Electric power comprehensive monitoring system construction method and system applied to front-end project

Abstract

The invention provides a lightweight electric power comprehensive monitoring system construction method and system applied to a front-end project, and the construction method comprises the following steps: configuring a dynamic display interface at a Web end of the front-end project, which is used for carrying out the preprocessing, caching and blocking processing of operation data of a target electric power system, the method comprises the following steps of: carrying out data preprocessing on a dynamic trend curve graph, drawing the dynamic updated trend curve graph by using a D3. Js library, carrying out curve drawing by using a request AnimationFrame, and running data preprocessing and dynamic updating calculation of the trend curve graph in Web Workers; and constructing an anomaly prediction model which comprises a multi-layer perceptron network created by using a TensorFlow.js library and is used for calculating the possibility of giving an alarm in the future according to the collected operation data of the target power system, and when the calculated probability exceeds a preset threshold value, carrying out first-class marking on the moment when the alarm is predicted to be given on the trend chart, and sending out an early warning notice at the same time. According to the method and the system, operation trend display and risk early warning of the target power system can be carried out at the Web end of the front-end project.

Description

Technical Field

[0001] The present invention relates to the technical field of power system monitoring, and in particular to a method and system for constructing a comprehensive power monitoring system applied to a front-end project. Background Art

[0002] In power systems, real-time monitoring and prediction of potential faults or abnormal conditions are crucial. Existing technologies usually rely on threshold detection or simple statistical analysis to identify anomalies, but these methods have the following disadvantages:

[0003] 1) Unable to capture complex trend changes, resulting in high false alarm rate or missing key events;

[0004] 2) Can only respond to problems after they occur, lacking the ability to predict potential problems;

[0005] 3) It is highly dependent on the alarm system. If the alarm system has an abnormality, it will not be able to identify the abnormal situation in time.

[0006] In the prior art, there are also some systems built using neural network models to monitor and predict potential faults or abnormal situations, but such systems usually rely on back-end servers for data analysis and processing, have high construction and operation costs, and cannot be flexibly configured and applied on the front end. The front end is limited in data processing capabilities and cannot independently implement the operation of such monitoring systems. Summary of the invention

[0007] The purpose of the present invention is to provide a lightweight method for constructing an integrated power monitoring system for front-end projects and a system constructed based on the method, which can display the operating trend and risk warning of the target power system on the Web side of the front-end project. The following technical solutions are specifically adopted:

[0008] The first aspect of the present invention provides a method for constructing an integrated power monitoring system for a front-end project, comprising the following steps:

[0009] A dynamic display interface is configured on the Web side of the front-end project, which is used to preprocess, cache and block the operating data of the target power system, and use the D3.js library to draw a dynamically updated trend curve chart. The trend curve chart uses requestAnimationFrame to perform curve drawing, and the data preprocessing and dynamic update calculation of the trend curve chart are run in Web Workers;

[0010] An abnormality prediction model is built on the Web side of the front-end project, which includes a multi-layer perceptron network created using the TensorFlow.js library, which is used to calculate the possibility of future alarms based on the collected target power system operation data, and when the calculated probability exceeds a preset threshold, the predicted alarm moment is marked as the first category on the trend chart, and an early warning notification is issued at the same time.

[0011] Furthermore, the operating data of the target power system includes real-time operating data and historical operating data, as well as real-time alarm data and historical alarm data, and the dynamic display interface is also configured to use a second type of mark on the trend curve chart to mark the specific time period where the alarm occurs based on the alarm data.

[0012] Furthermore, the preprocessing of the operation data includes:

[0013] Clean running data, handle outliers and fill in missing values;

[0014] Normalize the cleaned data and scale them uniformly to the range of 0 to 1;

[0015] The alarm data is converted into binary label form, where 1 represents an alarm event and 0 represents that the system is in normal operation.

[0016] Furthermore, the multilayer perceptron network comprises:

[0017] An input layer, which includes N neurons, and the shape of the input data is a vector containing N features, each neuron corresponds to one feature, and the N features correspond to the preset parameters in the running data;

[0018] The first hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the input layer and performs linear transformation and nonlinear activation;

[0019] The second hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the first hidden layer and performs further linear transformation and nonlinear activation;

[0020] The output layer, which contains 1 neuron, uses the Sigmoid activation function, receives data from the second hidden layer, and outputs a value between 0 and 1.

[0021] Furthermore, it also includes gradually updating and optimizing the abnormality prediction model based on the incremental learning method of TensorFlow.js, specifically using the following method:

[0022] Every time a real alarm occurs, the operating data within a preset time period before the real alarm moment is extracted;

[0023] Use the TensorFlow.js training method model.fit to train the anomaly prediction model based on the extracted data and specified training parameters.

[0024] The second aspect of the present invention provides a comprehensive power monitoring system applied to a front-end project, comprising at least:

[0025] A dynamic display interface is used to preprocess, cache and block the operating data of the target power system, and use the D3.js library to draw a dynamically updated trend curve graph. The trend curve graph uses requestAnimationFrame to perform curve drawing, and the data preprocessing and dynamic update calculation of the trend curve graph are run in Web Workers;

[0026] The abnormality prediction model includes a multi-layer perceptron network created using the TensorFlow.js library, which is used to calculate the possibility of future alarms based on the collected target power system operation data, and when the calculated probability exceeds a preset threshold, the first category mark is made on the trend chart at the moment when the alarm is predicted to occur, and an early warning notification is issued at the same time.

[0027] Furthermore, the operating data of the target power system includes real-time operating data and historical operating data, as well as real-time alarm data and historical alarm data. The dynamic display interface is also used to mark the specific time period where the alarm occurs on the trend curve chart with a second type of mark based on the alarm data.

[0028] Furthermore, the preprocessing of the operation data includes:

[0029] Clean running data, handle outliers and fill in missing values;

[0030] Normalize the cleaned data and scale them uniformly to the range of 0 to 1;

[0031] The alarm data is converted into binary label form, where 1 represents an alarm event and 0 represents that the system is in normal operation.

[0032] Furthermore, the multilayer perceptron network comprises:

[0033] An input layer, which includes N neurons, and the shape of the input data is a vector containing N features, each neuron corresponds to one feature, and the N features correspond to the preset parameters in the running data;

[0034] The first hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the input layer and performs linear transformation and nonlinear activation;

[0035] The second hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the first hidden layer and performs further linear transformation and nonlinear activation;

[0036] The output layer, which contains 1 neuron, uses the Sigmoid activation function, receives data from the second hidden layer, and outputs a value between 0 and 1.

[0037] Furthermore, the anomaly prediction model is gradually updated and optimized based on the incremental learning method of TensorFlow.js, specifically using the following method:

[0038] Every time a real alarm occurs, the operating data within a preset time period before the real alarm moment is extracted;

[0039] Use the TensorFlow.js training method model.fit to train the anomaly prediction model based on the extracted data and specified training parameters.

[0040] The beneficial effects of the present invention are as follows:

[0041] 1) The construction method of the present invention and the system constructed based on the method adopt a lightweight mode and can be deployed on the Web end of the front-end project for operation, which greatly reduces the construction and operation costs of the monitoring system. It can also be flexibly deployed as needed and can also be integrated with other power monitoring systems or enterprise management systems, making deployment and operation more flexible and convenient.

[0042] 2) Data preprocessing and dynamic update calculation of trend curve graph are run in Web Workers. Since WebWorker runs in an independent thread, time-consuming tasks can be performed in the background without blocking the rendering process of the main thread. Therefore, in the solution of the present invention, time-consuming calculations, data processing and other operations are performed in Web Workers. Even if the task is very time-consuming, the user interface can still remain responsive, providing a better user experience.

[0043] 3) Using requestAnimationFrame to perform curve drawing has higher efficiency and accuracy, and can adjust the execution frequency of the callback function according to the refresh rate of the display, thereby reducing unnecessary drawing and calculation and improving the system's operating performance.

[0044] 4) Divide the data into small blocks and process and render them one by one instead of processing all the data at once, further reducing the hardware configuration requirements of the front-end.

[0045] 5) The monitoring system constructed based on the method of the present invention can continuously monitor the operating status of the power system and predict potential abnormal risks through data analysis. When a possible abnormal situation is detected, a graphical early warning prompt can be provided through the display interface and an early warning can be issued, so that the staff can take measures quickly, thereby effectively avoiding or reducing the impact of the fault.

[0046] 6) The monitoring system constructed based on the method of the present invention has self-learning and optimization functions, which enables it to better adapt to changes in the power system and continuously maintain high-accuracy anomaly detection.

[0047] 7) The operation data, alarm data and abnormal analysis are displayed through an intuitive graphical interface, providing operators with more decision-making basis, thereby improving the work efficiency and accuracy of operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 The figure is a schematic diagram of a process of predicting anomalies using an anomaly prediction model in an embodiment of the present invention.

[0049] Figure 2 It is a schematic diagram of the overall operation flow of the comprehensive power monitoring system in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the present invention.

[0051] Terminology explanation:

[0052] D3.js (Data-Driven Documents) is a JavaScript library for creating dynamic, interactive data visualizations. It can use HTML, SVG, and CSS to display data, allowing developers to create various charts, graphics, and data visualizations in a data-driven way.

[0053] TensorFlow.js: is an open source library for machine learning and deep learning using JavaScript, allowing developers to run machine learning models in a browser or in a Node.js environment.

[0054] Epoch: In deep learning model training, N consecutive epochs means that the model completely traverses the entire training dataset N times. Specifically, each epoch means that the model performs a complete forward and backward propagation process on the entire training dataset, including data input, loss calculation, and weight update.

[0055] Example 1

[0056] The first embodiment of the present invention provides a method for constructing an integrated power monitoring system for a front-end project, comprising the following steps:

[0057] A dynamic display interface is configured on the Web side of the front-end project, which is used to preprocess, cache and block the operating data of the target power system, and use the D3.js library to draw a dynamically updated trend curve chart. In particular, the trend curve chart uses requestAnimationFrame to perform curve drawing, and the data preprocessing and dynamic update calculation of the trend curve chart are run in Web Workers;

[0058] An abnormality prediction model is built on the Web side of the front-end project, which includes a multi-layer perceptron network created using the TensorFlow.js library. It is used to calculate the possibility of future alarms based on the collected target power system operation data. When the calculated probability exceeds the preset threshold, the predicted alarm moment is marked in the first category on the trend chart of the dynamic display interface (such as giving a mark point and flashing prompt), and an early warning notification is issued at the same time.

[0059] By drawing trend curves on the dynamic display interface, not only can the changing trends of various indicators be displayed intuitively, but also the real-time nature of the information can be ensured, helping monitoring personnel to promptly discover potential problems and respond to them. Drawing the operating data of multiple characteristic parameters on the same trend chart for comparative analysis not only improves the readability and analysis efficiency of the data, but also enhances the user's insight into the system status, providing strong support for decision-making. At the same time, interactive exploration can also be achieved, allowing users to interact with trend charts by clicking, dragging, and other operations, such as zooming in, zooming out, and filtering specific areas, in order to analyze the data more deeply.

[0060] As a preferred implementation scheme, in this embodiment, the operation data of the target power system includes real-time operation data and historical operation data, as well as real-time alarm data and historical alarm data. The operation data includes key parameters such as current, voltage, power, and temperature.

[0061] For alarm data, the dynamic display interface is also configured to use the second type of mark (such as marking the corresponding line segment in yellow or red) on the trend curve chart according to the alarm data to mark the specific time period where the alarm occurred, so that monitoring personnel can promptly grasp the changes in the system status and quickly respond to any possible problems or abnormal conditions.

[0062] As a preferred implementation, in this embodiment, the preprocessing of the operation data includes:

[0063] Clean operational data, process outliers and fill in missing values ​​to ensure data integrity and accuracy;

[0064] Normalizing the cleaned data and scaling it to a range of 0 to 1 helps speed up model training and improve its performance.

[0065] Convert alarm data into binary labels, where 1 indicates an alarm event has occurred and 0 indicates the system is operating normally. This simplifies the data structure and facilitates subsequent analysis and modeling.

[0066] As a preferred implementation, in this embodiment, the multilayer perceptron network includes:

[0067] Input layer: This layer directly receives the raw input data. If the input data is a vector containing N features, then the input layer will have N neurons, each corresponding to one feature. For the scenario of analyzing the alarm probability of the power system, these N features include key parameters such as current, voltage, power and temperature, which will be used together to analyze the probability of alarm generation;

[0068] The first hidden layer receives data from the input layer. Assuming the input layer has N neurons, the first hidden layer also contains N neurons. Each neuron performs a linear transformation on the input and uses the ReLU activation function.

[0069] The second hidden layer receives data from the first hidden layer. The first hidden layer has performed preliminary feature extraction and nonlinear transformation on the input data. The second hidden layer further extracts and combines features based on this to capture higher-order features. This multi-level feature extraction can better represent the complex structure of the data.

[0070] The output layer contains 1 neuron. After the neurons in the output layer perform linear transformation on the output of the second hidden layer, they use the Sigmoid activation function to obtain an output between 0 and 1.

[0071] See also Figure 1,When the anomaly prediction model is constructed, it will be initially trained and ,deployed and run after the model is evaluated to have good performance, and ,alarm prediction will be performed based on the input real-time data.

[0072] In an illustrative example, the initial training process is as follows:

[0073] Use the preprocessed data to train the model, using 80% of the data as the training set for training and 20% of the data as the validation set for validation. Define the TensorFlow.js early stopping callback function to monitor the loss of the validation set. Specifically, after each epoch, get the test results on the validation set and record the best validation set accuracy so far; if the validation set accuracy does not improve or the loss starts to increase within several consecutive epochs, stop training.

[0074] After the initial training model is deployed and running, when the real-time operation data is pushed, the model will calculate the possibility of future alarms based on the input data. If the calculated probability exceeds the preset threshold, the moment will be specially marked on the trend curve chart of the dynamic display interface, and an early warning notification will be issued at the same time. This can help operators to promptly notice potential risk points and take corresponding measures to avoid or reduce losses.

[0075] As a preferred implementation scheme, the construction method in this embodiment also includes gradually updating and optimizing the abnormal prediction model based on the incremental learning method of TensorFlow.js, which allows the model to continuously learn and evolve under the continuous input of new data, thereby ensuring that its prediction ability and accuracy continue to improve over time. In this way, continuous improvement of model performance can be achieved to better adapt to new data patterns and environmental changes.

[0076] In this embodiment, the following method is specifically used to update and optimize the abnormality prediction model:

[0077] Every time a real alarm occurs, the operating data within a preset time period before the real alarm moment is extracted;

[0078] Use the TensorFlow.js training method model.fit to train the anomaly prediction model based on the extracted data and specified training parameters.

[0079] Example 2

[0080] The second embodiment of the present invention provides a comprehensive power monitoring system applied to a front-end project, including:

[0081] A dynamic display interface is used to preprocess, cache and block the operating data of the target power system, and use the D3.js library to draw a dynamically updated trend curve graph. The trend curve graph uses requestAnimationFrame to perform curve drawing, and the data preprocessing and dynamic update calculation of the trend curve graph are run in Web Workers;

[0082] The anomaly prediction model, which includes a multi-layer perceptron network created using the TensorFlow.js library, is used to calculate the probability of future alarms based on the collected target power system operation data, and when the calculated probability exceeds a preset threshold, the moment when the alarm is predicted to occur is marked as the first category on the trend chart, and an early warning notification is issued at the same time.

[0083] In this embodiment, the specific construction method and operation mode of the above-mentioned dynamic display interface and abnormal prediction model can refer to the description in Example 1, and will not be elaborated here.

[0084] See also Figure 2 When the electric power integrated monitoring system constructed in this embodiment is actually deployed at the front end, the operation process is as follows:

[0085] After the configuration is completed, the user opens the dynamic display interface on the Web end, and the abnormal prediction model establishes a connection with the data service of the target power system, queries the data service for the operation history data and alarm history data of the corresponding bit number, and draws a trend curve on the dynamic display interface based on the data returned by the data service, and marks the alarm section. During the initial operation, the model also needs to be initially trained based on the received historical data, and enter operation after the initial training is completed;

[0086] During operation, the abnormal prediction model subscribes to the real-time operation data from the real-time data service of the target power system, and the real-time data service returns the subscription result; at the same time, it subscribes to the real-time alarm data from the real-time alarm service of the target power system, and the real-time alarm service returns the subscription result;

[0087] Monitor real-time operation data. After receiving the real-time operation data pushed by the real-time data service, update the trend curve chart on the dynamic display interface. At the same time, the abnormality prediction model calculates the abnormality probability. When the calculated probability exceeds the preset threshold, the first category mark is made on the trend chart for the predicted alarm moment, and an early warning notification is issued at the same time;

[0088] Listen to real-time alarm data, receive real-time alarm data pushed by the real-time alarm service, update the alarm section on the dynamic display interface, and gradually update and optimize the abnormal prediction model based on the incremental learning method of TensorFlow.js.

[0089] The embodiments of the present invention are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing an integrated power monitoring system for a front-end project, characterized in that: The steps include: A dynamic display interface is configured on the Web side of the front-end project, which is used to preprocess, cache and block the operating data of the target power system, and use the D3.js library to draw a dynamically updated trend curve chart. The trend curve chart uses requestAnimationFrame to perform curve drawing, and the data preprocessing and dynamic update calculation of the trend curve chart are run in Web Workers; An abnormality prediction model is built on the Web side of the front-end project, which includes a multi-layer perceptron network created using the TensorFlow.js library, which is used to calculate the possibility of future alarms based on the collected target power system operation data, and when the calculated probability exceeds a preset threshold, the predicted alarm moment is marked as the first category on the trend chart, and an early warning notification is issued at the same time.

2. The method for constructing a comprehensive power monitoring system for a front-end project according to claim 1, characterized in that: The operating data of the target power system includes real-time operating data and historical operating data, as well as real-time alarm data and historical alarm data. The dynamic display interface is also configured to use a second type of mark on the trend curve chart to mark the specific time period in which the alarm occurs according to the alarm data.

3. The method for constructing a comprehensive power monitoring system for a front-end project according to claim 2, characterized in that: The preprocessing of the operation data includes: Clean running data, handle outliers and fill in missing values; Normalize the cleaned data and scale them uniformly to the range of 0 to 1; The alarm data is converted into binary label form, where 1 represents an alarm event and 0 represents that the system is in normal operation.

4. The method for constructing a comprehensive power monitoring system for a front-end project according to claim 1, characterized in that: The multi-layer perceptron network comprises: An input layer, which includes N neurons, and the shape of the input data is a vector containing N features, each neuron corresponds to one feature, and the N features correspond to the preset parameters in the running data; The first hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the input layer and performs linear transformation and nonlinear activation; The second hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the first hidden layer and performs further linear transformation and nonlinear activation; The output layer, which contains 1 neuron, uses the Sigmoid activation function, receives data from the second hidden layer, and outputs a value between 0 and 1.

5. The method for constructing a comprehensive power monitoring system for a front-end project according to any one of claims 1 to 4, characterized in that: It also includes an incremental learning method based on TensorFlow.js to gradually update and optimize the abnormal prediction model, specifically using the following method: Every time a real alarm occurs, the operating data within a preset time period before the real alarm moment is extracted; Use the TensorFlow.js training method model.fit to train the anomaly prediction model based on the extracted data and specified training parameters.

6. A comprehensive power monitoring system applied to front-end projects, characterized in that: At least include the following configured on the front-end project web side: A dynamic display interface is used to preprocess, cache and block the operating data of the target power system, and use the D3.js library to draw a dynamically updated trend curve graph. The trend curve graph uses requestAnimationFrame to perform curve drawing, and the data preprocessing and dynamic update calculation of the trend curve graph are run in Web Workers; The abnormality prediction model includes a multi-layer perceptron network created using the TensorFlow.js library, which is used to calculate the possibility of future alarms based on the collected target power system operation data, and when the calculated probability exceeds a preset threshold, the first category mark is made on the trend chart at the moment when the alarm is predicted to occur, and an early warning notification is issued at the same time.

7. The integrated power monitoring system for front-end projects according to claim 6, characterized in that: The operating data of the target power system includes real-time operating data and historical operating data, as well as real-time alarm data and historical alarm data. The dynamic display interface is also used to mark the specific time period in which the alarm occurs on the trend curve chart using a second type of mark according to the alarm data.

8. The integrated power monitoring system for front-end projects according to claim 7, characterized in that: The preprocessing of the operation data includes: Clean running data, handle outliers and fill in missing values; Normalize the cleaned data and scale them uniformly to the range of 0 to 1; The alarm data is converted into binary label form, where 1 represents an alarm event and 0 represents that the system is in normal operation.

9. The integrated power monitoring system for front-end projects according to claim 6, characterized in that: The multi-layer perceptron network comprises: An input layer, which includes N neurons, and the shape of the input data is a vector containing N features, each neuron corresponds to one feature, and the N features correspond to the preset parameters in the running data; The first hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the input layer and performs linear transformation and nonlinear activation; The second hidden layer, which contains N neurons, uses the ReLU activation function, receives data from the first hidden layer and performs further linear transformation and nonlinear activation; The output layer, which contains 1 neuron, uses the Sigmoid activation function, receives data from the second hidden layer, and outputs a value between 0 and 1.

10. The integrated power monitoring system applied to a front-end project according to any one of claims 6 to 9, characterized in that: The anomaly prediction model is gradually updated and optimized based on the incremental learning method of TensorFlow.js, specifically using the following method: Every time a real alarm occurs, the operating data within a preset time period before the real alarm moment is extracted; Use the TensorFlow.js training method model.fit to train the anomaly prediction model based on the extracted data and specified training parameters.

Images