An intelligent data optimization management platform based on power operation and maintenance

Abstract

The present invention belongs to the technical field of data optimization management. The present invention discloses an intelligent data optimization management platform based on power operation and maintenance, including: a relationship network construction module constructs a device relationship network according to equipment installation information and historical operation records; a model construction and training module uses historical operation data to train a state recognition model; a data acquisition and processing module collects and preprocesses real-time operation data; a data analysis module analyzes whether the operation data is abnormal based on the device relationship network; a state recognition module uses a state recognition model to identify the device state when an abnormality is detected; and a monitoring and alarm module monitors the operation data and alarms the abnormality. By constructing a device relationship network, the overall analysis of the operation data of multiple devices is realized, the mutual influence between devices is considered, the accuracy and comprehensiveness of abnormality detection are improved, valuable reference information can be provided for operation and maintenance personnel, and the efficiency of power operation and maintenance management is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of data optimization management, and more specifically, to an intelligent data optimization management platform based on power operation and maintenance. Background Art

[0002] As an important infrastructure of modern society, the safe and stable operation of the power system is crucial to the national economy and people's lives. With the continuous expansion of the scale of the power grid and the continuous growth of electricity demand, the complexity and uncertainty of the power system are also increasing, which brings huge challenges to power operation and maintenance. Traditional power operation and maintenance methods mainly rely on manual inspections and experience judgments, which have problems such as low efficiency, poor real-time performance, and insufficient accuracy, and are difficult to meet the operation and maintenance needs of modern power grids.

[0003] In recent years, with the rapid development of new-generation information technologies such as the Internet of Things, big data, and artificial intelligence, intelligent operation and maintenance has gradually become a development trend in the power industry. Intelligent operation and maintenance collects massive amounts of equipment operation data and uses advanced data analysis and artificial intelligence algorithms to achieve real-time monitoring of equipment status, fault warning, and intelligent diagnosis, greatly improving the efficiency and accuracy of power operation and maintenance. However, existing systems often analyze each device in isolation, ignoring the correlation between devices and easily overlooking some potential problems, which in turn leads to unreasonable arrangements for operation and maintenance work and wastes human and material resources. Summary of the invention

[0004] In order to overcome the problem that the prior art analyzes each device in isolation and ignores the correlation between devices, the present invention proposes an intelligent data optimization management platform based on power operation and maintenance to solve the above problem.

[0005] The present invention provides the following technical solutions:

[0006] An intelligent data optimization management platform based on power operation and maintenance, including:

[0007] A relationship network building module is used to obtain installation information of all devices and historical operation records of all devices, wherein the operation records include operation data and status data, and to build a device relationship network based on the installation information and historical operation data;

[0008] A model building and training module is used to train a state recognition model for identifying the state of the equipment based on the historical operation data of the equipment;

[0009] The data collection and processing module is used to collect the real-time operation data of each device through sensors, pre-process the collected operation data, and store the pre-processed operation data after adding a timestamp;

[0010] A data analysis module is used to obtain the stored operation data according to the device relationship network at a preset frequency and analyze whether there is any abnormality in the operation data;

[0011] The state recognition module is used to obtain the corresponding operation data and input the operation data into the state recognition model to obtain the equipment state when the data analysis module analyzes the operation data and finds abnormalities;

[0012] The monitoring and alarm module is used to monitor the real-time operating data obtained and to issue an alarm when the operating data exceeds the preset threshold. It is also used to monitor and output the device status identified by the status recognition model.

[0013] Preferably, the installation information includes the connection relationship between devices and the distance between devices, the operation data includes voltage, load and temperature, and the status data includes description information and corresponding codes.

[0014] Preferably, the step of constructing a device relationship network based on the installation information and historical operation data includes:

[0015] S1. Obtain any device as the central device, and use the central device and all devices that are connected to the central device and whose distance from the central device does not exceed a preset distance limit to form an initial device set;

[0016] S2, using each device in the initial device set as a node and the connection relationship between each device as an edge to build the initial network;

[0017] S3. Obtain the association relationship corresponding to each edge in the initial network according to the running data, and delete the edges with empty association relationships in the initial network;

[0018] S4, starting from the node corresponding to the central device in the initial network, using the search algorithm to traverse the entire initial network, and using the traversed nodes and edges and the association relationship corresponding to each edge to form the target network;

[0019] S5. Obtain the number of nodes in the target network; if the number of nodes is greater than a preset value, use the target network as the device relationship network; otherwise, return to step S1.

[0020] Preferably, the step of obtaining the association relationship corresponding to each edge in the initial network according to the operating data includes:

[0021] Get any edge in the initial network, and get two nodes connected by the edge, which are recorded as the first node and the second node respectively;

[0022] Obtain n pieces of historical operation data of the device corresponding to the first node to form a first operation data set, and obtain n pieces of historical operation data of the device corresponding to the second node to form a second operation data set according to the acquisition time of each operation data in the first operation data set;

[0023] Extracting one of the three data of voltage, load and temperature from the first operation data set and the second operation data set respectively, forming two corresponding data arrays respectively recorded as a first array and a second array;

[0024] Calculate the volatility index using the volatility calculation method based on the first array and the second array;

[0025] The fluctuation indexes of voltage, load and temperature are calculated respectively, and the maximum value of the three fluctuation indexes is recorded as the maximum fluctuation index;

[0026] Determine whether the maximum fluctuation index is less than the preset fluctuation threshold. If so, set the association relationship of the edge to null;

[0027] If not, obtain the first array and the second array corresponding to the maximum volatility index, and calculate the proportional index using the proportional calculation method according to the first array and the second array;

[0028] Determine whether the ratio index is greater than or equal to the preset ratio threshold. If so, set the association relationship of the edge to null;

[0029] If not, the maximum fluctuation index, the type of operating data corresponding to the maximum fluctuation index, and the proportion index are used as the association relationship corresponding to the edge.

[0030] Preferably, the steps of the fluctuation calculation method include:

[0031] Receive two sets of data, recorded as array A and array B,

[0032] The volatility index is calculated using the following formula:

[0033]

[0034] In the formula, BD represents the volatility index, i represents the subscript, m represents the number of elements in array A, and A i represents the i-th element in array A, B i Represents the i-th element in array B, sgx() represents the sign function. If the value in () is greater than 0, the value of sgx() is 1, otherwise the value of sgx() is 0.

[0035] Preferably, the steps of the ratio calculation method include:

[0036] Receive two sets of data, recorded as array C and array D,

[0037] Create an array C' with one less element than array C, and follow the formula Assign values ​​to the elements in array C', where C' i Represents the i-th element in array C', C i Represents the i-th element in array C;

[0038] Create an array D' with one less element than array D, and follow the formula Assign values ​​to the elements in array D', where D' i represents the i-th element in array D', D i Represents the i-th element in array D;

[0039] The proportionality index is calculated using the following formula:

[0040]

[0041] Where BL represents the scale index, q represents the number of elements in array C', represents the average value of the elements in array C', Represents the average value of the elements in array D'.

[0042] Preferably, the training step of the state recognition model includes:

[0043] Use the operation records of all devices in the device relationship network at any time in history as a piece of learning data, obtain multiple pieces of learning data to form a learning data set, and divide the learning data set into a training set and a validation set;

[0044] Build a machine learning framework and define the input layer and output layer, set the loss function and optimizer, and build a machine learning model;

[0045] Use the running data in the learning dataset as the input layer data, use the encoding corresponding to the running data as the output layer data, and use the training set to train the constructed machine learning model according to the set batch size and number of training rounds;

[0046] The trained machine learning model is tested using the validation set, and the machine learning model that meets the preset accuracy is output as the state recognition model.

[0047] Preferably, the step of acquiring the stored operation data according to the device relationship network and analyzing whether the operation data has an abnormality includes:

[0048] Get all the edges in the device relationship network to form an edge set;

[0049] Obtain any edge in the edge set as the target edge, obtain the running data type corresponding to the maximum fluctuation index in the association relationship corresponding to the target edge as the target type, the maximum fluctuation index as the target fluctuation, and the proportion index as the target proportion;

[0050] Obtain two devices corresponding to two nodes connected by the target edge and record them as the first device and the second device respectively;

[0051] Acquire data of all target types within the most recent time t of the first device and the second device from the stored operation data to form a first operation data group and a second operation data group respectively;

[0052] Performing data alignment on the first operation data group and the second operation data group;

[0053] Calculate the fluctuation index of the first and second operation data groups after alignment using a fluctuation calculation method and record it as the actual fluctuation, and calculate the fluctuation index of the first and second operation data groups after alignment using a ratio calculation method and record it as the actual ratio;

[0054] Determine whether the error between the actual fluctuation and the target fluctuation and the error between the actual ratio and the target ratio are both within a preset error range;

[0055] If yes, the target edge is marked as normal; if no, the target edge is marked as abnormal;

[0056] When any edge in the edge set is marked as an anomaly, the analysis result is set to have an anomaly; otherwise, the analysis result is set to have no anomaly.

[0057] Preferably, the step of aligning the first operation data group and the second operation data group comprises:

[0058] Selecting the group with less data in the first operation data group and the second operation data group as a reference group, and the other group as a selection group;

[0059] Extract the timestamp of each data from the reference group, and find the data closest to each timestamp in the selected group in turn to form an alignment group;

[0060] If the reference group is the first running data group, use the reference group as the first running data after alignment, and use the alignment group as the second running data after alignment; otherwise, use the alignment group as the first running data after alignment, and use the reference group as the second running data after alignment.

[0061] Preferably, the obtaining of corresponding operating data and inputting the operating data into a state recognition model to obtain the device state includes: obtaining the most recent operating data of all devices in the device relationship network; inputting the obtained operating data into the state recognition model to obtain the code of the model output; the monitoring and outputting of the device state recognized by the state recognition model includes: obtaining the code of the model output, obtaining corresponding description information according to the code of the model output, and outputting the description information.

[0062] The present invention provides an intelligent data optimization management platform based on power operation and maintenance, which has the following beneficial effects:

[0063] Through the relationship network construction module, the relationship network between devices is constructed using the installation information and historical operation records of the equipment, providing a basis for the subsequent comprehensive analysis of the operation data, thereby realizing the overall analysis of the operation data of multiple devices. The data analysis module is based on the equipment relationship network for analysis. It can integrate the data of multiple devices to analyze the operation status of the equipment, rather than focusing only on the operation data of a single device itself. It can consider the mutual influence between devices and improve the accuracy and comprehensiveness of anomaly detection; the state recognition module uses the associated information in the relationship network, combined with the trained model, to obtain descriptive information, which can provide a reference for the operation and maintenance work of public personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a module schematic diagram of an intelligent data optimization management platform based on power operation and maintenance of the present invention;

[0065] Figure 2 The present invention is a flowchart of constructing a device relationship network for an intelligent data optimization management platform based on power operation and maintenance. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.

[0067] Example 1

[0068] See also Figure 1 In this embodiment, an intelligent data optimization management platform based on power operation and maintenance includes:

[0069] A relationship network building module is used to obtain installation information of all devices and historical operation records of all devices, wherein the operation records include operation data and status data, and to build a device relationship network based on the installation information and historical operation data;

[0070] The installation information includes the connection relationship between devices and the distance between devices, the operation data includes voltage, load and temperature, and the status data includes description information and corresponding codes.

[0071] In this embodiment, the installation information and history records of the device can be obtained by connecting to the relevant database using scripts or manual import, etc. The description information can be in various forms such as pictures, text and audio, and can usually be entered by the staff according to the actual situation.

[0072] See also Figure 2 The step of constructing a device relationship network based on the installation information and historical operation data includes:

[0073] S1. Obtain any device as the central device, and use the central device and all devices that are connected to the central device and whose distance from the central device does not exceed a preset distance limit to form an initial device set;

[0074] S2, using each device in the initial device set as a node and the connection relationship between each device as an edge to build the initial network;

[0075] S3. Obtain the association relationship corresponding to each edge in the initial network according to the running data, and delete the edges with empty association relationships in the initial network;

[0076] S4, starting from the node corresponding to the central device in the initial network, using the search algorithm to traverse the entire initial network, and using the traversed nodes and edges and the association relationship corresponding to each edge to form the target network;

[0077] S5. Obtain the number of nodes in the target network; if the number of nodes is greater than a preset value, use the target network as the device relationship network; otherwise, return to step S1.

[0078] The step of obtaining the association relationship corresponding to each edge in the initial network according to the operation data includes:

[0079] Get any edge in the initial network, and get two nodes connected by the edge, which are recorded as the first node and the second node respectively;

[0080] Obtain n pieces of historical operation data of the device corresponding to the first node to form a first operation data set, and obtain n pieces of historical operation data of the device corresponding to the second node to form a second operation data set according to the acquisition time of each operation data in the first operation data set;

[0081] Extracting one of the three data of voltage, load and temperature from the first operation data set and the second operation data set respectively, forming two corresponding data arrays respectively recorded as a first array and a second array;

[0082] Calculate the volatility index using the volatility calculation method based on the first array and the second array;

[0083] The fluctuation indexes of voltage, load and temperature are calculated respectively, and the maximum value of the three fluctuation indexes is recorded as the maximum fluctuation index;

[0084] Determine whether the maximum fluctuation index is less than the preset fluctuation threshold. If so, set the association relationship of the edge to null;

[0085] If not, obtain the first array and the second array corresponding to the maximum volatility index, and calculate the proportional index using the proportional calculation method according to the first array and the second array;

[0086] Determine whether the ratio index is greater than or equal to the preset ratio threshold. If so, set the association relationship of the edge to null;

[0087] If not, the maximum fluctuation index, the type of operating data corresponding to the maximum fluctuation index, and the proportion index are used as the association relationship corresponding to the edge.

[0088] The steps of the fluctuation calculation method include:

[0089] Receive two sets of data, recorded as array A and array B,

[0090] The volatility index is calculated using the following formula:

[0091]

[0092] In the formula, BD represents the volatility index, i represents the subscript, m represents the number of elements in array A, and A i represents the i-th element in array A, B i Represents the i-th element in array B, sgx() represents the sign function. If the value in () is greater than 0, the value of sgx() is 1, otherwise the value of sgx() is 0.

[0093] The steps of the ratio calculation method include:

[0094] Receive two sets of data, recorded as array C and array D,

[0095] Create an array C' with one less element than array C, and follow the formula Assign values ​​to the elements in array C', where C' i Represents the i-th element in array C', Ci Represents the i-th element in array C;

[0096] Create an array D' with one less element than array D, and follow the formula Assign values ​​to the elements in array D', where D' i represents the i-th element in array D', D i Represents the i-th element in array D;

[0097] The proportionality index is calculated using the following formula:

[0098]

[0099] Where BL represents the scale index, q represents the number of elements in array C', represents the average value of the elements in array C', Represents the average value of the elements in array D'.

[0100] In this embodiment, firstly, the devices that may have mutual influence are divided into a network by connection relationship and distance. It should be noted that the connection relationship can be a direct circuit connection or communication connection, etc. Then, the change of the data between the devices in the time series is used to determine whether there is influence between the devices. It is mainly reflected by the fluctuation index and the proportion index. The larger the fluctuation index, the more times the data changes in the same direction in the time series, and the greater the possibility of influence between the devices. The larger the proportion index, the greater the difference in the change ratio of the two devices in the time series, and the less likely the influence between the devices.

[0101] During specific implementation, the preset distance limit can be set to 1000 meters. 100 pieces of historical operation data can be collected. The preset fluctuation threshold can be set to 0.8. When the fluctuation index is less than 0.8, it is considered that the correlation between devices is weak. The preset ratio threshold can be set to 0.5. When the ratio index is greater than or equal to 0.5, it is considered that the difference in the change ratio between devices is too large and there is no significant correlation. The preset value of the number of target network nodes can be set to 5. When the number of nodes in the target network is greater than 5, it is considered that the network scale is sufficient to reflect the relationship between devices. When traversing the initial network, the breadth-first search (BFS) algorithm is used. This algorithm is suitable for finding the shortest path in the network structure and can effectively traverse the entire initial network.

[0102] In actual applications, these preset values ​​can be adjusted according to specific situations or personnel experience, etc. Through the above steps, the device relationship network can be accurately constructed, and the devices that really have an impact relationship can be effectively identified, while filtering out irrelevant or low-correlation devices, thereby improving the accuracy and practicality of the network.

[0103] A model building and training module is used to train a state recognition model for identifying the state of the equipment based on the historical operation data of the equipment;

[0104] The training steps of the state recognition model include:

[0105] Use the operation records of all devices in the device relationship network at any time in history as a piece of learning data, obtain multiple pieces of learning data to form a learning data set, and divide the learning data set into a training set and a validation set;

[0106] Build a machine learning framework and define the input layer and output layer, set the loss function and optimizer, and build a machine learning model;

[0107] Use the running data in the learning dataset as the input layer data, use the encoding corresponding to the running data as the output layer data, and use the training set to train the constructed machine learning model according to the set batch size and number of training rounds;

[0108] The trained machine learning model is tested using the validation set, and the machine learning model that meets the preset accuracy is output as the state recognition model.

[0109] In this embodiment, the training method of the state recognition model can be performed as follows:

[0110] First, obtain a data set containing 10,000 equipment operation records. Each record contains the equipment's operating data (voltage, load, and temperature) and the corresponding status data (description information and encoding). Next, combine the operating data into a feature vector. Then, use a stratified sampling method to divide the data set into a training set and a validation set in a ratio of 7:3 to ensure that the distribution ratio of each type of data in the two sets is consistent.

[0111] When building a machine learning framework, you can choose the TensorFlow framework to build a neural network model. The model structure includes an input layer, two hidden layers, and an output layer. The input layer receives three features, corresponding to the dimensions of the feature vector; the output layer produces values ​​with the same dimensions as the state encoding. The hidden layer uses the ReLU activation function to enhance the nonlinear expression ability of the model. For the loss function, the cross entropy loss is selected, which is suitable for multi-classification problems. The optimizer selects the Adam algorithm, which can adaptively adjust the learning rate, which is conducive to the rapid convergence of the model. When compiling the model, the accuracy is also set as an evaluation indicator to measure the classification performance of the model.

[0112] During the training process, the batch size is set to 128, the initial learning rate is set to 0.001, and the number of training rounds is set to 100. During the training process, the learning rate decay strategy is used to reduce the learning rate to 90% of the original every 10 rounds to help the model converge better. During training, 90% of the training set is used for training, and the remaining 10% is used for testing to monitor whether the model is overfitting.

[0113] After training is completed, the previously divided validation set is used to evaluate the model performance. The preset accuracy requirement can be that the classification accuracy is greater than 95%. If the test result meets this condition, the model is used as the final state recognition model. The model can predict the current state code of the device based on the input device operation data (voltage, load and temperature). In this way, real-time monitoring and anomaly detection of device status can be achieved, improving the intelligence level and efficiency of power operation and maintenance.

[0114] The data collection and processing module is used to collect the real-time operation data of each device through sensors, pre-process the collected operation data, and store the pre-processed operation data after adding a timestamp;

[0115] In this embodiment, data preprocessing includes but is not limited to common data cleaning, data format unification, and data compression.

[0116] A data analysis module is used to obtain the stored operation data according to the device relationship network at a preset frequency and analyze whether there is any abnormality in the operation data;

[0117] The step of acquiring the stored operation data according to the device relationship network and analyzing whether the operation data is abnormal includes:

[0118] Get all the edges in the device relationship network to form an edge set;

[0119] Obtain any edge in the edge set as the target edge, obtain the running data type corresponding to the maximum fluctuation index in the association relationship corresponding to the target edge as the target type, the maximum fluctuation index as the target fluctuation, and the proportion index as the target proportion;

[0120] Obtain two devices corresponding to two nodes connected by the target edge and record them as the first device and the second device respectively;

[0121] Acquire data of all target types within the most recent time t of the first device and the second device from the stored operation data to form a first operation data group and a second operation data group respectively;

[0122] Performing data alignment on the first operation data group and the second operation data group;

[0123] Calculate the fluctuation index of the first and second operation data groups after alignment using a fluctuation calculation method and record it as the actual fluctuation, and calculate the fluctuation index of the first and second operation data groups after alignment using a ratio calculation method and record it as the actual ratio;

[0124] Determine whether the error between the actual fluctuation and the target fluctuation and the error between the actual ratio and the target ratio are both within a preset error range;

[0125] If yes, the target edge is marked as normal; if no, the target edge is marked as abnormal;

[0126] When any edge in the edge set is marked as an anomaly, the analysis result is set to have an anomaly; otherwise, the analysis result is set to have no anomaly.

[0127] In this embodiment, the analysis task can be performed at a preset frequency (for example, once per minute) to ensure that abnormal conditions in the operation of the equipment are discovered in a timely manner. The analysis task first extracts all edges from the device relationship network and constructs an edge set. Assume that the set contains 10 edges. For each edge in the set, we regard it as an analysis object and focus on its associated key indicators: specific operating data types such as the maximum fluctuation index of voltage and its proportional index. Specifically for a certain target edge, we identify and mark the two devices connected to it as the first device and the second device. Assume that for this target edge, the target type of interest is voltage, and it is known that its target fluctuation is 0.85 and its target ratio is 0.35. Subsequently, the system retrieves the voltage data records of the two devices in the last hour from the stored database, and constructs the first operating data group and the second operating data group respectively.

[0128] Next, the data is aligned, and the two sets of data can be synchronized according to the timestamp to ensure that their data points correspond one to one and the number is consistent. After that, the fluctuation calculation and proportion calculation methods are applied to calculate the actual fluctuation value and the actual proportion value respectively.

[0129] On this basis, we set a preset error range (e.g. ±5%) to evaluate whether the deviation between the actual value and the target value is within an acceptable range. If the actual fluctuation or actual ratio of any edge exceeds this range, the edge is considered to be abnormal. The entire set of edges (10 edges in this example) must be evaluated one by one through the above process.

[0130] Through this systematic analysis method, we can effectively monitor the dynamic changes of the relationship between devices. Once an edge is marked as abnormal, it may indicate that the relationship between the two connected devices has changed unusually, which may be caused by factors such as device performance degradation, improper adjustment of operating parameters, or interference from the external environment. This timely detection of abnormal situations provides a valuable basis for subsequent fault diagnosis and rapid response, helping to improve the stability and reliability of the entire system.

[0131] The state recognition module is used to obtain the corresponding operation data and input the operation data into the state recognition model to obtain the equipment state when the data analysis module analyzes the operation data and finds abnormalities;

[0132] The monitoring and alarm module is used to monitor the real-time operating data obtained and to issue an alarm when the operating data exceeds the preset threshold. It is also used to monitor and output the device status identified by the status recognition model.

[0133] The obtaining of corresponding operating data and inputting the operating data into the state recognition model to obtain the device state includes: obtaining the latest operating data of all devices in the device relationship network; inputting the obtained operating data into the state recognition model to obtain the code output by the model; the monitoring and outputting of the device state recognized by the state recognition model includes: obtaining the code output by the model, obtaining corresponding description information according to the code output by the model, and outputting the description information.

[0134] In this embodiment, after analyzing the existence of anomalies, accurate codes are obtained through machine learning models, and corresponding descriptive information is obtained based on the codes and then output. The output can be in various forms such as charts and texts according to the descriptive information, aiming to intuitively and clearly display possible situations under the current data, and provide a reference for the staff's operation and maintenance work.

[0135] Example 2

[0136] Based on Example 1, the step of aligning the first operation data group and the second operation data group includes:

[0137] Selecting the group with less data in the first operation data group and the second operation data group as a reference group, and the other group as a selection group;

[0138] Extract the timestamp of each data from the reference group, and find the data closest to each timestamp in the selected group in turn to form an alignment group;

[0139] If the reference group is the first running data group, use the reference group as the first running data after alignment, and use the alignment group as the second running data after alignment; otherwise, use the alignment group as the first running data after alignment, and use the reference group as the second running data after alignment.

[0140] In this implementation, by selecting a group with less data as a reference, the amount of data to be retained can be quickly determined, improving efficiency. Then, for each timestamp in the reference group, the closest data point is found in the selected group. This method avoids complex interpolation calculations while ensuring the accuracy of alignment.

[0141] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0142] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, 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.

[0143] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0144] Finally: The above description is only a preferred 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 protection scope of the present invention.

Claims

1. An intelligent data optimization management platform based on power operation and maintenance, characterized in that: include: A relationship network building module is used to obtain installation information of all devices and historical operation records of all devices, wherein the operation records include operation data and status data, and to build a device relationship network based on the installation information and historical operation data; The step of constructing a device relationship network based on the installation information and historical operation data includes: S1. Obtain any device as the central device, and use the central device and all devices that are connected to the central device and whose distance from the central device does not exceed a preset distance limit to form an initial device set; S2. Use each device in the initial device set as a node and the connection relationship between each device as an edge to build the initial network; S3. Obtain the association relationship corresponding to each edge in the initial network according to the running data, and delete the edges with empty association relationships in the initial network; S4, starting from the node corresponding to the central device in the initial network, using the search algorithm to traverse the entire initial network, and using the traversed nodes and edges and the association relationship corresponding to each edge to form the target network; S5. Obtain the number of nodes in the target network; if the number of nodes is greater than a preset value, use the target network as the device relationship network; otherwise, return to step S1; The step of obtaining the association relationship corresponding to each edge in the initial network according to the running data includes: Get any edge in the initial network, and get the two nodes connected by the edge, which are recorded as the first node and the second node respectively; Obtain n pieces of historical operation data of the device corresponding to the first node to form a first operation data set, and obtain n pieces of historical operation data of the device corresponding to the second node to form a second operation data set according to the acquisition time of each operation data in the first operation data set; Extracting one of the three data of voltage, load and temperature from the first operation data set and the second operation data set respectively, forming two corresponding data arrays respectively recorded as a first array and a second array; Calculate the volatility index using the volatility calculation method based on the first array and the second array; The fluctuation indexes of voltage, load and temperature are calculated respectively, and the maximum value of the three fluctuation indexes is recorded as the maximum fluctuation index; Determine whether the maximum fluctuation index is less than the preset fluctuation threshold. If so, set the association relationship of the edge to null; If not, obtain the first array and the second array corresponding to the maximum volatility index, and calculate the proportional index using the proportional calculation method according to the first array and the second array; Determine whether the ratio index is greater than or equal to the preset ratio threshold. If so, set the association relationship of the edge to null; If not, the maximum fluctuation index, the type of operating data corresponding to the maximum fluctuation index, and the proportion index are used as the association relationship corresponding to the edge; The steps of the fluctuation calculation method include: Receive two sets of data, recorded as array A and array B, The volatility index is calculated using the following formula: In the formula, BD represents the volatility index, i represents the subscript, m represents the number of elements in array A, and A i represents the i-th element in array A, B i Represents the i-th element in array B, sgx() represents the sign function. If the value in () is greater than 0, the value of sgx() is 1, otherwise the value of sgx() is 0; The steps of the ratio calculation method include: Receive two sets of data, recorded as array C and array D, Create an array C' with one less element than array C, and follow the formula Assign values ​​to the elements in array C', where C' i Represents the i-th element in array C', C i Represents the i-th element in array C; Create an array D' with one less element than array D, and follow the formula Assign values ​​to the elements in array D', where D' i represents the i-th element in array D', D i Represents the i-th element in array D; The proportionality index is calculated using the following formula: Where BL represents the scale index, q represents the number of elements in array C', represents the average value of the elements in array C', Represents the average value of the elements in array D'; A model building and training module is used to train a state recognition model for identifying the state of the equipment based on the historical operation data of the equipment; The data collection and processing module is used to collect the real-time operation data of each device through sensors, pre-process the collected operation data, and store the pre-processed operation data after adding a timestamp; A data analysis module is used to obtain the stored operation data according to the device relationship network at a preset frequency and analyze whether there is any abnormality in the operation data; The state recognition module is used to obtain the corresponding operation data and input the operation data into the state recognition model to obtain the equipment state when the data analysis module analyzes the operation data and finds abnormalities; The monitoring and alarm module is used to monitor the real-time operating data obtained and to issue an alarm when the operating data exceeds the preset threshold. It is also used to monitor and output the device status identified by the status recognition model.

2. The intelligent data optimization management platform based on power operation and maintenance according to claim 1 is characterized in that: The installation information includes the connection relationship between devices and the distance between devices, the operation data includes voltage, load and temperature, and the status data includes description information and corresponding codes.

3. The intelligent data optimization management platform based on power operation and maintenance according to claim 2 is characterized in that: The training steps of the state recognition model include: Use the operation records of all devices in the device relationship network at any time in history as a piece of learning data, obtain multiple pieces of learning data to form a learning data set, and divide the learning data set into a training set and a validation set; Build a machine learning framework and define the input layer and output layer, set the loss function and optimizer, and build a machine learning model; Use the running data in the learning dataset as the input layer data, use the encoding corresponding to the running data as the output layer data, and use the training set to train the constructed machine learning model according to the set batch size and number of training rounds; The trained machine learning model is tested using the validation set, and the machine learning model that meets the preset accuracy is output as the state recognition model.

4. The intelligent data optimization management platform based on power operation and maintenance according to claim 3 is characterized in that: The step of acquiring the stored operation data according to the device relationship network and analyzing whether the operation data is abnormal includes: Get all the edges in the device relationship network to form an edge set; Obtain any edge in the edge set as the target edge, obtain the running data type corresponding to the maximum fluctuation index in the association relationship corresponding to the target edge as the target type, the maximum fluctuation index as the target fluctuation, and the proportion index as the target proportion; Obtain two devices corresponding to two nodes connected by the target edge and record them as the first device and the second device respectively; Acquire data of all target types within the most recent time t of the first device and the second device from the stored operation data to form a first operation data group and a second operation data group respectively; Performing data alignment on the first operation data group and the second operation data group; The fluctuation index of the first and second operation data groups after alignment is calculated using a fluctuation calculation method and recorded as the actual fluctuation, and the fluctuation index of the first and second operation data groups after alignment is calculated using a ratio calculation method and recorded as the actual ratio; Determine whether the error between the actual fluctuation and the target fluctuation and the error between the actual ratio and the target ratio are both within a preset error range; If yes, the target edge is marked as normal; if no, the target edge is marked as abnormal; When any edge in the edge set is marked as an anomaly, the analysis result is set to have an anomaly; otherwise, the analysis result is set to have no anomaly.

5. The intelligent data optimization management platform based on power operation and maintenance according to claim 4 is characterized in that: The step of aligning the first operation data group and the second operation data group comprises: Selecting the group with less data in the first operation data group and the second operation data group as a reference group, and the other group as a selection group; Extract the timestamp of each data from the reference group, and find the data closest to each timestamp in the selected group in turn to form an alignment group; If the reference group is the first running data group, use the reference group as the first running data after alignment, and use the alignment group as the second running data after alignment; otherwise, use the alignment group as the first running data after alignment, and use the reference group as the second running data after alignment.

6. The intelligent data optimization management platform based on power operation and maintenance according to claim 5 is characterized in that: The obtaining of corresponding operating data and inputting the operating data into the state recognition model to obtain the device state includes: obtaining the latest operating data of all devices in the device relationship network; inputting the obtained operating data into the state recognition model to obtain the code output by the model; the monitoring and outputting of the device state recognized by the state recognition model includes: obtaining the code output by the model, obtaining corresponding description information according to the code output by the model, and outputting the description information.

Images