Industrial load abnormal value identification and processing method and system based on nonparametric regression

Abstract

The invention discloses an industrial load abnormal value identification and processing method and system based on nonparametric regression, and the method comprises the steps: collecting industrial load data, and constructing a daily load vector data set according to the collected industrial load data; calculating a power consumption mode classification threshold, and dividing the power consumption mode into a basic power consumption mode and a special power consumption mode; extracting a daily load feature vector of a basic power utilization mode, calculating load similarity, classifying load levels according to the similarity, extracting a center load vector from different types of power utilization loads, calculating a data abnormal load domain, and screening out abnormal loads; and correcting the abnormal value by using a conventional weighted averaging method and a weighted averaging method based on load level mapping. According to the invention, for daily load data information of any time span, abnormal value identification and correction can be simply and effectively realized.

Description

Technical Field

[0001] The present invention belongs to the field of power systems, and in particular relates to a method, system, device and storage medium for identifying and processing abnormal values ​​of industrial loads based on non-parametric regression. Background Art

[0002] With the continuous improvement of the intelligent level of power networks and the deepening of power market reform, power load data has gradually become one of the most widely used data in the fields of power system operation, power consumption analysis and power demand management. These data play an irreplaceable role in optimizing power resource allocation, improving the operating efficiency of power systems and achieving precise control of power demand. However, power load data may deviate from its normal value due to a variety of factors. Therefore, it is particularly important to identify and correct abnormal values ​​in load data before practical application.

[0003] With the advancement of intelligent power systems, a large amount of power load data is collected in real time and uploaded to the centralized control center. Faced with such a huge amount of data, relying on manual means to identify outliers is not only inefficient but also impractical. Therefore, an efficient and reliable automated data processing method must be used to identify and correct outliers in load data.

[0004] At present, in terms of load anomaly identification, a variety of methods have been proposed in existing technical literature, including the use of sample clustering depth technology to identify distorted load shapes and the use of traditional t-test methods to detect abnormal data points; in response to the problem of ATR2 neural network mode drift, an improved abnormal data identification and adjustment model has been proposed; combining the fuzzy soft clustering idea with the Kohonen neural network model, an abnormal data identification and cleaning model based on the radial basis function (RBF) network has been formed. In the above load anomaly identification research, the research objects are mainly concentrated on bus load or system load, which cannot meet the industrial load anomaly identification of industrial park users. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a method, system, equipment and storage medium for identifying and processing industrial load anomaly values ​​based on non-parametric regression. The method can effectively identify abnormal values ​​of important information such as power consumption patterns and power consumption levels caused by interference during the recording of industrial load data, and process the abnormal values.

[0006] Technical solution: A method for identifying and processing abnormal values ​​of industrial loads based on nonparametric regression of the present invention comprises:

[0007] Collect industrial load data and construct a daily load vector data set based on the collected industrial load data;

[0008] Perform time series analysis on the daily load vectors in the daily load vector data set; use the statistical fuzzy matrix method to cluster the power consumption patterns of the daily load vector data set that has undergone time series analysis, distinguish between basic power consumption patterns and special power consumption patterns, and thus divide the daily load vector data set into a basic power consumption pattern data set and a special power consumption pattern data set;

[0009] Select daily load feature vectors from the basic power consumption pattern data set, use the daily load feature vectors to calculate the load level similarity between different daily load vectors in the basic power consumption pattern data set, set the load level classification threshold according to the load level similarity, and divide the basic power consumption pattern data set into multiple basic power consumption pattern sub-data sets with different load levels according to the load level classification threshold; extract the central load vector of the industrial load data in each basic power consumption pattern sub-data set based on the non-parametric regression method; construct an abnormal load data domain according to the central load vector, and filter out abnormal load data from the abnormal load data domain;

[0010] Load level mapping is introduced to map different load levels to approximate load levels, and then the conventional weighted mean method and the weighted mean method based on load level mapping are mixed to correct the abnormal load data.

[0011] Furthermore, the time series analysis of the daily load vector in the daily load vector data set is performed, including:

[0012] The similarity of electricity consumption patterns of two groups of daily load vectors in the daily load vector dataset is calculated as follows:

[0013] For two sets of daily load vectors [x i1 , x i2 , …, x in ] and [x k1 , x k2 , …, x kn ], where x ij and

[0014] x kj (j=1, 2, ..., n) represents the load at the jth moment on the i-th day and the k-th day, respectively.

[0015]

[0016] In the formula, w ik ∈[0,1],w ik The closer it is to 1, the more similar the electricity consumption patterns of the two days are.

[0017] Furthermore, the statistical fuzzy matrix method is used to cluster the power consumption patterns of the daily load vector data set after time series analysis, distinguishing the basic power consumption pattern from the special power consumption pattern, thereby dividing the daily load vector data set into the basic power consumption pattern data set and the special power consumption pattern data set, including:

[0018] According to the similarity of the power consumption patterns of two groups of daily load vectors in the daily load vector data set, the equivalent fuzzy matrix coefficients are calculated as follows:

[0019] The "°" symbol is introduced to represent the self-multiplication operation of the fuzzy relation matrix,

[0020] W s =W°W°…°W

[0021] (W°W) ij =max k {min{W ik , W jk}}

[0022] Where W s is the equivalent fuzzy matrix, k is the load of the kth comparison day; by calculating the above formula, we can obtain [W1,W2,…W 29 ,W 30 ]Equivalent fuzzy matrix coefficient set;

[0023] Use the k-nearest neighbor algorithm to obtain [W1,W2,…W 29 ,W 30 ]The equivalent fuzzy matrix coefficient set is clustered to automatically find the classification threshold, and the daily load vector data set is divided into a basic power consumption mode data set and a special power consumption mode data set according to the classification threshold.

[0024] Furthermore, a daily load feature vector is selected from the basic power consumption pattern data set, and the load level similarity between different daily load vectors in the basic power consumption pattern data set is calculated using the daily load feature vector. A load level classification threshold is set according to the load level similarity, and the basic power consumption pattern data set is divided into a plurality of basic power consumption pattern sub-data sets with different load levels according to the load level classification threshold, including:

[0025] Select the maximum load P in the daily load from the basic power consumption pattern data set max , minimum load P min , median load P median and the standard deviation σ relative to the median median The daily load characteristic vector P f :

[0026] P f =[P max P min Pmedian σ median ]

[0027] Using the daily load characteristic vector P f The load level similarity between different daily load vectors in the basic electricity consumption pattern dataset is calculated using the following formula:

[0028]

[0029] In the formula, s ij Indicates the similarity between the load levels on the i-th day and the j-th day; P f , i and P f , j represent the load characteristic vectors of the i-th day and the j-th day respectively;

[0030] When the load characteristic vectors of two days are the same, s ij =∞; through the k-nearest neighbor algorithm, the load level classification threshold is set, and the basic power consumption mode data set is divided into a plurality of basic power consumption mode sub-data sets with different load levels according to the load level classification threshold.

[0031] Furthermore, based on the non-parametric regression method, the central load vector is extracted from the industrial load data in each basic electricity consumption mode sub-dataset, including:

[0032] At time t, the central load nonparametric regression model is:

[0033]

[0034] Where N is the number of observations; w i Represents a point (t,y i )’s weight; x i represents the observed value of the load at time t; Represents an estimate of the center load.

[0035] In nonparametric regression analysis, the point (t,y i )’s weight w i The calculation formula is as follows:

[0036]

[0037] In the formula, K h (n) represents the kernel density function with scale parameter h; n represents the center load position mark, and n=N+1; n i represents the load observation value position mark, and n i ∈[1,N];

[0038] The density kernel function K(·) and scale parameter h used in the above formula play a decisive role in the value of the weight coefficient. The kernel function formula is:

[0039]

[0040] Where K(n) is the probability density function of the standard normal distribution. For the scale parameter h, the larger its value is, the more neighboring data information is considered, and the more the central load vector obtained can reflect the data distribution, so h is set to the number of load vectors participating in the extraction of the central load vector in the basic power consumption mode sub-dataset.

[0041] Furthermore, an abnormal load data domain is constructed according to the central load vector, and abnormal load data is screened out from the abnormal load data domain, including:

[0042] For the confidence level α, the distribution F of the parameter θ in the abnormal load data domain at the α level is defined as:

[0043]

[0044] In the formula, It means that the distribution function F(θ) is The quantile below ; It means that the distribution function F(θ) is at the confidence level At time t, the relationship between the observed load data and the central load is as follows:

[0045]

[0046] Where N is the number of load observations; i is the sequence of load observations; y is i represents the observed load, Indicates the central load; ε i represents the deviation between the central load and the load observation value; assuming that the deviation term ε i are independent and identically distributed, and have a mean of 0 and a variance of σ 2 The estimated value of is:

[0047]

[0048] The abnormal load data domain is:

[0049]

[0050] In the formula, Represents the standard normal distribution Quantile.

[0051] Furthermore, load level mapping is introduced to map different load levels to approximate load levels, and then the conventional weighted mean method and the weighted mean method based on load level mapping are mixed to correct the abnormal load data, including:

[0052] Load level mapping is introduced to map different load levels to approximate load levels. The mapping relationship is:

[0053]

[0054] In the formula, x t-1 Indicates the load value at the moment before the moment to be corrected; x' t and x' t-1 Respectively represent the load values ​​at time t and t-1 on the reference day; f(x' t ) represents the mapped load value; when it is necessary to correct the abnormal load value for more than two consecutive moments, start from the second value to be corrected x t-1 and x' t-1 It is still the load value at the first moment to be corrected;

[0055] The load data at the same time on days with similar load levels are selected as references, and the weighted mean is used as the correction value. The correction formula is as follows:

[0056]

[0057] Where N d , t represents the abnormal load correction value; M represents the number of selected nearby days; f(y d-i , t ) and γ i They represent the influence weights of the load observation value and the value to be corrected at time t on day di respectively;

[0058] When a day with similar load level is found on the day to be corrected, the conventional weighted mean method is used, and the correction formula is as follows:

[0059]

[0060] Where M represents the number of selected nearby days; y d-i , t represents the load observation value at time t on day di; γ i Indicates the influence weight of the value to be corrected.

[0061] Based on the same inventive concept, the present invention provides an industrial load abnormal value identification and processing system based on non-parametric regression, comprising:

[0062] A data acquisition module is used to collect industrial load data and construct a daily load vector data set based on the collected industrial load data;

[0063] The power consumption mode classification module is used to perform time series analysis on the daily load vectors in the daily load vector data set; the statistical fuzzy matrix method is used to cluster the power consumption modes of the daily load vector data set that has undergone time series analysis, and distinguish between basic power consumption modes and special power consumption modes, thereby dividing the daily load vector data set into a basic power consumption mode data set and a special power consumption mode data set;

[0064] The load anomaly value identification module is used to select daily load feature vectors from the basic power consumption mode data set, use the daily load feature vectors to calculate the load level similarity between different daily load vectors in the basic power consumption mode data set, set the load level classification threshold according to the load level similarity, and divide the basic power consumption mode data set into multiple basic power consumption mode sub-data sets with different load levels according to the load level classification threshold; based on the non-parametric regression method, extract the central load vector of the industrial load data in each basic power consumption mode sub-data set; construct an abnormal load data domain according to the central load vector, and filter out the abnormal load data from the abnormal load data domain;

[0065] The outlier correction module is used to introduce load level mapping, map different load levels to approximate load levels, and then use a combination of conventional weighted mean method and weighted mean method based on load level mapping to correct abnormal load data.

[0066] Based on the same inventive concept, the present invention provides an industrial load anomaly value identification and processing device based on non-parametric regression, characterized in that it includes a processor and a memory, wherein computer instructions are stored in the memory, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the steps of the above-mentioned industrial load anomaly value identification and processing method based on non-parametric regression.

[0067] Based on the same inventive concept, a computer-readable storage medium of the present invention is characterized in that the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for identifying and processing industrial load anomaly values ​​based on non-parametric regression.

[0068] Beneficial effects: Compared with the prior art, the significant technical effects of the present invention are as follows:

[0069] The present invention designs an efficient industrial load anomaly value identification and processing system. The system first classifies the power consumption mode based on the statistical fuzzy matrix, and then classifies the load level of the basic power consumption mode and extracts the central load of the corresponding load level. Since the central load is extracted under the same power consumption mode, the traditional clustering method solves the problem of the central load vector being offset due to the influence of the load vectors of other data sets, and more accurately fits the central load of the corresponding load level, thereby estimating the load anomaly value and correcting it. The industrial load anomaly value identification and processing system based on non-parametric regression, which is constructed by three modules, can simply and effectively realize the identification and correction of anomalies for daily load data information of any time span. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a flow chart of a method for identifying and processing abnormal values ​​of industrial loads based on non-parametric regression disclosed in an embodiment of the present invention;

[0071] Figure 2 This is a schematic diagram of a monthly load center vector of a chemical plant disclosed in an embodiment of the present invention;

[0072] Figure 3 It is a schematic diagram of a monthly load curve and identification result of a chemical plant disclosed in an embodiment of the present invention;

[0073] Figure 4 It is a schematic diagram of a monthly load correction curve of a chemical plant disclosed in an embodiment of the present invention;

[0074] Figure 5 It is a structural schematic diagram of an industrial load abnormal value identification and processing system based on non-parametric regression disclosed in an embodiment of the present invention;

[0075] Figure 6 The invention discloses a structural diagram of an industrial load abnormal value identification and processing device based on non-parametric regression. DETAILED DESCRIPTION

[0076] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It will be understood by those skilled in the art that the purposes and advantages that can be achieved with the present invention are not limited to the specific description of the above beneficial effects, and the above and other purposes that can be achieved by the present invention will be more clearly understood based on the following detailed description.

[0077] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed in the present invention can be implemented in hardware, software or a combination of the two. Whether it is implemented in hardware or software depends on the specific application and design of the technical solution. Professional and technical personnel can use different methods to implement the functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0078] Reference to "embodiments" in the present invention means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0079] Example 1

[0080] See also Figure 1 , Figure 1 1 is a flow chart of a method for identifying and processing abnormal values ​​of industrial loads based on non-parametric regression disclosed in an embodiment of the present invention. Figure 1 The industrial load abnormal value identification and processing method described is applied to the power system, such as for load abnormal value identification, etc., which is not limited in the embodiments of the present invention. Figure 1 As shown, the method for identifying and processing abnormal values ​​of industrial load based on non-parametric regression may include the following operations:

[0081] S1. Collect industrial load data and construct a daily load vector data set based on the collected industrial load data.

[0082] (1) In this example, the actual power load data of a chemical plant in September is used. The industrial load value of the month is collected with 15 minutes as a sampling point, and the sampled data is divided into 30 load data sets as follows:

[0083] {[x 1,1 ,x 1,1 ,…x 1,95 ,x 1,96 ],[x 2,1 ,x 2,1 ,…x 2,95 ,x 2,96 ],…,[x 30,1 ,x 30,1 ,…x 30,95 ,x 30,96 ]}

[0084] S2. Perform time series analysis on the daily load vectors in the daily load vector data set; use the statistical fuzzy matrix method to cluster the power consumption patterns of the daily load vector data set that has undergone time series analysis, distinguish between basic power consumption patterns and special power consumption patterns, and thus divide the daily load vector data set into a basic power consumption pattern data set and a special power consumption pattern data set. The specific steps are as follows:

[0085] S2.1. Perform time series analysis on the daily load vectors in the daily load vector data set.

[0086] The similarity of electricity consumption patterns of two groups of daily load vectors in the daily load vector dataset is calculated as follows:

[0087] For two sets of daily load vectors [x i1 , x i2 , …, x in ] and [x k1 , x k2 , …, x kn ], where x ij and

[0088] x kj (j=1, 2, ..., n) represents the load at the jth moment on the i-th day and the k-th day, respectively.

[0089]

[0090] In the formula, w ik ∈[0,1],w ik The closer it is to 1, the more similar the electricity consumption patterns of the two days are.

[0091] S2.2. Use the statistical fuzzy matrix method to cluster the power consumption patterns of the daily load vector data set that has undergone time series analysis, distinguish between basic power consumption patterns and special power consumption patterns, and thus divide the daily load vector data set into a basic power consumption pattern data set and a special power consumption pattern data set.

[0092] According to the similarity w of the electricity consumption patterns of two groups of daily load vectors in the daily load vector dataset ik , calculate the equivalent fuzzy matrix coefficients as follows:

[0093] The "°" symbol is introduced to represent the self-multiplication operation of the fuzzy relation matrix,

[0094] W s =W°W°…°W

[0095] (W°W) ij =max k {min{W ik , W jk}}

[0096] Where Ws is the equivalent fuzzy matrix, k is the load of the kth comparison day;

[0097] By calculating the above formula, we can obtain [W1,W2,…W 29 ,W 30 ]Equivalent fuzzy matrix coefficient set.

[0098] Use the k-nearest neighbor algorithm to obtain [W1,W2,…W 29 ,W 30 ]The equivalent fuzzy matrix coefficient set is clustered to automatically find the classification threshold, and the daily load vector data set is divided into a basic power consumption mode data set and a special power consumption mode data set according to the classification threshold.

[0099] In this embodiment, after the equivalent fuzzy matrix is ​​multiplied by itself s-1 times, the equivalent fuzzy matrix W s-1 They have similar transitivity and the following relationship:

[0100] W s-1 =W s =W s+1 =W s+2 =…

[0101] W s-1 The elements in are called the approximate coefficients of the transferable power consumption pattern similarity, which reflect the similarity of the power consumption patterns of different load vectors. s From the elements in , it can be found that the equivalent fuzzy relationship coefficient between the load vectors belonging to the basic power consumption mode is getting closer to 1, while the equivalent fuzzy relationship coefficient between the load vectors belonging to the special power consumption mode and the basic power consumption mode is quite different from the former. By selecting a classification threshold between the two, the load vectors are divided into the basic power consumption mode or the special power consumption mode.

[0102] S3. Select daily load characteristic vectors from the basic electricity consumption pattern data set divided in step S2, and use the daily load characteristic vectors to calculate the load level similarity between different daily load vectors in the basic electricity consumption pattern data set; set the load level classification threshold according to the load level similarity, and divide the basic electricity consumption pattern data set into multiple basic electricity consumption pattern sub-data sets with different load levels according to the load level classification threshold; based on the non-parametric regression method, extract the central load vectors of the industrial load data in different basic electricity consumption pattern sub-data sets respectively; construct an abnormal load data domain according to the central load vector, and filter out the abnormal load data from the abnormal load data domain.

[0103] The specific process of step S3 is as follows:

[0104] S3.1. Select a daily load characteristic vector from the basic electricity consumption pattern data set divided in step S2.

[0105] Select the maximum load P in the daily load from the basic power consumption pattern data set max , minimum load P min , median load P median and the standard deviation σ relative to the median median The daily load characteristic vector P f :

[0106] P f =[P max P min P median σ median ]

[0107] S3.2. Use the daily load feature vector to calculate the load level similarity between different daily load vectors in the basic electricity consumption pattern dataset.

[0108] In this embodiment, the daily load characteristic vector P is used f The load level similarity between different daily load vectors in the basic electricity consumption pattern dataset is calculated using the following formula:

[0109]

[0110] In the formula, s ij Indicates the similarity between the load levels on the i-th day and the j-th day; P f , i and P f , j represent the load characteristic vectors of the i-th day and the j-th day respectively;

[0111] S3.3. Use the daily load feature vector to calculate the load level similarity between different daily load vectors in the basic electricity consumption pattern dataset.

[0112] When the load characteristic vectors of two days are the same, s ij =∞; through the k-nearest neighbor algorithm, an appropriate load level classification threshold is selected, and the basic power consumption mode data set is divided into multiple basic power consumption mode sub-data sets with different load levels according to the load level classification threshold. In this example, four types of load levels are extracted.

[0113] S3.4. Based on the non-parametric regression method, the central load vectors of industrial load data in different basic electricity consumption mode sub-datasets are extracted respectively.

[0114] At time t, the central load nonparametric regression model is:

[0115]

[0116] Where N is the number of observations; w i Represents a point (t,yi )’s weight; x i represents the observed value of the load at time t; Represents an estimate of the center load.

[0117] In nonparametric regression analysis, the point (t,y i )’s weight w i The calculation formula is as follows:

[0118]

[0119] In the formula, K h (N i ) represents the kernel density function with scale parameter h; N represents the center load position mark; n i represents the load observation value position mark, and n i ∈[1,N].

[0120] The density kernel function K(·) and scale parameter h used in the above formula play a decisive role in the value of the weight coefficient. The kernel function formula is:

[0121]

[0122] Where K(n) is the probability density function of the standard normal distribution. For the scale parameter h, the larger its value is, the more neighboring data information is considered, and the more the central load vector obtained can reflect the data distribution, so h is set to the number of load vectors participating in the extraction of the central load vector in the basic power consumption mode sub-dataset.

[0123] In this embodiment, h is set to 4, which is the number of load vectors participating in the central vector extraction in the basic power consumption mode data set. The extracted central load vector is as follows: Figure 2 . Figure 2 In the figure, the load curves under various load levels are shown, and the thick line represents the extracted central load vector.

[0124] S3.5, construct an abnormal load data domain according to the central load vector calculated in step S3.4, and filter out abnormal load data from the abnormal load data domain. The details are as follows:

[0125] S3.5.1. Construct an abnormal load data domain based on the central load vector calculated in step S3.4.

[0126] Outlier identification is to confirm whether certain observations are distributed in the abnormal data domain.

[0127] In this embodiment, for the confidence level α, the abnormal load data domain of the distribution F of the parameter θ at the α level is defined as:

[0128]

[0129] In the formula, It means that the distribution function F(θ) is at the confidence level The quantile below ; It means that the distribution function F(θ) is at the confidence level At time t, the relationship between the observed load data and the central load is as follows:

[0130]

[0131] Where N is the number of load observations; i is the sequence of load observations; y is i represents the observed load, Indicates the central load; ε i represents the deviation between the central load and the load observation value; assuming the deviation ε i are independent and identically distributed, and have a mean of 0 and a variance of σ 2 The estimated value of is:

[0132]

[0133] In the formula, is the assumed deviation ε i are independent and identically distributed; N represents the number of observations;

[0134] The abnormal load data domain is:

[0135]

[0136] In the formula, Represents the standard normal distribution Quantile.

[0137] S3.5.2. Filter out abnormal load data from the abnormal load data domain and identify abnormal load values.

[0138] The specific process is as follows:

[0139] S3.5.2.1. For the basic electricity consumption pattern data set, select the corresponding abnormal data domain to identify load anomaly values ​​according to the load level of the daily load data subset.

[0140] S3.5.2.2. For special power consumption pattern data sets, load level matching is required first. For the "peak-valley-peak" power consumption pattern, according to the ratio of high and low load periods, select the feature vector formed by high load or low load to match the load level, and use the abnormal data domain of the corresponding load level to identify load anomalies. If the matching fails, go to step S3.5.2.3.

[0141] S3.5.2.3. For Et loads that have not been identified as load anomalies in the special power consumption mode data set, the maximum upper bound and minimum lower bound of all abnormal data domains are used to identify load anomalies in the load data. Figure 3 . Figure 3 The brown curve in the middle is the load curve for the month. The 4th, 9th, 18th and 24th are turning points of the load level. The electricity consumption patterns on the above dates are significantly different from those on other dates, so they are identified as special electricity consumption patterns.

[0142] S4. Introduce load level mapping, map different load levels to approximate load levels, and then use a combination of conventional weighted mean method and weighted mean method based on load level mapping to correct abnormal load data. The details are as follows:

[0143] S4.1. Introduce load level mapping to map different load levels to approximate load levels.

[0144] In this embodiment, load level mapping is introduced to map different load levels to approximate load levels. The mapping relationship is:

[0145]

[0146] In the formula, x t-1 Indicates the load value at the moment before the moment to be corrected; x t ′ and x t-1 ′ respectively represent the load values ​​at time t and time t-1 on the reference day; f(x' t ) represents the mapped load value; when it is necessary to correct the load abnormality values ​​for more than two consecutive moments, in order to reduce the influence of the correction error, x t-1 and x t-1 ′ is still the load value at the first moment to be corrected.

[0147] S4.2. The conventional weighted mean method and the weighted mean method based on load level mapping are used in combination to correct abnormal load data.

[0148] When no day with similar load level is found for the day to be corrected, the weighted mean method based on load level mapping is used for correction, as follows:

[0149] The load data at the same time on days with similar load levels are selected as references, and the weighted mean is used as the correction value. The correction formula is as follows:

[0150]

[0151] Where N d , t represents the abnormal load correction value; M represents the number of selected nearby days; f(yd-i , t ) represents the load observation value at time t on day di; γ i Indicates the influence weight of the value to be corrected.

[0152] When a day with similar load level can be found for the day to be corrected, the conventional weighted mean method is used, and the correction formula is as follows:

[0153]

[0154] Where M represents the number of selected nearby days; y d-i , t represents the load observation value at time t on day di, γ i Indicates the influence weight of the value to be corrected. Figure 4 .

[0155] Example 2

[0156] See also Figure 5 , Figure 5 The present invention discloses a structural diagram of an industrial load abnormal value identification and processing system based on non-parametric regression, which can realize load abnormal value identification, and specifically includes:

[0157] A data acquisition module is used to collect industrial load data and construct a daily load vector data set based on the collected industrial load data;

[0158] The power consumption mode classification module is used to perform time series analysis on the daily load vectors in the daily load vector data set; the statistical fuzzy matrix method is used to cluster the power consumption modes of the daily load vector data set that has undergone time series analysis, and distinguish between basic power consumption modes and special power consumption modes, thereby dividing the daily load vector data set into a basic power consumption mode data set and a special power consumption mode data set;

[0159] The load anomaly value identification module is used to select daily load feature vectors from the basic power consumption mode data set, use the daily load feature vectors to calculate the load level similarity between different daily load vectors in the basic power consumption mode data set, set the load level classification threshold according to the load level similarity, and divide the basic power consumption mode data set into multiple basic power consumption mode sub-data sets with different load levels according to the load level classification threshold; based on the non-parametric regression method, extract the central load vector of the industrial load data in each basic power consumption mode sub-data set; construct an abnormal load data domain according to the central load vector, and filter out the abnormal load data from the abnormal load data domain;

[0160] The outlier correction module is used to introduce load level mapping, map different load levels to approximate load levels, and then use a combination of conventional weighted mean method and weighted mean method based on load level mapping to correct abnormal load data.

[0161] In an optional embodiment, the method for identifying and processing industrial load anomaly values ​​based on non-parametric regression includes: a) collecting industrial load data and constructing a daily load vector data set based on the collected industrial load data; b) calculating the power consumption mode classification threshold, distinguishing between basic power consumption modes and special power consumption modes, thereby dividing the daily load vector data set into a basic power consumption mode data set and a special power consumption mode data set; c) extracting the daily load characteristic vector of the basic power consumption mode, calculating the load similarity, classifying the load levels according to the similarity, extracting the central load vector for different types of power loads, calculating the data abnormal load domain, and screening out abnormal loads; d) using the conventional weighted mean method and the weighted mean method based on load level mapping to correct the abnormal values.

[0162] Example 3

[0163] See also Figure 6 , Figure 6 1 is a schematic diagram of a structure of an industrial load abnormal value identification and processing device based on non-parametric regression disclosed in an embodiment of the present invention. Figure 6 The described device can be applied to a power system, such as for identifying abnormal load values, etc., which is not limited in the embodiments of the present invention.

[0164] like Figure 6 As shown, the device may include a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the steps of the method described in the above embodiment and can achieve the technical effect consistent with the above method.

[0165] The memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the memory may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). A program / utility having a set (at least one) of program modules may be stored in, for example, the memory, such program modules including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment. The program modules typically perform the functions and / or methods in the embodiments described herein.

[0166] The processor executes various functional applications and data processing by running the program stored in the memory, such as implementing the method provided in the first embodiment of the present invention.

[0167] Example 4

[0168] Embodiment 4 of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in the above embodiment are implemented and the technical effect consistent with the above method can be achieved.

[0169] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, a device or a device or used in combination with it.

[0170] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0171] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0172] Computer program code for performing the operation of the present invention may be written in one or more programming languages ​​or combinations thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0173] Of course, the storage medium containing computer executable instructions provided by an embodiment of the present invention, whose computer executable instructions are not limited to the above method operations, can also execute related operations in the method provided by any embodiment of the present invention.

[0174] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression, characterized in that: include: Collect industrial load data and construct a daily load vector data set based on the collected industrial load data; Perform time series analysis on the daily load vectors in the daily load vector data set; use the statistical fuzzy matrix method to cluster the power consumption patterns of the daily load vector data set that has undergone time series analysis, distinguish between basic power consumption patterns and special power consumption patterns, and thus divide the daily load vector data set into a basic power consumption pattern data set and a special power consumption pattern data set; Select daily load feature vectors from the basic electricity consumption pattern data set, use the daily load feature vectors to calculate the load level similarity between different daily load vectors in the basic electricity consumption pattern data set, set the load level classification threshold according to the load level similarity, and divide the basic electricity consumption pattern data set into multiple basic electricity consumption pattern sub-data sets with different load levels according to the load level classification threshold; extract the central load vectors of industrial load data in different basic electricity consumption pattern sub-data sets based on the non-parametric regression method; construct an abnormal load data domain according to the central load vector, and filter out abnormal load data from the abnormal load data domain; Load level mapping is introduced to map different load levels to approximate load levels, and then the conventional weighted mean method and the weighted mean method based on load level mapping are mixed to correct the abnormal load data.

2. The method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression according to claim 1 is characterized in that: Perform time series analysis on the daily load vectors in the daily load vector data set, including: The similarity of electricity consumption patterns of two groups of daily load vectors in the daily load vector dataset is calculated as follows: For two sets of daily load vectors [x i1 , x i2 , …, x in ] and [x k1 , x k2 , …, x kn ], where x ij and x kj (j=1, 2, ..., n) represents the load at the jth moment on the i-th day and the k-th day, respectively. In the formula, w ik ∈[0,1],w ik The closer it is to 1, the more similar the electricity consumption patterns of the two days are.

3. The method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression according to claim 2 is characterized in that: The statistical fuzzy matrix method is used to cluster the power consumption patterns of the daily load vector data set after time series analysis, distinguishing the basic power consumption pattern from the special power consumption pattern, so that the daily load vector data set is divided into the basic power consumption pattern data set and the special power consumption pattern data set, including: According to the similarity of the power consumption patterns of two groups of daily load vectors in the daily load vector data set, the equivalent fuzzy matrix coefficients are calculated as follows: Introduction The symbol represents the self-multiplication operation of the fuzzy relation matrix. W s =W°E°…°W Where W s is the equivalent fuzzy matrix, k is the load of the kth comparison day; by calculating the above formula, we can obtain [W1,W2,…W 29 ,W 30 ]Equivalent fuzzy matrix coefficient set; Use the k-nearest neighbor algorithm to obtain [W1,W2,…W 29 ,W 30 ]The equivalent fuzzy matrix coefficient set is clustered to automatically find the classification threshold, and the daily load vector data set is divided into a basic power consumption mode data set and a special power consumption mode data set according to the classification threshold.

4. The method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression according to claim 1 is characterized in that: Select daily load feature vectors from the basic power consumption pattern data set, use the daily load feature vectors to calculate the load level similarity between different daily load vectors in the basic power consumption pattern data set, set the load level classification threshold according to the load level similarity, and divide the basic power consumption pattern data set into multiple basic power consumption pattern sub-data sets with different load levels according to the load level classification threshold, including: Select the maximum load P in the daily load from the basic power consumption pattern data set max , minimum load P min , median load P median and the standard deviation σ relative to the median median The daily load characteristic vector P f : P f =[P max P min P median σ median ] Using the daily load characteristic vector P f The load level similarity between different daily load vectors in the basic electricity consumption pattern dataset is calculated using the following formula: In the formula, s ij Indicates the similarity between the load levels on the i-th day and the j-th day; P f , i and P f , j represent the load characteristic vectors of the i-th day and the j-th day respectively; When the load characteristic vectors of two days are the same, s ij =∞; through the k-nearest neighbor algorithm, the load level classification threshold is set, and the basic power consumption mode data set is divided into a plurality of basic power consumption mode sub-data sets with different load levels according to the load level classification threshold.

5. The method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression according to claim 1 is characterized in that: Based on the non-parametric regression method, the central load vectors of industrial load data in different basic electricity consumption mode sub-datasets are extracted respectively, including: At time t, the central load nonparametric regression model is: Where N is the number of observations; w i Represents a point (t,y i )’s weight; x i represents the observed value of the load at time t; Represents an estimate of the center load. In nonparametric regression analysis, the point (t,y i )’s weight w i The calculation formula is as follows: In the formula, K h (N i ) represents the kernel density function with scale parameter h; N represents the center load position mark; n i represents the load observation value position mark, and n i ∈[1,N]; The density kernel function K(·) and scale parameter h used in the above formula play a decisive role in the value of the weight coefficient. The kernel function formula is: Where K(n) is the probability density function of the standard normal distribution. For the scale parameter h, the larger its value is, the more neighboring data information is considered, and the more the central load vector obtained can reflect the data distribution. h is set to the number of load vectors participating in the extraction of the central load vector in the basic power consumption mode sub-dataset.

6. The method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression according to claim 1 is characterized in that: An abnormal load data domain is constructed according to the central load vector, and abnormal load data is screened out from the abnormal load data domain, including: For the confidence level α, the distribution F of the parameter θ in the abnormal load data domain at the α level is defined as: In the formula, It means that the distribution function F(θ) is at the confidence level The quantile below ; It means that the distribution function F(θ) is at the confidence level At time t, the relationship between the observed load data and the central load is as follows: Where N is the number of load observations; i is the sequence of load observations; y is i represents the observed load, Indicates the central load; ε i represents the deviation between the central load and the load observation value; assuming the deviation ε i are independent and identically distributed, and have a mean of 0 and a variance of σ 2 The estimated value of is: In the formula, is the assumed deviation ε i are independent and identically distributed; N represents the number of observations; The abnormal load data domain is: In the formula, Represents the standard normal distribution Quantile.

7. The method for identifying and processing abnormal values ​​of industrial load based on nonparametric regression according to claim 1 is characterized in that: Load level mapping is introduced to map different load levels to approximate load levels, and then the conventional weighted mean method and the weighted mean method based on load level mapping are mixed to correct the abnormal load data, including: Load level mapping is introduced to map different load levels to approximate load levels. The mapping relationship is: In the formula, x t-1 Indicates the load value at the moment before the moment to be corrected; x' t and x' t-1 Respectively represent the load values ​​at time t and t-1 on the reference day; f(x' t ) represents the mapped load value; when it is necessary to correct the abnormal load value for more than two consecutive moments, start from the second value to be corrected x t-1 and x' t-1 It is still the load value at the first moment to be corrected; The load data at the same time on days with similar load levels are selected as references, and the weighted mean is used as the correction value. The correction formula is as follows: Where N d , t represents the abnormal load correction value; M represents the number of selected nearby days; f(y d-i , t ) represents the load observation value at time t on day di; γ i Indicates the impact weight of the value to be corrected; When a day with similar load level is found on the day to be corrected, the conventional weighted mean method is used, and the correction formula is as follows: Where M represents the number of selected nearby days; y d-i , t represents the load observation value at time t on day di; γ i Indicates the influence weight of the value to be corrected.

8. A system for identifying and processing abnormal values ​​of industrial loads based on nonparametric regression, characterized in that: include: A data acquisition module is used to collect industrial load data and construct a daily load vector data set based on the collected industrial load data; The power consumption mode classification module is used to perform time series analysis on the daily load vector in the daily load vector data set; The statistical fuzzy matrix method is used to cluster the power consumption patterns of the daily load vector data set after time series analysis, distinguishing the basic power consumption pattern from the special power consumption pattern, so as to divide the daily load vector data set into the basic power consumption pattern data set and the special power consumption pattern data set; The load anomaly value identification module is used to select daily load feature vectors from the basic power consumption mode data set, use the daily load feature vectors to calculate the load level similarity between different daily load vectors in the basic power consumption mode data set, set the load level classification threshold according to the load level similarity, and divide the basic power consumption mode data set into multiple basic power consumption mode sub-data sets with different load levels according to the load level classification threshold; based on the non-parametric regression method, extract the central load vector of the industrial load data in different basic power consumption mode sub-data sets; construct an abnormal load data domain according to the central load vector, and filter out the abnormal load data from the abnormal load data domain; The outlier correction module is used to introduce load level mapping, map different load levels to approximate load levels, and then use a combination of conventional weighted mean method and weighted mean method based on load level mapping to correct abnormal load data.

9. An industrial load abnormal value identification and processing device based on non-parametric regression, characterized in that: The electronic device comprises a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the steps of the method for identifying and processing industrial load anomaly values ​​based on non-parametric regression as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for identifying and processing industrial load anomaly values ​​based on non-parametric regression as claimed in any one of claims 1 to 7.

Images