Missing value filling method and device, electronic equipment and storage medium

Abstract

The invention discloses a missing value filling method and device, electronic equipment and a storage medium. The method comprises the steps of determining a first data set of a target object; determining a time interval corresponding to the acquisition moment of each piece of first data according to the acquisition moment of each piece of first data; and filling the first data set according to the local data features of the first data set and the time interval to generate a target data set. According to the technical scheme, the first data set of the target object is determined; determining a time interval corresponding to the acquisition moment of each piece of first data according to the acquisition moment of each piece of first data; and filling the first data set according to the local data feature and the time interval of the first data set, and generating the target data set, thereby ensuring that the generated target data set can have the time sequence feature of the first data set, further enabling the target data set to better conform to the real situation, and improving the accuracy of the data set.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a missing value filling method, device, electronic device and storage medium. Background Art

[0002] With the rapid development of data processing technology, there are many methods for processing outliers and null values ​​in arrays, such as direct deletion, mean replacement, median replacement, etc.

[0003] However, there are many different types of existing processing methods. When using them, it is difficult to determine the processing method that is more suitable for the current situation, and the processing results are often difficult to ensure that they have the data characteristics of the original data, which leads to certain errors in the final processing results. Summary of the invention

[0004] The present invention provides a missing value filling method, device, electronic device and storage medium to solve the problem that when traditional methods fill time series data, the influence of time series characteristics on data cannot be considered during filling.

[0005] According to one aspect of the present invention, a missing value filling method is provided, the method comprising:

[0006] Determine a first data group of a target object, the first data group includes at least one first data, the first data includes at least one data element, the collection time of each first data is different, the collection order of each first data in the first data group is continuous, the target object is a wind turbine device, at least one data element in at least one first data in the first data group is a null value, and the number of elements is not greater than a preset number of null values;

[0007] Determine, according to the collection time of each first data, the time interval corresponding to the collection time of each first data;

[0008] Randomly extracting a preset number of data from the first data group, and determining local data features based on the extracted data;

[0009] The first data group is filled according to the local data features and time interval of the first data group to generate a target data group.

[0010] According to another aspect of the present invention, there is provided a missing value filling device, the device comprising:

[0011] A data acquisition module is used to determine a first data group of a target object, wherein the first data group includes at least one first data, the first data includes at least one data element, the acquisition time of each first data is different, the acquisition order of each first data in the first data group is continuous, the target object is a wind turbine device, at least one data element of at least one first data exists in the first data group, and the number of elements is not greater than a preset number or a null value;

[0012] A time interval determination module, used to determine the time interval corresponding to the collection moment of each first data according to the collection moment of each first data;

[0013] A feature acquisition module, used for randomly extracting a preset amount of data from the first data group, and determining local data features based on the extracted data;

[0014] The data filling module is used to fill the first data group according to the local data characteristics and time interval of the first data group to generate a target data group.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] at least one processor; and

[0017] a memory communicatively connected to at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by at least one processor, and the computer program is executed by at least one processor so that the at least one processor can execute the missing value filling method of any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the missing value filling method of any embodiment of the present invention when executed.

[0020] The technical solution of the embodiment of the present invention determines the first data group of the target object; determines the time interval corresponding to the collection time of each first data according to the collection time of each first data, so that when the first data group is subsequently filled, the time interval can be used to determine the timing relationship between each first data, thereby improving the accuracy of the final filling result; fills the first data group according to the local data characteristics and time interval of the first data group to generate a target data group, thereby ensuring that the generated target data group can have the timing characteristics of the first data group, thereby making the target data group more consistent with the actual situation and improving the accuracy of the data group; through the combined use of the above steps, it is possible to overcome the problem that the direct deletion method will cause a greater degree of information loss and the direct filling method ignores the timing.

[0021] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flowchart of a missing value filling method provided according to Embodiment 1 of the present invention;

[0024] Figure 2 is a flowchart of another missing value filling method provided according to the second embodiment of the present invention;

[0025] Figure 3 is a structural schematic diagram of a missing value filling device provided according to Embodiment 3 of the present invention;

[0026] Figure 4 It is a structural schematic diagram of an electronic device for implementing the missing value filling method of an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] Embodiment 1

[0030] Figure 1 A flowchart of a missing value filling method is provided for the first embodiment of the present invention. This embodiment is applicable to the case where the time series data generated during the operation of the wind turbine contains null values ​​and needs to be filled. The method can be executed by a missing value filling device, which can be implemented in the form of hardware and / or software. The missing value filling device can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes:

[0031] S110. Determine a first data group of a target object, wherein the first data group includes at least one first data, the first data includes at least one data element, the collection time of each first data is different, the collection order of each first data in the first data group is continuous, the target object is a wind turbine equipment, at least one data element in at least one first data in the first data group is a null value, and the number of elements is not greater than a preset number of null values.

[0032] The first data group can be a time series data group, including at least one first data, the first data is a few bits of time series data, each first data contains at least one data element, the collection time of each first data is different, the collection order of each first data in the first data group is continuous, and the first data group is the operating parameters of the wind turbine equipment.

[0033] The first data set can be identified by the following formula:

[0034] x=(x0,x1,...,x n-2 , x n-1 )∈R d×n ;

[0035] In the formula, x represents the first data set, x i represents the i-th first data in the first data group; Represents the j-th attribute value of the i-th first data in the first data group x.

[0036] For example, the first data group includes three first data, each of which includes three data elements: wind force, temperature, and rotation speed. In this case, the first data group may be in the following form:

[0037]

[0038] Wherein, null represents a null value, each column represents a first data, the first row represents the wind force, the second row represents the temperature, and the third row represents the rotation speed. In the first data group, there must be at least one first data element with a null value.

[0039] Traditional methods of dealing with missing values ​​are mainly divided into direct deletion and traditional filling methods. Direct deletion will lead to greater information loss, which is not conducive to mining the essential characteristics and missing patterns hidden in the data. Data filling based on traditional machine learning algorithms or data filling methods based on neural networks consider the filling of missing values ​​in non-time series data, but do not consider the filling of missing values ​​in time series data. Therefore, this solution is to fill each data in the time series data group, and consider the impact of the time interval between each time series data on the time series data during the filling process, as well as the data characteristics between each time series data, so that the final filled time series data can be more accurate.

[0040] S120: Determine the time interval corresponding to the collection moment of each first data according to the collection moment of each first data.

[0041] In order to ensure the accuracy of the filling results, it is necessary to consider the time intervals between the collection moments of different first data. The reason is that as the time interval increases, the timing relationship in the data will gradually weaken, while the decrease in the time interval will cause the timing relationship to become stronger, so it needs to be determined.

[0042] The time intervals between the collection moments corresponding to the first data may be the same or different.

[0043] The time tag t of each first data in the first data group is defined as:

[0044] t=(t0,t1,...,t n-2 , t n-1 );

[0045] By determining the time interval corresponding to the collection time of each first data according to the collection time of each first data, when the first data group is subsequently filled, the time interval can be used to determine the timing relationship between each first data, thereby improving the accuracy of the final filling result.

[0046] S130: randomly extract a preset amount of data from the first data group, and determine local data features based on the extracted data.

[0047] S140: Fill the first data group according to the local data features and time interval of the first data group to generate a target data group.

[0048] After the first data group is obtained, a preset number of first data are randomly extracted from the first data group, and local data features corresponding to the extracted first data are determined.

[0049] When filling data, the timing relationship between each first data in the first data group is determined according to the time interval, and then the first data group is filled in combination with the local data features corresponding to the extracted first data, so as to obtain the target data group.

[0050] By filling the first data group according to the local data characteristics and time interval of the first data group, a target data group is generated, thereby ensuring that the generated target data group can have the timing characteristics of the first data group, thereby making the target data group more consistent with the actual situation and improving the accuracy of the data group.

[0051] After the target data set is generated according to the local data features and time interval of the first data set, steps A1-A2 are also included:

[0052] Step A1, determining target data similarity between the first data group and the target data group, wherein the target data similarity is determined based on similarity between data features of the first data group and overall data features of the target data group;

[0053] Step A2: if the target data similarity is less than or equal to the preset threshold, regenerate the target data group, and determine the target data similarity of the first data group and the regenerated target data group again until the target data similarity is greater than the preset threshold.

[0054] In order to ensure the accuracy of the target data set finally generated, the target data set needs to be verified. Therefore, the similarity between the target data set and the first data set is determined by the overall data characteristics of the target data set and the first data set to determine the similarity between the generated target data set and the first data set.

[0055] If the target data similarity is greater than a preset threshold, it indicates that the generated target data group is accurate, and the target data group will no longer be adjusted.

[0056] If the target data similarity is less than or equal to the preset threshold, it indicates that the generated target data group is inaccurate. At this time, the target data group generation process will be re-executed until the target data similarity is greater than the preset threshold.

[0057] By determining the target data similarity between the first data group and the target data group, and using the target data similarity and the preset threshold for judgment, the accuracy of the filling result is guaranteed, and the filling is performed again based on the accuracy briefing of the filling result, so as to ensure the stability of the final output result.

[0058] According to the technical solution of the embodiment of the present invention, by determining the first data group of the target object; by determining the time interval corresponding to the collection time of each first data according to the collection time of each first data, so that when the first data group is subsequently filled, the time interval can be used to determine the timing relationship between each first data, thereby improving the accuracy of the final filling result; by filling the first data group according to the local data characteristics and time interval of the first data group, generating a target data group, thereby ensuring that the generated target data group can have the timing characteristics of the first data group, thereby making the target data group more consistent with the actual situation and improving the accuracy of the data group; through the combined use of the above steps, it is possible to overcome the problem that the direct deletion method will cause a greater degree of information loss and the direct filling method ignores the timing.

[0059] Embodiment 2

[0060] Figure 2 The present invention provides another flowchart of a missing value filling method. Based on the above embodiment, this embodiment further optimizes the process of filling the first data group according to the local data features and time intervals of the first data group in the above embodiment to generate the target data group. This embodiment can be combined with various optional solutions in one or more of the above embodiments. Figure 2 As shown, the missing value filling method of this embodiment may include the following steps:

[0061] S210. Determine a first data group of a target object, wherein the first data group includes at least one first data, the first data includes at least one data element, the collection time of each first data is different, the collection order of each first data in the first data group is continuous, the target object is a wind turbine equipment, at least one data element in at least one first data in the first data group is a null value, and the number of elements is not greater than a preset number of null values.

[0062] S220: Determine, according to the collection time of each first data, the time interval corresponding to the collection time of each first data.

[0063] S230: randomly extract a preset amount of data from the first data group, and determine local data features based on the extracted data.

[0064] S240: Based on the generator, fill the first data group according to the local data features and time interval of the first data group to generate a target data group.

[0065] To ensure the efficiency of filling, the present application considers using a generator to perform the operation of generating a target data group.

[0066] Accordingly, filling the first data group according to the local data features and time interval of the first data group to generate the target data group may include steps B1-B7:

[0067] Step B1, determining a mask matrix of a first data group;

[0068] Step B2: determining the time intervals between the collection moments of the first data according to the collection moments of the first data;

[0069] Step B3, randomly extracting a preset number of data from the first data group to obtain a second data group;

[0070] Step B4, determining data features of the second data group to obtain local data features;

[0071] Step B5, generating an interval matrix according to each time interval between the collection moments of each first data;

[0072] Step B6: Generate a decay factor of the historical memory vector in the generator according to the interval matrix;

[0073] Step B7: Fill the first data group based on the attenuation factor of the historical memory vector and the local data characteristics to obtain the target data group.

[0074] The mask matrix may be a matrix having the same dimension as the processing object (such as an image matrix or other data matrix), wherein each element is usually a specific value and is used to perform a specific operation on the processing object.

[0075] In this application, the mask matrix m∈R d×n Is the following formula:

[0076]

[0077] Interval matrix δ∈R d×n Is the following formula:

[0078]

[0079] The decay factor of the historical memory vector is a value used to control the degree of decay of the historical memory vector over time or in an iterative process. In a model involving historical information, in order to balance the importance of current information and past information, the decay factor is introduced to gradually reduce the influence of the historical memory vector.

[0080] The formula for the attenuation factor β of this application is:

[0081]

[0082] where w β With b β is the neural network that needs to be trained, e represents a natural constant, Represents the time interval of each dimension at the current moment.

[0083] The generator and discriminator of the present application are both located in a loop filling neural network model to achieve filling of the first data group.

[0084] Correspondingly, the update formula of the loop filling neural network model is as follows:

[0085]

[0086]

[0087] in It is the reset gate. To update the gate, is the candidate memory vector, σ is the ReLU function, and W and b are the parameters of the cyclic filling neural network model.

[0088] The goal of the generator in the loop filling neural network model is to generate data that the discriminator cannot judge whether the sample is true or false. The ultimate goal of the discriminator is to distinguish the true and false samples generated by the generator as much as possible, thus forming a generative adversarial network. Correspondingly, the optimization formula of the generative adversarial network is as follows:

[0089]

[0090] Among them, P data (x) represents the distribution of data input to the discriminator, P z (z) represents the distribution of data generated by the generator, z represents the random input vector of the generator, D(x) represents the output probability of the discriminator, that is, the possibility that the input sample is a real sample, and G(z) represents the output of the generator (new sample). The current loss function of the discriminator model is calculated by the optimization formula, and the parameters of the discriminator model are updated according to the loss function using the gradient descent method.

[0091] In an optional solution, the first stage training process of the generator may include steps C1-C5:

[0092] Step C1, determining a third data group of a target object, wherein the third data group includes at least one third data, the third data includes at least three data elements, the collection time of each third data is different, the collection order of each third data in the third data group is continuous, the target object is a wind turbine device, and each data element in each third data in the third data group is not a null value;

[0093] Step C2, randomly eliminating data in the third data group to obtain a fourth data group;

[0094] Step C3, determining the time interval between the collection moments of each fourth data group in the fourth data group;

[0095] Step C4, inputting the time intervals between the collection moments of each fourth data group and the fourth data group into the generator for data filling to obtain a fifth data group;

[0096] Step C5: adjust the generator according to the third data group and the fifth data group, regenerate the third data group and the fifth data group after the adjustment, and adjust the generator again until the number of adjustments reaches a preset number.

[0097] In order to ensure the accuracy of the training results, the generator will be trained for the first time.

[0098] During training, sample data of the target object, i.e., the third data group, is obtained, and part of the data in the third data group is eliminated by random elimination, thereby obtaining the fourth data group.

[0099] The time intervals between the collection moments of each fourth data group in the fourth data group are obtained, and are input into the generator together with the fourth data group, and the generator fills them to obtain the fifth data group.

[0100] Since the third data group is the original data and the fifth data group is the filling data, the generator can be adjusted with the help of the difference between the third data group and the fifth data group, and the above steps can be repeated, so that the data generated by the generator is closer to the original data, thereby improving the accuracy of the data generated by the generator.

[0101] S250 : Determine target data similarity between the first data group and the target data group based on the discriminator.

[0102] Optionally, the first stage of training the discriminator includes:

[0103] Based on the generator that has completed the first stage of training process, the discriminator is trained until the number of training times reaches the preset number.

[0104] After the generator completes the first stage of training, the discriminator will be trained through the generator. During training, the input and output data of the generator are sent to the discriminator, which performs similarity judgment to determine the truth or falsehood of the data input to the discriminator.

[0105] In an optional solution, after the generator and the discriminator complete the first stage of training respectively, steps D1-D2 are also included:

[0106] Step D1: Based on the discriminator of the previous stage, the generator of the current stage is trained until the generator of the current stage meets the judgment requirements of the discriminator of the previous stage or the number of training times reaches a preset number of times, thereby obtaining a generator that has completed training in the current stage;

[0107] Step D2: Based on the generator that has completed training in the current stage, train the generator of the previous stage until the number of training times reaches a preset number, obtain the discriminator that has completed training in the current stage, and start the training of the generator and discriminator of the next stage until the preset stage is reached.

[0108] After the generator and the discriminator have completed the first stage of training, the discriminator of the previous stage will be used to train the generator of the current stage. After the generator has completed the training of the current stage, the generator of the current stage will be used to train the discriminator of the previous stage to obtain the discriminator of the current stage. This process is repeated until the preset stage is reached, and all the training of the generator and the discriminator is completed.

[0109] Through repeated training of the generator and the discriminator, the accuracy between the generator and the discriminator can be improved and stabilized, thereby ensuring the accuracy and stability of the final calculation filling result.

[0110] S260: If the data similarity is less than or equal to the preset threshold, regenerate the target data group, and determine the target data similarity of the first data group and the regenerated target data group again until the target data similarity is greater than the preset threshold.

[0111] In addition to S250, the generator is also responsible for judging whether the target data set is qualified according to the data similarity. When the data similarity is less than the preset threshold, the generator will be required to regenerate the target data set to ensure the accuracy of the final result.

[0112] According to the technical solution of the embodiment of the present invention, by determining the first data group of the target object, according to the collection time of each first data, the time interval corresponding to the collection time of each first data is determined, based on the generator, according to the local data features and time interval of the first data group, the first data group is filled to generate a target data group, based on the discriminator, the target data similarity of the first data group and the target data group is determined, if the data similarity is less than or equal to the preset threshold, the target data group is regenerated, and the target data similarity of the first data group and the regenerated target data group is determined again until the target data similarity is greater than the preset threshold, thereby accurately determining the impact of the timing on the filled data, and completing data filling and data generation with the help of the generator and the discriminator, thereby improving the filling efficiency and ensuring the accuracy and stability of the final result.

[0113] Embodiment 3

[0114] Figure 3 The present invention provides a structural block diagram of a missing value filling device, which can be applied to situations where time series data generated during the operation of a wind turbine contains null values ​​and needs to be filled. The missing value filling device can be implemented in the form of hardware and / or software, and the missing value filling device can be configured in an electronic device with data processing capabilities. Figure 3 As shown, the missing value filling device of this embodiment may include: a data acquisition module 310, a time interval determination module 320, a feature acquisition module 330 and a data filling module 340. Among them:

[0115] The data acquisition module 310 is used to determine a first data group of a target object, wherein the first data group includes at least one first data, the first data includes at least one data element, the acquisition time of each first data is different, the acquisition order of each first data in the first data group is continuous, the target object is a wind turbine device, at least one data element in at least one first data in the first data group is a null value, and the number of elements is not greater than a preset number of null values;

[0116] A time interval determination module 320, configured to determine the time interval corresponding to the collection moment of each first data according to the collection moment of each first data;

[0117] A feature acquisition module 330, configured to randomly extract a preset amount of data from the first data group, and determine local data features based on the extracted data;

[0118] The data filling module 340 is used to fill the first data group according to the local data characteristics and time interval of the first data group to generate a target data group. The local data characteristics are data characteristics of data extracted after randomly extracting a preset number of data from the first data group.

[0119] Based on the above embodiment, optionally, after the data filling module 340, the device further includes:

[0120] Determining target data similarity between the first data group and the target data group, wherein the target data similarity is determined based on similarity between data features of the first data group and overall data features of the target data group;

[0121] If the data similarity is less than or equal to the preset threshold, the target data set is regenerated, and the target data similarity of the first data set and the regenerated target data set is determined again until the target data similarity is greater than the preset threshold.

[0122] Based on the above embodiment, optionally, the data filling module 340 includes:

[0123] Based on the generator, the first data group is filled according to the local data features and time interval of the first data group to generate a target data group;

[0124] Accordingly, based on the generator, the first data group is filled according to the local data features and time interval of the first data group to generate the target data group, including:

[0125] determining a mask matrix for the first data set;

[0126] Determine, according to the collection time of each first data, each time interval between the collection time of each first data;

[0127] Randomly extracting a preset number of data from the first data group to obtain a second data group;

[0128] determining data features of the second data group to obtain local data features;

[0129] Generate an interval matrix according to each time interval between the collection moments of each first data;

[0130] Generate a decay factor of the historical memory vector in the generator according to the interval matrix;

[0131] Based on the attenuation factor of the historical memory vector and the local data characteristics, the first data group is filled to obtain the target data group.

[0132] Based on the above embodiment, optionally, the first stage training process of the generator includes:

[0133] Determine a third data group of the target object, the third data group includes at least one third data, the third data includes at least three data elements, the collection time of each third data is different, the collection order of each third data in the third data group is continuous, the target object is a wind turbine device, and each data element in each third data in the third data group is not a null value;

[0134] Randomly remove data from the third data group to obtain a fourth data group;

[0135] Determine the time interval between the collection moments of each fourth data group in the fourth data group;

[0136] Inputting the fourth data group into the generator for data filling to obtain a fifth data group;

[0137] The generator is adjusted according to the third data group and the fifth data group, and the third data group and the fifth data group are regenerated after the adjustment, and the generator is adjusted again until the number of adjustments reaches a preset number of times.

[0138] Based on the above embodiment, optionally, determining the target data similarity between the first data group and the target data group includes:

[0139] Determine target data similarity between the first data group and the target data group based on the discriminator;

[0140] Correspondingly, the first stage training process of the discriminator includes:

[0141] Determine a sixth data group of the target object, the sixth data group includes at least one sixth data, the sixth data includes at least one data element, the collection time of each sixth data is different, the collection order of each sixth data in the sixth data group is continuous, the target object is a wind turbine device, at least one data element in at least one sixth data in the sixth data group is a null value, and the number of elements is not greater than a preset number of null values;

[0142] Based on the generator trained in the first stage, data is filled in the sixth data group to obtain a seventh data group;

[0143] The discriminator is readjusted through the seventh data set and the sixth data set, and the seventh data set and the sixth data set are regenerated after the adjustment, and the discriminator is adjusted again until the number of adjustments reaches a preset number of times.

[0144] On the basis of the above embodiment, optionally, the discriminator and the generator are adjusted based on a gradient descent algorithm.

[0145] Based on the above embodiment, optionally, after the generator and the discriminator complete the first stage of training respectively, the method further includes:

[0146] Based on the discriminator of the previous stage, the generator of the current stage is trained until the generator of the current stage meets the judgment requirements of the discriminator of the previous stage or the number of training times reaches the preset number of times, and the generator of the current stage is obtained.

[0147] Based on the generator that has completed training in the current stage, the generator of the previous stage is trained until the number of training times reaches the preset number, and the discriminator that has completed training in the current stage is obtained, and the training of the generator and discriminator of the next stage is started until the preset stage is reached.

[0148] The missing value filling device provided in the embodiment of the present invention can execute the missing value filling method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0149] Embodiment 4

[0150] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0151] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0152] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0153] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 11 performs the various methods and processes described above, such as a missing value filling method.

[0154] In some embodiments, the missing value filling method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the missing value filling method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the missing value filling method in any other appropriate manner (e.g., by means of firmware).

[0155] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0156] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0157] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 foregoing.

[0158] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0159] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0160] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0161] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0162] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A missing value filling method, characterized in that: include: Determine a first data group of a target object, wherein the first data group includes at least one first data, the first data includes at least one data element, the collection time of each first data is different, the collection order of each first data in the first data group is continuous, the target object is a wind turbine device, at least one data element in at least one first data in the first data group is a null value, and the number of elements is not greater than a preset number of null values; Determine, according to the collection time of each first data, the time interval corresponding to the collection time of each first data; Randomly extracting a preset number of data from the first data group, and determining local data features based on the extracted data; The first data group is filled according to the local data features and time interval of the first data group to generate a target data group.

2. The method according to claim 1, characterized in that After generating a target data group according to the local data features and time interval of the first data group, the method further includes: Determining target data similarity between the first data group and the target data group, wherein the target data similarity is determined based on similarity between data features of the first data group and overall data features of the target data group; If the target data similarity is less than or equal to a preset threshold, the target data group is regenerated, and the target data similarity of the first data group and the regenerated target data group is determined again until the target data similarity is greater than the preset threshold.

3. The method according to claim 1, characterized in that Filling the first data group according to the local data features and time interval of the first data group to generate a target data group includes: Based on the generator, the first data group is filled according to the local data features and time interval of the first data group to generate a target data group; Correspondingly, based on the generator, the first data group is filled according to the local data features and time interval of the first data group to generate a target data group, including: determining a mask matrix for the first data set; Determine, according to the collection time of each first data, each time interval between the collection time of each first data; Randomly extracting a preset number of data from the first data group to obtain a second data group; determining data features of the second data group to obtain local data features; Generate an interval matrix according to each time interval between the collection moments of each first data; Generate a decay factor of the historical memory vector in the generator according to the interval matrix; Based on the attenuation factor of the historical memory vector and the local data characteristics, the first data group is filled to obtain a target data group.

4. The method according to claim 2, characterized in that: The first stage training process of the generator includes: Determine a third data group of a target object, wherein the third data group includes at least one third data, the third data includes at least three data elements, the collection time of each third data is different, the collection order of each third data in the third data group is continuous, the target object is a wind turbine device, and each data element in each third data in the third data group is not a null value; Randomly eliminating data in the third data group to obtain a fourth data group; Determining the time interval between the collection moments of each fourth data group in the fourth data group; Inputting the time intervals between the collection moments of the fourth data groups and the fourth data groups into a generator for data filling to obtain a fifth data group; The generator is adjusted according to the third data group and the fifth data group, and the third data group and the fifth data group are regenerated after the adjustment, and the generator is adjusted again until the number of adjustments reaches a preset number of times.

5. The method according to claim 4, characterized in that Determining target data similarity between the first data group and the target data group includes: Determining target data similarity between the first data group and the target data group based on the discriminator; Correspondingly, the first stage training process of the discriminator includes: Based on the generator that has completed the first stage of training process, the discriminator is trained until the number of training times reaches the preset number.

6. The method according to claim 2, characterized in that Based on the gradient descent algorithm, the discriminator and generator are adjusted.

7. The method according to claim 5, characterized in that After the generator and discriminator complete the first stage of training, they also include: Based on the discriminator of the previous stage, the generator of the current stage is trained until the generator of the current stage meets the judgment requirements of the discriminator of the previous stage or the number of training times reaches the preset number of times, and the generator of the current stage is obtained. Based on the generator that has completed training in the current stage, the generator of the previous stage is trained until the number of training times reaches the preset number, and the discriminator that has completed training in the current stage is obtained, and the training of the generator and discriminator of the next stage is started until the preset stage is reached.

8. A missing value filling device, characterized in that: include: A data acquisition module is used to determine a first data group of a target object, wherein the first data group includes at least one first data, the first data includes at least one data element, the acquisition time of each first data is different, the acquisition order of each first data in the first data group is continuous, the target object is a wind turbine device, at least one data element in at least one first data in the first data group is a null value, and the number of elements is not greater than a preset number of null values; A time interval determination module, used to determine the time interval corresponding to the collection moment of each first data according to the collection moment of each first data; A feature acquisition module, used for randomly extracting a preset amount of data from the first data group, and determining local data features based on the extracted data; The data filling module is used to fill the first data group according to the local data characteristics and time interval of the first data group to generate a target data group.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the missing value filling method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the missing value filling method according to any one of claims 1 to 7 when executed.

Images