A method, device, equipment and medium for analyzing abnormality of measurement indicators

Abstract

The present application discloses a method, device, equipment and medium for analyzing abnormality of metering indicators. The method comprises: when an abnormality occurs in a metering indicator, obtaining archive data and operation data of each target metering terminal associated with the metering indicator at the current moment; performing fault analysis on the archive data and operation data through a preset fault classification model to obtain a fault combination corresponding to each target metering terminal; counting the frequency of the fault combination of all target metering terminals, and outputting the fault combination with the highest frequency; improving the existing technology of analyzing abnormality of indicators through expert judgment combined with on-site operation and maintenance investigation, which has the technical problems of large workload and low efficiency.

Description

Technical Field

[0001] The present application relates to the field of power grid technology, and in particular to a method, device, equipment and medium for analyzing abnormalities in measurement indicators. Background Art

[0002] In power grid metering automation systems, a large number of terminals are involved, particularly in low-voltage centralized meter reading systems, which can involve millions of meters. Metering indicators such as the online rate and meter reading rate of collected data are easily affected by external factors, leading to abnormal metering indicators.

[0003] When metering automation systems have abnormal metering indicators, it is necessary to analyze the causes by reviewing a large number of reports. Existing technologies mainly use expert judgment combined with on-site operation and maintenance inspections to analyze indicator anomalies, which results in heavy workload and low efficiency. Summary of the Invention

[0004] The present application provides a method, device, equipment and medium for analyzing abnormalities in measurement indicators, which are used to improve the existing technology of analyzing abnormalities in indicators through expert judgment combined with on-site operation and maintenance inspection, which has technical problems such as large workload and low efficiency.

[0005] In view of this, the first aspect of the present application provides a method for analyzing abnormality of measurement indicators, comprising:

[0006] When a measurement indicator is abnormal, obtain the archive data and operation data of each target measurement terminal associated with the measurement indicator at the current moment;

[0007] Performing fault analysis on the archive data and the operating data using a preset fault classification model to obtain a fault combination corresponding to each target metering terminal;

[0008] The frequencies of the fault combinations of all the target metering terminals are counted, and the fault combination with the highest frequency is output.

[0009] Optionally, the configuration process of the preset fault classification model is:

[0010] Combine the fault causes that lead to abnormal measurement indicators to obtain several fault combinations;

[0011] Determine the fault combination corresponding to the faulty metering terminal based on the fault work order data;

[0012] Constructing training samples based on the archival data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination;

[0013] The convolutional neural network is trained using the training samples to obtain the preset fault classification model.

[0014] Optionally, constructing a training sample based on the archival data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination includes:

[0015] fusing the archive data corresponding to the faulty metering terminal and the operating data on the day of the fault to obtain fused data;

[0016] One-hot encoding is performed on the fault combination corresponding to the fault metering terminal, and the obtained encoding result is used as the fault category label of the fusion data to obtain a training sample.

[0017] Optionally, the loss function of the convolutional neural network is:

[0018]

[0019] Where y p is the probability of predicting fault category j, x i is the i-th training sample, θ is the vector parameter matrix of the neuron, λ is the regularization adjustment parameter, k is the predicted fault category, and m is the number of training samples.

[0020] Optionally, the archival data includes: operator, terminal type, terminal model, terminal manufacturer, terminal installation date, terminal district office, terminal software version, terminal master station type, terminal base station type, terminal base station status, terminal power outage status, terminal work order status and terminal signal strength status.

[0021] A second aspect of the present application provides a measurement indicator abnormality analysis device, comprising:

[0022] An acquisition unit, configured to acquire, when an abnormality occurs in a metering indicator, archive data and operation data of each target metering terminal associated with the metering indicator at the current moment;

[0023] a fault analysis unit, configured to perform fault analysis on the archive data and the operation data using a preset fault classification model to obtain a fault combination corresponding to each target metering terminal;

[0024] The output unit is configured to count the frequencies of the fault combinations of all the target metering terminals and output the fault combination with the highest frequency.

[0025] Optionally, the configuration process of the preset fault classification model is:

[0026] Combine the fault causes that lead to abnormal measurement indicators to obtain several fault combinations;

[0027] Determine the fault combination corresponding to the faulty metering terminal based on the fault work order data;

[0028] Constructing training samples based on the archival data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination;

[0029] The convolutional neural network is trained using the training samples to obtain the preset fault classification model.

[0030] Optionally, the archival data includes: operator, terminal type, terminal model, terminal manufacturer, terminal installation date, terminal district office, terminal software version, terminal master station type, terminal base station type, terminal base station status, terminal power outage status, terminal work order status and terminal signal strength status.

[0031] A third aspect of the present application provides a measurement indicator anomaly analysis device, the device comprising a processor and a memory;

[0032] The memory is used to store program code and transmit the program code to the processor;

[0033] The processor is used to execute the measurement indicator anomaly analysis method described in any one of the first aspects according to the instructions in the program code.

[0034] In a fourth aspect, the present application provides a computer-readable storage medium, which is used to store program code. When the program code is executed by a processor, it implements the measurement indicator anomaly analysis method described in any one of the first aspects.

[0035] It can be seen from the above technical solutions that this application has the following advantages:

[0036] The present application provides a method for analyzing metering indicator anomalies, comprising: when a metering indicator anomaly occurs, obtaining archival data and operating data of each target metering terminal associated with the metering indicator at the current moment; performing fault analysis on the archival data and operating data using a preset fault classification model to obtain a fault combination corresponding to each target metering terminal; and counting the frequency of the fault combinations of all target metering terminals, and outputting the fault combination with the highest frequency.

[0037] In this application, when an abnormality occurs in a metering indicator, the archive data and operation data of each target metering terminal related to the metering indicator at the current moment are obtained, and the archive data and operation data of each target metering terminal are automatically analyzed through a preset fault classification model to obtain the fault combination of each target metering device. The fault combination with the highest frequency output at the end is the main reason for the abnormality of the metering indicator at the current moment. No human participation is required, which improves the efficiency of the abnormality analysis of the metering indicator and improves the existing technology of analyzing the abnormality of indicators through expert judgment combined with on-site operation and maintenance inspection, which has the technical problems of large workload and low efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0039] Figure 1 A flow chart of a method for analyzing abnormality of measurement indicators provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of a mathematical model structure of a neuron provided in an embodiment of the present application;

[0041] Figure 3 A schematic diagram of the structure of a measurement indicator abnormality analysis device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 are within the scope of protection of this application.

[0043] For easier understanding, see Figure 1 , an embodiment of a measurement indicator abnormality analysis method provided by the present application includes:

[0044] Step 101: When a metering indicator is abnormal, obtain the archive data and operation data of each target metering terminal associated with the metering indicator at the current moment.

[0045] When it is found that the online rate and / or meter reading rate at the current moment decreases, that is, the metering indicator is abnormal, the archive data and operation data of each target metering terminal associated with the metering indicator at the current moment are obtained, wherein the operation data is the data generated during the operation of the metering terminal.

[0046] The embodiment of the present application adds new fault classification dimensions, namely the power outage analysis dimension and the signal strength dimension. Among them, the power outage analysis dimension includes: 1. Planned power outage, matching the power outage status of the fault terminal at the current moment according to the power outage plan; 2. Line power outage, under the same line, more than 80% of the terminals are out of power, and it is judged that other terminals under the line are also in a power outage state; 3. Continuous power outage state, if the terminal in the power outage state at the previous moment is still offline at the next moment, it is judged that the state continues to the next moment. Work order analysis dimension: according to the battery duration of the meter, if there is a work order maintenance plan for the terminal within three days, it is judged that the fault terminal has a work order abnormality; Signal strength dimension: according to the recent time of the fault metering terminal, if the signal strength in the recent period is too low, it is judged that there is a signal strength abnormality.

[0047] Therefore, the final archival data of the metering terminal obtained includes: the operator of the metering terminal, the terminal type, the terminal model, the terminal manufacturer, the terminal installation date, the terminal district office, the terminal software version, the terminal master station type, the terminal base station type, the terminal base station status, the terminal power outage status, the terminal work order status and the terminal signal strength status.

[0048] Step 102: Perform fault analysis on the archive data and the operation data using a preset fault classification model to obtain a fault combination corresponding to each target metering terminal.

[0049] The acquired archival data and operating data of each target metering terminal are input into a preset fault classification model for fault analysis to obtain the fault combination corresponding to each target metering terminal.

[0050] The configuration process of the preset fault classification model is as follows:

[0051] S1. Combining the fault causes that lead to abnormal measurement indicators to obtain several fault combinations;

[0052] Combine all possible causes of metering indicator anomalies in pairs, or in pairs, to obtain several fault combinations. Refer to Table 1, and combine them based on multiple dimensions, such as district office, terminal type, terminal manufacturer, operator, and installation year.

[0053] Table 1 Failure factors

[0054]

[0055] Assume that the number of distinct categories in dimension k is n k , then the number of fault combinations obtained by any three-dimensional combination is:

[0056] ∑n i +∑n i n k +∑n i nk n j , i<k<j;

[0057] Where n i is the number of non-repeating categories of dimension i, n j is the number of distinct categories of dimension j.

[0058] S2. Determine the fault combination corresponding to the faulty metering terminal based on the fault work order data;

[0059] The cause of the fault is determined based on the keywords of the fault description in the fault work order data and the keywords of the description of the repaired fault, and then the fault combination corresponding to each faulty metering terminal is matched.

[0060] S3. Construct training samples based on the archival data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination;

[0061] The archival data corresponding to the faulty metering terminal and the operating data on the day of the fault are fused to obtain fused data. The fault combination corresponding to the faulty metering terminal is one-hot encoded, and the encoding result is used as the fault category label of the fused data to obtain training samples.

[0062] For fused data, when a fault metering terminal has missing data in a certain dimension, the mode or mean interpolation method can be used to fill it in; for the dimension of category classification, one-hot encoding is used; for the dimension of numerical category, decentralized normalization can be used; for the category that needs to be classified (that is, the fault category in the fault combination), one-hot encoding is used.

[0063] One-hot encoding, also known as single-bit encoding, primarily uses an N-bit state register to encode N states. Each state has its own register bit, and only one bit is active at any given time. One-hot encoding represents categorical variables as binary vectors. This first requires mapping the categorical values ​​to integer values; then, each integer value is represented as a binary vector.

[0064] After obtaining a number of training samples, the number of training samples in each category can be balanced according to the number of training samples in each category, and new training samples can be obtained by generating similar data.

[0065] S4. Train the convolutional neural network using training samples to obtain a preset fault classification model.

[0066] You can randomly sample 80% of the data for network training and set aside 20% for network validation. Use cross-validation and grid search to adjust network parameters until the network accuracy improves by less than a preset threshold, such as 0.0005. Grid search is an exhaustive search method for specified parameter values. It optimizes the parameters of the estimation function through cross-validation to find the optimal learning algorithm. Specifically, the possible values ​​of each parameter are permuted and combined, and all possible combinations are listed to form a "grid." Each combination is then used for network training, and performance is evaluated using cross-validation.

[0067] From the characteristics and functions of neurons, we can know that neurons are multi-input and single-output information processing units, and their processing of information is nonlinear. According to the characteristics and functions of neurons, neurons can be abstracted into a simple mathematical model, and its structure can be referred to Figure 2 .

[0068] In 2, x1, x2, ..., x n is the input of the neuron, that is, the information from the axons of the previous n neurons, Σ is the threshold of neuron i; w i1 , w i2 ..., w in They are neuron i pairs x1, x2, ..., x n The weight coefficient, that is, the transmission efficiency of the synapse; Y i is the output of neuron i; f[·] is the activation function, which determines the effect of neuron i on the input x1, x2, ..., x n How the common stimulation reaches the threshold is output.

[0069] from Figure 2 The neuron model can be expressed as follows:

[0070] a1=x1*w 11 +x2*w 12 …x n *w 1n ;

[0071] The activation function formula is defined as:

[0072]

[0073] The loss function L(·) is:

[0074]

[0075] in, is the true label of the i-th input data, is the predicted value of the i-th input data, and m is the number of input data.

[0076] In this embodiment of the application, by adding a weight decay term To modify the cost function, the attenuation term will penalize excessive parameter values. Therefore, the cost function in the embodiment of the present application is:

[0077]

[0078] Where y p is the probability of predicting fault category j, x i is the i-th training sample, θ is the vector parameter matrix of the neuron, λ is the regularization adjustment parameter, k is the predicted fault category, and m is the number of training samples.

[0079] In an embodiment of the present application, the preset fault classification model obtained through training can be used to mine effective information from operating data and terminal archive data, and to capture key issues or common issues of indicator abnormalities without relying on human active search and analysis of abnormal indicator related reports, which can effectively improve the technical problems of large workload and relatively single analysis points in existing measurement indicator abnormality analysis; and, in an embodiment of the present application, potential attributes of measurement indicator abnormality-related devices are combined through effective coding classification to express the commonality of problems causing measurement indicator abnormalities, and by constructing a neural network, combinations of obvious features that cause measurement indicator abnormalities in multiple combinations are discovered, thereby improving the efficiency of measurement indicator abnormality analysis.

[0080] Step 103: Count the frequencies of the fault combinations of all target metering terminals, and output the fault combination with the highest frequency.

[0081] The frequency of the fault combination of all target metering terminals is counted, and the fault combination with the highest frequency is output. The fault category in this fault combination is the main reason causing the abnormal metering indicator at the current moment.

[0082] Please refer to Table 2, which shows a verification result obtained when verifying the preset classification model in an embodiment of the present application.

[0083] Table 2 Model validation results

[0084]

[0085] It should be noted that the hierarchical analysis method can be used to summarize the key factors affecting the metering indicators of different regions, consider the equipment failures, network parameters, etc. in the region as factors to construct a judgment matrix, combine the actual operating status of the metering terminals in different regions as low-level factors, and calculate the hierarchical ranking weights of the key factors as a reference for the key factors.

[0086] In an embodiment of the present application, when an abnormality occurs in a metering indicator, the archive data and operation data of each target metering terminal related to the metering indicator at the current moment are obtained, and the archive data and operation data of each target metering terminal are automatically analyzed through a preset fault classification model to obtain a fault combination of each target metering device. The fault combination with the highest frequency output at the end is the main reason for the abnormality of the metering indicator at the current moment. No human participation is required, which improves the efficiency of the metering indicator abnormality analysis and improves the existing technology of performing indicator abnormality analysis through expert judgment combined with on-site operation and maintenance inspection, which has the technical problems of large workload and low efficiency.

[0087] The above is an embodiment of a measurement indicator abnormality analysis method provided by the present application, and the following is an embodiment of a measurement indicator abnormality analysis device provided by the present application.

[0088] Please refer to Figure 3 , an embodiment of the present application provides a measurement indicator abnormality analysis device, comprising:

[0089] An acquisition unit, configured to acquire, when an abnormality occurs in a metering indicator, archive data and operation data of each target metering terminal associated with the metering indicator at the current moment;

[0090] The fault analysis unit is used to perform fault analysis on the archive data and operation data using a preset fault classification model to obtain the fault combination corresponding to each target metering terminal;

[0091] The output unit is used to count the frequencies of the fault combinations of all target metering terminals and output the fault combination with the highest frequency.

[0092] As a further improvement, the configuration process of the preset fault classification model is as follows:

[0093] Combine the fault causes that lead to abnormal measurement indicators to obtain several fault combinations;

[0094] Determine the fault combination corresponding to the faulty metering terminal based on the fault work order data;

[0095] Construct training samples based on the archival data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination;

[0096] The convolutional neural network is trained through training samples to obtain a preset fault classification model.

[0097] As a further improvement, the archival data includes: operator, terminal type, terminal model, terminal manufacturer, terminal installation date, terminal district office, terminal software version, terminal master station type, terminal base station type, terminal base station status, terminal power outage status, terminal work order status and terminal signal strength status.

[0098] In an embodiment of the present application, when an abnormality occurs in a metering indicator, the archive data and operation data of each target metering terminal related to the metering indicator at the current moment are obtained, and the archive data and operation data of each target metering terminal are automatically analyzed through a preset fault classification model to obtain a fault combination of each target metering device. The fault combination with the highest frequency output at the end is the main reason for the abnormality of the metering indicator at the current moment. No human participation is required, which improves the efficiency of the metering indicator abnormality analysis and improves the existing technology of performing indicator abnormality analysis through expert judgment combined with on-site operation and maintenance inspection, which has the technical problems of large workload and low efficiency.

[0099] The embodiment of the present application also provides a measurement indicator abnormality analysis device, the device including a processor and a memory;

[0100] The memory is used to store program codes and transmit the program codes to the processor;

[0101] The processor is used to execute the measurement indicator abnormality analysis method in the aforementioned method embodiment according to the instructions in the program code.

[0102] An embodiment of the present application also provides a computer-readable storage medium, which is used to store program code. When the program code is executed by a processor, it implements the measurement indicator anomaly analysis method in the aforementioned method embodiment.

[0103] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0104] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof 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.

[0105] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0106] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0107] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0108] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0109] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for executing all or part of the steps of the method described in each embodiment of the present application through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English full name: Read-Only Memory, English abbreviation: ROM), random access memory (English full name: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program code.

[0110] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for analyzing abnormality of measurement indicators, characterized in that: include: When an abnormality occurs in a metering indicator, the archive data and operation data of each target metering terminal associated with the metering indicator at the current moment are obtained; wherein the archive data includes: operator, terminal type, terminal model, terminal manufacturer, terminal installation date, terminal district bureau, terminal software version, terminal master station type, terminal base station type, terminal base station status, terminal power failure status, terminal work order status and terminal signal strength status, and the operation data is the data generated during the operation of the metering terminal; Performing fault analysis on the archive data and the operation data by using a preset fault classification model to obtain a fault combination corresponding to each target metering terminal; Counting the frequencies of the fault combinations of all the target metering terminals, and outputting the fault combination with the highest frequency; The configuration process of the preset fault classification model is: Combine the fault causes that lead to abnormal measurement indicators to obtain several fault combinations; Determine the fault combination corresponding to the faulty metering terminal according to the fault work order data; Constructing training samples based on the archive data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination; Training a convolutional neural network using the training samples to obtain the preset fault classification model; The constructing of training samples according to the archive data corresponding to the fault metering terminal, the operation data on the day of the fault, and the fault combination includes: Fusing the archive data corresponding to the faulty metering terminal and the operation data on the day of the fault to obtain fused data; One-hot encoding is performed on the fault combination corresponding to the fault metering terminal, and the obtained encoding result is used as the fault category label of the fused data to obtain a training sample.

2. A measurement index abnormality analysis device, characterized in that: include: An acquisition unit is used to acquire the archive data and operation data of each target metering terminal associated with the metering indicator at the current moment when an abnormality occurs in the metering indicator; wherein the archive data includes: operator, terminal type, terminal model, terminal manufacturer, terminal installation date, terminal district bureau, terminal software version, terminal master station type, terminal base station type, terminal base station status, terminal power failure status, terminal work order status and terminal signal strength status; and the operation data is data generated during the operation of the metering terminal; A fault analysis unit, used to perform fault analysis on the archive data and the operation data by using a preset fault classification model to obtain a fault combination corresponding to each target metering terminal; An output unit, used to count the frequencies of the fault combinations of all the target metering terminals and output the fault combination with the highest frequency; The configuration process of the preset fault classification model is: Combine the fault causes that lead to abnormal measurement indicators to obtain several fault combinations; Determine the fault combination corresponding to the faulty metering terminal according to the fault work order data; Constructing training samples based on the archive data corresponding to the faulty metering terminal, the operating data on the day of the fault, and the fault combination; Training a convolutional neural network using the training samples to obtain the preset fault classification model; The constructing of training samples according to the archive data corresponding to the fault metering terminal, the operation data on the day of the fault, and the fault combination includes: Fusing the archive data corresponding to the faulty metering terminal and the operation data on the day of the fault to obtain fused data; One-hot encoding is performed on the fault combination corresponding to the fault metering terminal, and the obtained encoding result is used as the fault category label of the fused data to obtain a training sample.

3. A measurement index abnormality analysis device, characterized in that: The device comprises a processor and a memory; The memory is used to store program codes and transmit the program codes to the processor; The processor is used to execute the measurement indicator abnormality analysis method described in claim 1 according to the instructions in the program code.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code, and when the program code is executed by the processor, the measurement indicator abnormality analysis method according to claim 1 is implemented.

Images