite load account in the regulation period (in the fourth type of load data), Y is the combination of the load sample sub-vector of the rigid load in the regulation period (i.e., the seventh type of load data), Nk is the dimension of the combination of the load sample sub-vector, and the obtained That is, the flexible load data of the composite load user, that is, the load data that can be used for grid regulation.

[0068] Through the above steps S102 to S106, the purpose of accurately identifying the user-side flexible load can be achieved, thereby realizing the technical effects of improving the accuracy of user-side flexible load identification and realizing the balanced and stable operation of the grid system, thereby solving the technical problems of low accuracy of user-side flexible load identification in the related art, and further leading to poor balance of the distribution grid regulation.

[0069] Based on the above embodiments and optional embodiments, an optional implementation of the present application is provided, Figure 2 is a flowchart of an optional flexible load identification method according to an embodiment of the present application, as Figure 2 shown, applied to a distribution grid system containing a distributed microgrid, and the distributed microgrid by default includes a power generation device, an energy storage device, and an adjustable load, i.e., the distributed microgrid can be directly regarded as a flexible load, and the method includes:

[0070] First, the initial load data of the distribution grid needs to be obtained, and the initial load data includes user initial load and distributed microgrid initial load. Each user in the grid corresponds to an electricity account, and the initial load data corresponding to each electricity account is the user initial load data. The user initial load data is calculated according to the electric power data required by the power system at different time points in a day, i.e., the user initial load data can be regarded as an n-dimensional vector data, and n is the number of time periods divided in a day. In the distribution grid, the user side includes N electricity accounts, and therefore the initial load data of N electricity accounts needs to be collected. The initial load of each electricity account is a sample data Xi=(xi1, xi2,..., xin), and the initial load data corresponding to the multiple electricity accounts of the user side respectively constitutes a sample set A={X1, X2,..., XN}.

[0071] The initial load data corresponding to the multiple electricity accounts of the user side respectively is preprocessed to eliminate possible four types of abnormal data (i.e., blank data, repeated data, impact data, and invalid data). After preprocessing the load data of the user side, the abnormal data can be effectively removed, and the target load data corresponding to the multiple electricity accounts respectively is obtained, thereby improving the accuracy level of subsequent load identification.

[0072] Firstly, the sample center initial value is determined. The user side load mainly has two modes: rigid load, rigid load + flexible load, and the flexible load is mainly interruptible load. The regulation and control of flexible load is unified and coordinated according to the agreement signed by the power grid system, so the users who have not signed the agreement are the users without flexible load, that is, the initial sample center of the user data containing only rigid load is selected. The sample as a whole (i.e. the target load data corresponding to multiple electricity accounts) is divided into two categories according to the load mode, including: the first category of load data composed of rigid load, and the second category of load data composed of composite load (i.e. rigid load and flexible load). The first category of load data is used to filter the first category of sample clustering, and the remaining second category of load data is re-clustered using the K-means algorithm. When the second category of load data is used in the K-means algorithm, the number of categories that the data set should be divided into, i.e. K value, needs to be determined to avoid local optimal solution caused by improper selection of K value. Since the flexible load on the user side currently mainly has the following forms: air conditioner cluster, water heater cluster, electric vehicle pile and air source heat pump, which are aimed at individual users and industrial users, the proportion of these types of flexible load is different. Through combination, there are a total of combination situations, i.e. the combination number interval is

[0073] The concept of dichotomy is used to determine the target clustering number of the above-mentioned second category of load data.

[0074] Since the composite load electricity account is considered to have only rigid load demand when unregulated, the fourth type of load data of the composite load electricity account during the regulation period is found, and the seventh type of load data corresponding to the rigid load of the target composite type closest to the composite load electricity account is found. Based on the average value between the fourth type of load data and the seventh type of load data, the flexible load identification result corresponding to the composite load electricity account in the target power grid is determined. The above method can be but is not limited to realized by searching for the corresponding load curve, that is, after finding the load curve of the rigid load closest to the composite load electricity account during the unregulated period, then comparing the average value of the difference between the composite load curve and the corresponding rigid load curve during the regulation period, the flexible load value of the composite load electricity account can be obtained as the flexible load identification result corresponding to the composite load electricity account. The average value difference calculation formula is: In the formula, X is the combination of the load sample sub-vector of the composite load electricity account during the regulation period (in the fourth type of load data), Y is the combination of the load sample sub-vector of the rigid load during the regulation period (i.e. the seventh type of load data), Nk is the dimension of the combination of the load sample sub-vector, and the obtained is the flexible load data of the composite load user, that is, the load data that can be used for power grid regulation.

[0075] At this point, the collection and processing of the load data of the user side and the load identification have been completed, and the identification of the flexible load can achieve accurate differentiation according to the selection of the load data dimension.

[0076] The model applied in the application also includes a distributed micro-grid. The distributed micro-grid includes a power generation device, an energy storage device and a variable load by default. The power generation device refers to power generation modes such as wind power generation, photovoltaic power generation and hydroelectric power generation. These power generation modes are limited by environmental factors, and the power generation power is randomly fluctuating. Therefore, the energy storage device and the adjustable load need to be configured to maintain the balance of the power generation power. The energy storage device is represented by a battery storage, and the variable load refers to a motor type device. The output power of the device can be adjusted to adjust the output power of the distributed micro-grid. In the model, the distributed micro-grid should be used as an output source of grid-connected power. Therefore, the load of the user side is positive, and the load of the distributed micro-grid is negative.

[0077] Since the distributed micro-grid requires the output power to be as smooth as possible when outputting power, the energy storage device and the variable load arranged in the micro-grid can also assist in maintaining the load balance of the distributed micro-grid, so that the load level of the distributed micro-grid can be adjusted within a certain range, and therefore, in the scheme of the present application, the distributed micro-grid is regarded as an output source with flexible load, that is, Pv+PE=Po+Pc, wherein Pv is the load of the power generation device, Pvmin(minimum power generation power)<Pv<Pvmax(maximum power generation power), PE is the load of the energy storage device, PEmin(battery charging and discharging minimum power)<PE<PEmax(battery charging and discharging maximum power), Po is the output power of the distributed micro-grid, and Pc is the flexible load. Since the flexible load can be divided into interruptible load, adjustable load and transferable load, the three types of loads can be realized by the power-off and voltage-regulated charging of the energy storage device, and therefore, PE is a factor directly affecting the flexible load.

[0078] The variable load directly affects the output power of the distributed micro-grid, and further adding the variable load obtains the adjusted load model of the distributed micro-grid: Po=Pv-Pq+PE-Pc=Pv+PE-Pq-Pc, wherein Pq is the variable load, Pq is a control variable within a range, Pqmin(minimum power of the variable load)<Pq<Pqmax(maximum power of the variable load), and the load model of the distributed micro-grid in the above formula is an ideal output model of the distributed micro-grid.

[0079] The load model of the distributed micro-grid established above can keep the load of the distributed micro-grid consistent with the load on the user side, thereby facilitating control. Taking the control during the peak power consumption time as an example, the control includes the following steps:

[0080] First, the over-load demand of the power user side exceeding the load demand outside the power system is obtained, and the available flexible load of the user side is calculated, and then Po=over-load demand-available flexible load is obtained.

[0081] The formula is substituted into the formula above, and Pv+PE-Pq-Pc=over-load demand-available flexible load is obtained, and then Pc=available flexible load is set, so that Pv+PE-Pq=over-load demand is obtained, and the value of Pq can be obtained according to the actual Pv and PE data. By setting the variable load to the corresponding Pq, the distributed micro-grid and the user side load can be coordinated and consistent, thereby maintaining the balance of the distribution network.

[0082] The embodiment of the present application can achieve the following technical effects: a user-side load identification method based on an improved K-means algorithm is designed, the user is divided into two categories of rigid load and composite load by using a step-by-step K-means algorithm, and part of the flexible load is screened out from the composite load based on a KNN algorithm; a load model of a distributed micro-grid is designed, the identification result of the flexible load on the user side is introduced into the load model of the distributed micro-grid, and the balance of the power distribution network is achieved by maintaining the consistency of the flexible load on both sides.

[0083] In the embodiment, a flexible load identification device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be described again. As used below, the term "module" "device" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware or a combination of software and hardware is also possible and is contemplated.

[0084] According to the embodiment of the present application, a device embodiment for implementing the above-mentioned flexible load identification method is also provided, Figure 3 is a structural schematic diagram of a flexible load identification device according to the embodiment of the present application, as Figure 3 shown, the above-mentioned flexible load identification device comprises an acquisition module 300, a classification module 302 and an identification module 304, wherein:

[0085] The above-mentioned acquisition module 300 is used to acquire target load data corresponding to a plurality of power consumption accounts in a target power grid in a first sampling period;

[0086] The above-mentioned classification module 302 is connected to the above-mentioned acquisition module 300 and is used to perform first classification processing on the target load data corresponding to the plurality of power consumption accounts, to obtain a first load classification result, wherein the first load classification result comprises first load data composed of rigid load and second load data composed of composite load, and the composite load comprises rigid load and flexible load;

[0087] The above-mentioned identification module 304 is connected to the above-mentioned classification module 302 and is used to obtain a flexible load identification result corresponding to the target power grid by using a K-nearest neighbor algorithm based on the second load data.

[0088] In the embodiment of the present application, by setting the acquisition module 300, the target load data corresponding to a plurality of power consumption accounts in the target power grid in a first sampling period is acquired; the classification module 302 is connected to the acquisition module 300, and the target load data corresponding to the plurality of power consumption accounts is subjected to first classification processing to obtain a first load classification result, wherein the first load classification result includes: first load data composed of rigid load, and second load data composed of composite load, and the composite load includes rigid load and flexible load; the identification module 304 is connected to the classification module 302, and based on the second load data, the K-nearest neighbor algorithm is used to obtain the flexible load identification result corresponding to the target power grid, so that the purpose of accurately identifying the user-side flexible load is achieved, thereby realizing the technical effect of improving the accuracy of user-side flexible load identification and realizing the balanced and stable operation of the power grid system, and further solving the technical problems of low accuracy of user-side flexible load identification in the related art, and further leading to poor balance of power distribution network regulation.

[0089] It should be noted that each of the above modules can be implemented by software or hardware. For example, for the latter, each of the above modules can be located in the same processor, or each of the above modules can be located in different processors in any combination.

[0090] It should be noted that the acquisition module 300, the classification module 302, and the identification module 304 correspond to steps S102 to S106 in the embodiment, and the above modules have the same instances and application scenarios as the corresponding steps, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules as part of the device can run in a computer terminal.

[0091] It should be noted that the optional or preferred embodiments of the present embodiment can refer to the related description in the embodiment, which will not be repeated here.

[0092] The flexible load identification device described above can further include a processor and a memory, and the acquisition module 300, the classification module 302, and the identification module 304 are stored in the memory as program modules, and the processor executes the above program modules stored in the memory to realize the corresponding functions.

[0093] The processor includes a core, and the core retrieves the corresponding program module from the memory, and the above core can be set to one or more. The memory can include a non-persistent memory in a computer readable medium, a random access memory (RAM), and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM), and the memory includes at least one memory chip.

[0094] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in the embodiment, the non-volatile storage medium comprises a stored program, wherein the program, when executed, controls a device in which the non-volatile storage medium is located to perform any of the flexible load identification methods.

[0095] Optionally, in the embodiment, the non-volatile storage medium can be located in any one of a group of computer terminals in a computer network, or in any one of a group of mobile terminals, and the non-volatile storage medium comprises a stored program.

[0096] Optionally, the program, when executed, controls a device in which the non-volatile storage medium is located to perform the following functions: obtaining target load data corresponding to a plurality of electricity accounts in a target power grid in a first sampling period; performing first classification processing on the target load data corresponding to the plurality of electricity accounts to obtain a first load classification result, wherein the first load classification result comprises: first type load data composed of rigid load, and second type load data composed of composite load, and the composite load comprises rigid load and flexible load; and based on the second type load data, using a K-nearest neighbor algorithm to obtain a flexible load identification result corresponding to the target power grid.

[0097] According to an embodiment of the present application, an embodiment of a processor is also provided. Optionally, in the embodiment, the processor is configured to execute a program, wherein the program, when executed, performs any of the flexible load identification methods.

[0098] According to an embodiment of the present application, an embodiment of a computer program product is also provided, which, when executed on a data processing device, is adapted to execute a program that initializes steps of any of the flexible load identification methods.

[0099] Optionally, the computer program product, when executed on a data processing device, is adapted to execute a program that initializes steps of a method: obtaining target load data corresponding to a plurality of electricity accounts in a target power grid in a first sampling period; performing first classification processing on the target load data corresponding to the plurality of electricity accounts to obtain a first load classification result, wherein the first load classification result comprises: first type load data composed of rigid load, and second type load data composed of composite load, and the composite load comprises rigid load and flexible load; and based on the second type load data, using a K-nearest neighbor algorithm to obtain a flexible load identification result corresponding to the target power grid.

[0100] The electronic device provided by the embodiment of the present application comprises a processor, a memory and a program stored in the memory and executable on the processor, and the processor implements the following steps when executing the program: obtaining target load data corresponding to a plurality of power consumption accounts in a target power grid in a first sampling period; performing first classification processing on the target load data corresponding to the plurality of power consumption accounts to obtain a first load classification result, wherein the first load classification result comprises first load data composed of rigid load and second load data composed of composite load, and the composite load comprises rigid load and flexible load; and obtaining a flexible load identification result corresponding to the target power grid based on the second load data by using a K-nearest neighbor algorithm.

[0101] The serial numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0102] In the above embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0103] In several embodiments provided in the present application, it should be understood that the disclosed technical contents can be implemented by other ways. Among them, the above-mentioned device embodiments are only schematic, for example, the division of the above-mentioned modules can be a logical function division, and actual implementation can have another division way, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some interface, indirect coupling or communication connection of modules or modules, which can be electrical or other forms.

[0104] The above-mentioned modules described as separate components can be or can not be physically separated, and the components displayed as modules can be or can not be physical modules, that is, they can be located in one place, or can be distributed to a plurality of modules. According to actual needs, part or all of the modules can be selected to achieve the purpose of the embodiment of the present application.

[0105] In addition, each functional module in each embodiment of the present application can be integrated in a processing module, or each module can exist physically, or two or more modules can be integrated in one module. The above-mentioned integrated module can be realized in the form of hardware or in the form of software functional module.

[0106] If the above-mentioned integrated modules are realized in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable nonvolatile storage medium. Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the embodiments of the present application. The aforementioned non-volatile storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0107] The above is only the preferred embodiment of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A flexible load identification method, characterized by, The method comprises: obtaining target load data corresponding to a plurality of power consumption accounts in a target power grid in a first sampling period; performing first classification processing on the target load data corresponding to the plurality of power consumption accounts to obtain first load classification results, wherein the first load classification results comprise first type load data composed of rigid load and second type load data composed of composite load, and the composite load comprises rigid load and flexible load; based on the second type load data, using a K-nearest neighbor algorithm to obtain a flexible load identification result corresponding to the target power grid, comprising: obtaining a plurality of flexible load forms corresponding to the plurality of power consumption accounts, and a first combination number of flexible load combinations composed of one or more of the plurality of flexible load forms, wherein the plurality of flexible load forms at least include: air conditioner cluster, water heater cluster, electric vehicle charging pile, air source heat pump; determining a first combination interval corresponding to the first combination number; using clustering quality as reference index data, using dichotomy to perform dichotomy processing on the first combination interval to determine a target clustering number for clustering the second type load data; based on the target clustering number and the second type load data, using the K-nearest neighbor algorithm to obtain the flexible load identification result.

2. The method of claim 1, wherein, The first load classification result is obtained by performing first classification processing on the target load data corresponding to the plurality of power consumption accounts, comprising: obtaining load modes corresponding to a plurality of power consumption accounts, wherein the load mode is a rigid load mode or a composite load mode; target load data corresponding to a first power consumption account in the plurality of power consumption accounts whose load mode is the rigid load mode is taken as the first type load data; target load data corresponding to a second power consumption account in the plurality of power consumption accounts whose load mode is the composite load mode is taken as the second type load data; based on the first type load data and the second type load data, the first load classification result is obtained.

3. The method of claim 1, wherein, The flexible load identification result corresponding to the target power grid is obtained based on the second type load data by using a K-nearest neighbor algorithm, comprising: dividing the first sampling period into a plurality of sub-periods, and determining period types corresponding to the plurality of sub-periods, wherein the period type is a non-regulation period or a regulation period; determining third type load data corresponding to the non-regulation period in the second type load data, and fourth type load data corresponding to the regulation period in the second type load data; determining fifth type load data corresponding to the non-regulation period in the first type load data, load types corresponding to the fifth type load data, and sixth type load data corresponding to the regulation period in the first type load data; based on the fifth type load data and the load types corresponding to the fifth type load data, using the K-nearest neighbor algorithm to determine a target load type corresponding to the third type load data; determining seventh type load data corresponding to the target load type from the sixth type load data; determine the flexible load identification result based on the fourth type of load data and the seventh type of load data.

4. The method of claim 3, wherein, The determination of the flexible load identification result based on the fourth type of load data and the seventh type of load data comprises: determining the average value of the difference between the fourth type of load data and the seventh type of load data; determining the flexible load identification result based on the average value of the difference.

5. The method according to any one of claims 1 to 4, characterized in that, The target load data corresponding to each of the plurality of electricity accounts in the target power grid within the first sampling period comprises: obtaining initial load data corresponding to each of the plurality of electricity accounts within the first sampling period; determining whether there is abnormal data in the initial load data corresponding to each of the plurality of electricity accounts; in the case that there is no abnormal data in the initial load data corresponding to each of the plurality of electricity accounts, the initial load data corresponding to each of the plurality of electricity accounts is taken as the target load data corresponding to each of the plurality of electricity accounts.

6. The method of claim 5, wherein, In the case that the abnormal data includes first type of abnormal data, the method further comprises: in the case that the first type of abnormal data exists in the initial load data corresponding to any one of the plurality of electricity accounts, obtaining first historical sampling data of the any one electricity account corresponding to the first type of abnormal data, wherein the first historical sampling data corresponds to the sampling time of the first type of abnormal data, and the first type of abnormal data is any one of the following: blank data, repeated data, and impact data; based on the first historical sampling data, using the K-nearest neighbor algorithm to obtain first supplementary data corresponding to the first type of abnormal data; replacing the first type of abnormal data with the first supplementary data to obtain target load data corresponding to the any one electricity account.

7. The method of claim 5, wherein, In the case that the abnormal data includes second type of abnormal data, the method further comprises: in the case that the second type of abnormal data exists in the initial load data corresponding to any one of the plurality of electricity accounts, obtaining second historical sampling data of the any one electricity account corresponding to the second type of abnormal data, wherein the second historical sampling data corresponds to the sampling time of the first sampling period, and the second type of abnormal data at least includes invalid data; based on the second historical sampling data, using the K-nearest neighbor algorithm to obtain second supplementary data corresponding to the second type of abnormal data; taking the second supplementary data as the target load data corresponding to the any one electricity account.

8. A flexible load identification device, characterized by comprises: an acquisition module, configured to acquire target load data corresponding to each of a plurality of electricity accounts in a target power grid within a first sampling period; a classification module, configured to perform first classification processing on the target load data corresponding to each of the plurality of electricity accounts to obtain a first load classification result, wherein the first load classification result comprises first type of load data composed of rigid load and second type of load data composed of composite load, and the composite load comprises rigid load and flexible load; The identification module is configured to obtain a flexible load identification result corresponding to the target power grid based on the second type of load data by using a K-nearest neighbor algorithm, including: obtaining a plurality of flexible load forms corresponding to the plurality of electricity accounts, and a first combination number of flexible load combinations composed of one or more of the plurality of flexible load forms, wherein the plurality of flexible load forms at least include: an air conditioner cluster, a water heater cluster, an electric vehicle charging pile, and an air source heat pump; determining a first combination interval corresponding to the first combination number; using clustering quality as reference index data, performing dichotomy processing on the first combination interval by using dichotomy to determine a target clustering number for clustering the second type of load data; and obtaining the flexible load identification result based on the target clustering number and the second type of load data by using the K-nearest neighbor algorithm.

Images