An intelligent monitoring system for a centralized control center based on real-time monitoring

Abstract

The present invention relates to the field of new energy station monitoring technology, specifically to an intelligent monitoring system of a centralized control center based on real-time monitoring, including a real-time monitoring unit, a data processing unit, a graph modeling unit, an anomaly detection unit, a centralized control unit, a distributed storage unit and an anomaly alarm unit, specifically: the real-time monitoring unit is used to perform status monitoring on a target new energy station to obtain status data of the target new energy station. The present invention monitors the status of a target new energy station in real time through a real-time monitoring unit, and can continuously and timely obtain the operating data of the station to ensure a comprehensive grasp of the station status, which helps to quickly discover anomalies and respond, and the data processing unit performs abnormal identification on the transmitted status data based on a trained abnormal data identification model, and can effectively identify and remove abnormal data to ensure the accuracy of data for subsequent analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy station monitoring, and in particular to an intelligent monitoring system of a centralized control center based on real-time monitoring. Background Art

[0002] With the development of digital transformation of power companies, the awareness of independent control of power companies' monitoring systems and power information systems has been continuously improved, the monitoring systems and equipment facilities of power companies' centralized control centers have been continuously improved, the number of new energy stations has increased sharply, and the amount of data connected to the monitoring system of the centralized control center has continued to increase. The monitoring system data security, independent control and large-scale data storage capabilities are crucial.

[0003] Traditional systems may not be able to monitor the status of new energy sites in real time and continuously, resulting in delayed detection and response to anomalies, increasing potential risks; and traditional systems often rely on manual or simple algorithms for data analysis, which can easily lead to misjudgment or missed anomalies, affecting the accuracy of decision-making; and traditional systems lack intelligent anomaly detection based on data models, and may not be able to accurately identify complex or intermittent abnormal events, increasing the probability of false alarms; and traditional systems usually lack intelligent anomaly level assessment mechanisms, which are prone to excessive alarms or missed reports, affecting operational efficiency; and data storage in traditional systems may be centralized, which is not only susceptible to single point failures, but also has low data access efficiency, increasing the difficulty and risk of system maintenance. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide an intelligent monitoring system for a centralized control center based on real-time monitoring.

[0005] The technical solution adopted to solve the above technical problems is: an intelligent monitoring system of a centralized control center based on real-time monitoring, including a real-time monitoring unit, a data processing unit, a graph modeling unit, an anomaly detection unit, a centralized control unit, a distributed storage unit and an anomaly alarm unit, specifically: The real-time monitoring unit is used to monitor the status of the target new energy station to obtain the status data of the target new energy station and transmit the status data of the target new energy station to the data processing unit; The data processing unit is used to receive the state data of the target new energy station transmitted by the real-time monitoring unit, and perform abnormal identification on the state data of the target new energy station based on the trained abnormal data identification model to obtain abnormal data, delete the abnormal data to obtain the standard state data of the target new energy station, and transmit the standard state data of the target new energy station to the spectrum modeling unit and the distributed storage unit; The graph modeling unit is used to receive the standard status data of the target new energy station transmitted by the data processing unit, and to perform statistics on the standard status data of a preset time period to obtain the standard status data at each time point within the preset time period, to construct a state graph of the preset time period based on the standard status data at each time point within the preset time period, and to transmit the state graph of the preset time period to the abnormality detection unit.

[0006] Preferably, the anomaly detection unit is used to receive the state graph of the preset time period transmitted by the graph modeling unit, and perform anomaly detection on the state graph of the preset time period based on a trained anomaly detection model to obtain the abnormal nodes corresponding to the state graph of the preset time period, that is, to obtain the abnormal time points within the preset time period, and transmit the abnormal time points within the preset time period to the centralized control unit.

[0007] Preferably, the distributed storage unit receives the standard status data of the target new energy station transmitted by the data processing unit, and stores the standard status data of the target new energy station in a preset distributed storage network based on the resource data of the preset distributed storage network, wherein the distributed storage network includes multiple storage nodes.

[0008] Preferably, the centralized control unit is used to receive the abnormal time points within the preset time period transmitted by the abnormality detection unit, and obtain a control command list corresponding to the abnormal time points within the preset time period, and obtain a standard status data list corresponding to the abnormal time points within the preset time period from the distributed storage network, obtain an average value list of historical standard status data corresponding to the control command list, obtain the abnormality level corresponding to the control command list corresponding to the abnormal time points within the preset time period based on the difference between the average value list of historical standard status data and the standard status data list, generate a corresponding abnormal alarm signal based on the abnormal level, and transmit the abnormal alarm signal to the abnormal alarm unit.

[0009] Preferably, the abnormal alarm unit is used to receive the abnormal alarm signal transmitted by the centralized control unit, and to issue an abnormal alarm based on the abnormal alarm signal using an audible and visual alarm device.

[0010] Preferably, the state graph includes multiple state nodes and multiple state connection edges, wherein the state nodes correspond to the standard state data at each time point, and the state connection edges correspond to the correlation between two state nodes, wherein the correlation between the state nodes is the correlation of the time points corresponding to the state nodes, and wherein the correlation of the time points is determined by a preset time window.

[0011] Preferably, the real-time monitoring unit includes a unit monitoring module, a photovoltaic monitoring module, an energy storage monitoring module and a meteorological monitoring module; The unit monitoring module is used to monitor the status of multiple wind turbines in the target new energy station to obtain status data of the multiple wind turbines; The photovoltaic monitoring module is used to monitor the status of multiple photovoltaic components in the target new energy station to obtain status data of the multiple photovoltaic components; The energy storage monitoring module is used to monitor the status of multiple energy storage devices in the target new energy station to obtain status data of the multiple energy storage devices; The meteorological monitoring module is used to perform meteorological monitoring on the target new energy station to obtain meteorological data of the target new energy station.

[0012] Preferably, the abnormal data identification model adopts an improved BP neural network, and the BP neural network is optimized based on the firefly algorithm, wherein the BP neural network is optimized based on the firefly algorithm, including: A group of firefly individuals are randomly generated, and the position of each firefly individual represents the combination of each weight and bias vector of the BP neural network; For each combination of weights and bias values ​​represented by an individual firefly, the output of the BP neural network is calculated by forward propagation, and the brightness is calculated according to the error between the output of the BP neural network and the actual target, wherein the brightness is a mean square error; Updating the light intensity and attractiveness of the individual fireflies, and updating the position of the individual fireflies based on the light intensity and attractiveness of the individual fireflies; Determine whether the training error reaches the convergence value or the number of iterations reaches the maximum value. If the training error reaches the convergence value and the number of iterations reaches the maximum value, the iteration ends and the weight and bias value combination of the neural network is output, otherwise the number of iterations is increased by 1.

[0013] Preferably, the update formula of the light intensity of the individual fireflies is as follows: ; in, and Indicates The first iteration and the The iteration The light intensity of a firefly individual, represents the specific gravity of fluorescein, represents the light intensity absorption coefficient, Indicates The position of the individual fireflies at the iteration, Indicates The brightness corresponding to the position of the individual fireflies at the iteration; The update formula of the attractiveness of the firefly individual is as follows: ; in, represents the attractiveness of individual fireflies, represents the maximum attraction, Indicates Firefly individuals and The distance between individual fireflies, and , and Indicates Firefly individuals and The location of individual fireflies; The update formula of the position of the firefly individual is as follows: ; in, and Indicates The first iteration and the The iteration The position of individual fireflies, Indicates The iteration The position of individual fireflies, Indicates The random factor of firefly individuals, Represents the step size scaling factor.

[0014] Preferably, the anomaly detection model adopts a graph neural network, and the anomaly detection model includes a graph convolutional network, a convolutional network, a hybrid module and an anomaly judgment module. The graph convolutional network is used to perform convolution operations on the feature matrix and the adjacency matrix of the state spectrum of the preset time period to obtain the time correlation characteristics of the standard state data. The convolutional network is used to perform convolution operations on the state feature matrix of the standard state data of the preset time period to obtain the periodic characteristics of the standard state data. The hybrid module is used to perform feature fusion on the time correlation characteristics and periodic characteristics of the standard state data to obtain an abnormal probability result. The abnormal judgment module is used to judge whether the preset standard state data is abnormal based on the abnormal probability result.

[0015] Preferably, the graph convolution network includes an input layer, a first graph convolution layer, a second graph convolution layer, a first fully connected layer and an output layer in sequence, and the convolution network includes a first convolution layer, a spatial attention layer, a second convolution layer, a pooling layer and a second fully connected layer in sequence, wherein the spatial attention layer performs average pooling and maximum pooling on the input features to obtain an average pooling vector and a maximum pooling vector, the average pooling vector and the maximum pooling vector are processed by a convolution layer and a Sigmoid activation function to obtain a spatial attention output weight, the spatial attention output weight is multiplied by the input features by the same element to obtain a spatial attention vector, the mixing module splices the time correlation feature and the periodic feature to obtain a spliced ​​vector feature, the fully connected layer based on one neuron extracts the time series feature of the spliced ​​vector feature, and the abnormal probability of the time series feature is output by the Sigmoid activation function.

[0016] Preferably, the expression of the first graph convolutional layer is as follows: ; in, represents the feature matrix of the first graph convolutional layer, represents the activation function, The adjacency matrix representing the state graph for a preset period of time, The characteristic matrix representing the state spectrum of a preset period, represents the weight matrix of the first graph convolutional layer, represents the bias vector of the first graph convolutional layer; The expression of the second graph convolutional layer is as follows: ; in, represents the feature matrix of the second graph convolutional layer, represents the weight matrix of the second graph convolutional layer, Represents the bias vector of the second graph convolutional layer; The expression of the first fully connected layer is as follows: ; in, represents the feature matrix of the first fully connected layer, represents the weight matrix of the first fully connected layer, represents the bias vector of the first fully connected layer; The calculation formula of the abnormal probability is as follows: ; in, represents the predicted label probability, represents the time-dependent characteristics, Represents periodic characteristics, represents the weight matrix of the hybrid module, represents the bias vector of the hybrid module, Represents the concatenated vector feature.

[0017] The beneficial effects of the present invention are as follows: (1) The present invention uses a real-time monitoring unit to monitor the status of the target new energy station in real time, and can continuously and timely obtain the station's operating data to ensure a comprehensive grasp of the station status, which helps to quickly discover anomalies and respond. The data processing unit uses a trained abnormal data recognition model to identify abnormalities in the transmitted status data, and can effectively identify and remove abnormal data to ensure the accuracy of data for subsequent analysis. This data cleaning process can avoid misjudgment and improve the reliability of the system. (2) The present invention uses a graph modeling unit, and the system can construct a state graph based on standard state data within a preset time period. This graph representation not only helps to fully understand the system's status at different time points, but also provides clear basic data for subsequent anomaly detection. The abnormal detection unit uses the state graph for analysis, which can accurately detect abnormal time points within the preset time period, ensuring that abnormal events can be identified in the shortest time possible. This graph-based anomaly detection method can reduce false alarms and improve detection accuracy; (3) After the centralized control unit receives the abnormal time point, the present invention can evaluate the severity of the abnormality based on the difference between the historical standard status data and the control command list, and generate a corresponding abnormal alarm signal based on this. This intelligent abnormal level evaluation mechanism enables the system to issue appropriate responses to abnormalities of different degrees, avoiding excessive alarms or missed alarms, and stores the standard status data in the distributed storage network through the distributed storage unit to ensure high availability and security of the data. At the same time, the distributed storage architecture can improve data access efficiency and reduce the risk of data loss. The abnormal alarm unit triggers the sound and light alarm device according to the abnormal alarm signal, which can quickly issue an alarm to relevant personnel to ensure that the operator responds in time and takes corresponding measures. This mechanism improves the emergency response capability of the system and reduces the risk of potential losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of the system architecture of the overall system in an embodiment of the present invention; Figure 2 This is a block diagram of the internal modules of a real-time monitoring unit in an embodiment of the present invention.

[0019] Figure numerals: 1. Real-time monitoring unit; 101. Unit monitoring module; 102. Photovoltaic monitoring module; 103. Energy storage monitoring module; 104. Meteorological monitoring module; 2. Data processing unit; 3. Spectrum modeling unit; 4. Anomaly detection unit; 5. Centralized control unit; 6. Distributed storage unit; 7. Abnormal alarm unit. DETAILED DESCRIPTION

[0020] Embodiment 1, as Figure 1 As shown, the present invention proposes an intelligent monitoring system for a centralized control center based on real-time monitoring, including a real-time monitoring unit 1, a data processing unit 2, a graph modeling unit 3, an anomaly detection unit 4, a centralized control unit 5, a distributed storage unit 6 and an anomaly alarm unit 7, specifically: The real-time monitoring unit 1 is used to monitor the state of the target new energy station to obtain the state data of the target new energy station and transmit the state data of the target new energy station to the data processing unit 2; The data processing unit 2 is used to receive the state data of the target new energy station transmitted by the real-time monitoring unit 1, and perform abnormal identification on the state data of the target new energy station based on the trained abnormal data identification model to obtain abnormal data, delete the abnormal data to obtain the standard state data of the target new energy station, and transmit the standard state data of the target new energy station to the spectrum modeling unit 3 and the distributed storage unit 6; The spectrum modeling unit 3 is used to receive the standard status data of the target new energy station transmitted by the data processing unit 2, and to count the standard status data of a preset time period to obtain the standard status data at each time point within the preset time period, and to construct a state spectrum of the preset time period based on the standard status data at each time point within the preset time period, and to transmit the state spectrum of the preset time period to the abnormality detection unit 4.

[0021] In the present invention, the target new energy station refers to a new energy power generation station that needs to be monitored and data analyzed, such as a wind farm, a solar power station, etc. These stations use renewable energy (such as wind energy, solar energy, etc.) to generate electricity, so their operating efficiency and status need to be tracked in real time through intelligent monitoring; status data refers to data that describes the working conditions of each device or system in the new energy station at a specific time point; the abnormal data identification model is a model trained by machine learning or data analysis technology, which is used to identify abnormal data that may exist in the target new energy station. The model can determine whether the status data at a certain moment is abnormal based on historical data and pattern recognition methods; the preset time period refers to a specific time range when the system is set, for example, this can be a working period of a certain day, an operating cycle of a certain month, or a specific time period for analyzing the performance of the equipment; the state map is a graphical representation constructed by visualizing the standard state data at each time point in the preset time period. The state map is usually used to show the state changes of the equipment or system throughout the monitoring cycle, helping engineers or operators to quickly understand the trends and characteristics of the system operation.

[0022] In an optional embodiment, the anomaly detection unit 4 is used to receive the state graph of a preset time period transmitted by the graph modeling unit 3, and perform anomaly detection on the state graph of the preset time period based on a trained anomaly detection model to obtain the abnormal nodes corresponding to the state graph of the preset time period, that is, to obtain the abnormal time points within the preset time period, and transmit the abnormal time points within the preset time period to the centralized control unit 5.

[0023] It should be noted that anomaly detection refers to identifying data points or events that deviate from normal patterns or expected results by analyzing data. The goal of anomaly detection is to discover potential faults, problems or abnormal conditions that may affect the stability, efficiency or safety of the system; in the state map, abnormal nodes refer to those time points or data points that are identified as abnormal. These nodes represent situations in which the operating status of the equipment or system deviates from the normal situation within a preset time period. Abnormal nodes may reflect equipment failure, system failure or other types of abnormal conditions; abnormal time points refer to specific time points identified as abnormal by the anomaly detection model within a certain period of time. The status data corresponding to these time points are significantly different from the historical normal state, which may indicate equipment failure or system abnormality.

[0024] In an optional embodiment, the distributed storage unit 6 receives the standard status data of the target new energy station transmitted by the data processing unit 2, and stores the standard status data of the target new energy station in a preset distributed storage network based on the resource data of the preset distributed storage network, wherein the distributed storage network includes multiple storage nodes.

[0025] It should be noted that a distributed storage network is a storage system that stores data in multiple physical locations or nodes. It improves data reliability, availability and scalability by distributing data to multiple nodes (such as servers, storage devices, etc.). The preset distributed storage network refers to the storage architecture determined in the design phase to ensure that data can be distributed to different storage nodes according to specific needs; resource data refers to the computing resources and storage resources of the distributed storage network; storage nodes are the basic units in the distributed storage system, usually referring to the servers or devices responsible for storing data. In the distributed storage system, each storage node stores a part of the data, and multiple nodes work together to provide large-capacity, high-reliability data storage services.

[0026] In an optional embodiment, the centralized control unit 5 is used to receive the abnormal time points within a preset time period transmitted by the abnormality detection unit 4, and obtain a control command list corresponding to the abnormal time points within the preset time period, and obtain a standard status data list corresponding to the abnormal time points within the preset time period from the distributed storage network, obtain an average value list of historical standard status data corresponding to the control command list, obtain the abnormality level corresponding to the control command list corresponding to the abnormal time points within the preset time period based on the difference between the average value list of historical standard status data and the standard status data list, generate a corresponding abnormal alarm signal based on the abnormal level, and transmit the abnormal alarm signal to the abnormal alarm unit 7.

[0027] It should be noted that the control command list refers to a collection of corresponding operation commands issued by the system according to different abnormal situations when an abnormality occurs. These commands may include device switch control, parameter adjustment, system restart, etc., which are used to deal with the abnormal state of the system; historical standard status data refers to the standard status data of the equipment in the past period of time. By accumulating and analyzing historical data, it can provide a reference benchmark for future equipment operation; the average value list refers to a list formed by averaging a series of historical data values, which is usually used to describe the average level of a parameter in a certain period of time, which helps to evaluate the normal range and trend of the data.

[0028] In an optional embodiment, the abnormal alarm unit 7 is used to receive the abnormal alarm signal transmitted by the centralized control unit 5, and to issue an abnormal alarm according to the sound and light alarm device based on the abnormal alarm signal.

[0029] It should be noted that the sound and light alarm equipment is a device that sends out alarms through sound and light signals at the same time. The purpose is to attract people's attention and alert people that abnormal conditions have occurred in the system or equipment. These devices usually include alarm bells, alarm lights, flashing lights, buzzers, etc., combined with the dual reminders of sound and visual signals to ensure that the alarm information can be effectively conveyed in different environments.

[0030] Embodiment 2, as Figure 2 As shown, the present invention proposes an intelligent monitoring system for a centralized control center based on real-time monitoring. Compared with the first embodiment, this embodiment also includes: the state graph includes multiple state nodes and multiple state connection edges, wherein the state node corresponds to the standard state data at each time point, and the state connection edge corresponds to the correlation between two state nodes, wherein the correlation between the state nodes is the correlation of the time points corresponding to the state nodes, wherein the correlation of the time points is determined by a preset time window.

[0031] In this embodiment, the correlation of time points is determined by a preset time window, specifically: based on the preset time window, it is determined whether the time points are in the same time window. If so, it means that the time points are correlated, otherwise, it means that the time points are not correlated.

[0032] In an optional embodiment, the real-time monitoring unit 1 includes a unit monitoring module 101, a photovoltaic monitoring module 102, an energy storage monitoring module 103 and a meteorological monitoring module 104; The unit monitoring module 101 is used to monitor the status of multiple wind turbines in the target new energy station to obtain status data of the multiple wind turbines; The photovoltaic monitoring module 102 is used to monitor the status of multiple photovoltaic components in the target new energy station to obtain status data of the multiple photovoltaic components; The energy storage monitoring module 103 is used to monitor the status of multiple energy storage devices in the target new energy station to obtain status data of the multiple energy storage devices; The weather monitoring module 104 is used to perform weather monitoring on the target new energy station to obtain weather data of the target new energy station.

[0033] It should be noted that the unit monitoring module is mainly used to monitor the working status of the wind turbine. It collects and analyzes the operating data of the wind turbine, such as power generation, rotation speed, wind speed, temperature, etc., to ensure the normal operation of the wind turbine; the photovoltaic monitoring module is used to monitor the working status of photovoltaic components (solar panels), which convert solar energy into electrical energy. The monitoring module usually includes monitoring indicators such as the output power, temperature, and light intensity of photovoltaic components; the energy storage monitoring module is used to monitor the status and performance of energy storage devices (such as batteries, supercapacitors, etc.). It collects information such as voltage, current, charge and discharge times, and remaining power of the energy storage device; the meteorological monitoring module is used to monitor the meteorological data of the area where the target new energy station is located. Common meteorological data include wind speed, wind direction, temperature, humidity, air pressure, precipitation, etc., which have an important impact on wind power generation and photovoltaic power generation.

[0034] In an optional embodiment, the abnormal data identification model adopts an improved BP neural network, and the BP neural network is optimized based on the firefly algorithm, wherein the optimization of the BP neural network based on the firefly algorithm includes: A group of firefly individuals are randomly generated, and the position of each firefly individual represents the combination of each weight and bias vector of the BP neural network; For each combination of weights and bias values ​​represented by an individual firefly, the output of the BP neural network is calculated through forward propagation, and the brightness is calculated based on the error between the output of the BP neural network and the actual target, where the brightness is the mean square error; Update the light intensity and attractiveness of individual fireflies, and update the position of individual fireflies based on the light intensity and attractiveness of individual fireflies; Determine whether the training error reaches the convergence value or the number of iterations reaches the maximum value. If the training error reaches the convergence value and the number of iterations reaches the maximum value, the iteration ends and the weight and bias value combination of the neural network is output, otherwise the number of iterations is increased by 1.

[0035] It should be noted that the BP neural network is a common artificial neural network, which is trained through the back propagation algorithm and is used to solve problems such as pattern recognition and regression analysis. Its basic principle is to calculate the error between the network output and the actual target, and adjust the weights of each node (neuron) in the network by back propagating the error; the firefly algorithm is an optimization algorithm that simulates the behavior of fireflies. It belongs to the swarm intelligence algorithm. Its basic idea is that each "firefly" individual has attractiveness (brightness). The higher the brightness of the firefly, the stronger its attraction to other individuals. Through iterative optimization, the brightness of the group gradually approaches the optimal solution; forward propagation is the process of data transmission in the neural network. The input data is calculated by the neurons of each layer and passed to the output layer to finally get the prediction result.

[0036] In an optional embodiment, the update formula of the light intensity of individual fireflies is as follows: ; in, and Indicates The first iteration and the The iteration The light intensity of a firefly individual, represents the specific gravity of fluorescein, represents the light intensity absorption coefficient, Indicates The position of the individual fireflies at the iteration, Indicates The brightness corresponding to the position of the individual firefly at the iteration; The update formula of the attractiveness of individual fireflies is as follows: ; in, represents the attractiveness of individual fireflies, represents the maximum attraction, Indicates Firefly individuals and The distance between individual fireflies, and , and Indicates Firefly individuals and The location of individual fireflies; The update formula for the position of individual fireflies is as follows: ; in, and Indicates The first iteration and the The iteration The position of individual fireflies, Indicates The iteration The position of individual fireflies, Indicates The random factor of firefly individuals, Represents the step size scaling factor.

[0037] In an optional embodiment, the anomaly detection model adopts a graph neural network, and the anomaly detection model includes a graph convolutional network, a convolutional network, a hybrid module and an anomaly judgment module. The graph convolutional network is used to perform convolution operations on the feature matrix and adjacency matrix of the state spectrum of a preset time period to obtain the time correlation characteristics of the standard state data. The convolutional network is used to perform convolution operations on the state feature matrix of the standard state data of the preset time period to obtain the periodic characteristics of the standard state data. The hybrid module is used to perform feature fusion on the time correlation characteristics and periodic characteristics of the standard state data to obtain an abnormal probability result. The abnormal judgment module is used to judge whether the preset standard state data is abnormal based on the abnormal probability result.

[0038] It should be noted that graph neural networks are a type of neural networks used to process graph-structured data. In the graph, nodes represent entities and edges represent the relationships between entities. Graph convolutional networks are an important variant of graph neural networks, specifically used for learning graph-structured data. GCN uses convolution operations (similar to the convolution operations in traditional convolutional neural networks) to fuse the feature information of nodes and the information of adjacent nodes to generate richer node representations. CNN extracts local features through operations such as convolutional layers and pooling layers, and performs high-level feature learning through fully connected layers. Convolution operations can capture spatial or temporal features in input data. Time correlation features refer to the time dependency in data, which means the relationship or influence between the data at the current time point and the data at the previous and subsequent time points in time series data. Periodic features refer to regular fluctuations or periodic changes in data.

[0039] In an optional embodiment, a graph convolutional network includes an input layer, a first graph convolutional layer, a second graph convolutional layer, a first fully connected layer and an output layer in sequence, and the convolutional network includes a first convolutional layer, a spatial attention layer, a second convolutional layer, a pooling layer and a second fully connected layer in sequence, wherein the spatial attention layer performs average pooling and maximum pooling on the input features to obtain an average pooling vector and a maximum pooling vector, the average pooling vector and the maximum pooling vector are processed by the convolutional layer and the Sigmoid activation function to obtain a spatial attention output weight, the spatial attention output weight is multiplied by the input features by the same element to obtain a spatial attention vector, the mixing module splices the time correlation features and the periodic features to obtain a spliced ​​vector feature, the fully connected layer based on a neuron extracts the temporal features of the spliced ​​vector features, and the abnormal probability of the temporal features is output by the Sigmoid activation function.

[0040] In an optional embodiment, the expression of the first graph convolutional layer is as follows: ; in, represents the feature matrix of the first graph convolutional layer, represents the activation function, The adjacency matrix representing the state graph for a preset period of time, The characteristic matrix representing the state spectrum of a preset period, represents the weight matrix of the first graph convolutional layer, represents the bias vector of the first graph convolutional layer; The expression of the second graph convolutional layer is as follows: ; in, represents the feature matrix of the second graph convolutional layer, represents the weight matrix of the second graph convolutional layer, Represents the bias vector of the second graph convolutional layer; The expression of the first fully connected layer is as follows: ; in, represents the feature matrix of the first fully connected layer, represents the weight matrix of the first fully connected layer, represents the bias vector of the first fully connected layer; The calculation formula of abnormal probability is as follows: ; in, represents the predicted label probability, represents the time-dependent characteristics, Represents periodic characteristics, represents the weight matrix of the hybrid module, represents the bias vector of the hybrid module, Represents the concatenated vector feature.

[0041] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. An intelligent monitoring system for a centralized control center based on real-time monitoring, characterized in that: The system comprises a real-time monitoring unit (1), a data processing unit (2), a graph modeling unit (3), an anomaly detection unit (4), a centralized control unit (5), a distributed storage unit (6) and an anomaly alarm unit (7), specifically: The real-time monitoring unit (1) is used to monitor the status of the target new energy station to obtain status data of the target new energy station, and transmit the status data of the target new energy station to the data processing unit (2); The data processing unit (2) is used to receive the state data of the target new energy station transmitted by the real-time monitoring unit (1), and to perform abnormality identification on the state data of the target new energy station based on a trained abnormal data identification model to obtain abnormal data, delete the abnormal data to obtain standard state data of the target new energy station, and transmit the standard state data of the target new energy station to the spectrum modeling unit (3) and the distributed storage unit (6); The spectrum modeling unit (3) is used to receive the standard status data of the target new energy station transmitted by the data processing unit (2), and to perform statistics on the standard status data of a preset period of time to obtain the standard status data at each time point within the preset period of time, to construct a state spectrum of the preset period of time based on the standard status data at each time point within the preset period of time, and to transmit the state spectrum of the preset period of time to the abnormality detection unit (4).

2. According to claim 1, the intelligent monitoring system of the centralized control center based on real-time monitoring is characterized in that: The anomaly detection unit (4) is used to receive the state graph of the preset time period transmitted by the graph modeling unit (3), and perform anomaly detection on the state graph of the preset time period based on the trained anomaly detection model to obtain an abnormal node corresponding to the state graph of the preset time period, that is, to obtain an abnormal time point within the preset time period, and transmit the abnormal time point within the preset time period to the centralized control unit (5); The distributed storage unit (6) receives the standard status data of the target new energy station transmitted by the data processing unit (2), and stores the standard status data of the target new energy station in a preset distributed storage network based on resource data of the preset distributed storage network, wherein the distributed storage network includes a plurality of storage nodes.

3. According to claim 2, the intelligent monitoring system of the centralized control center based on real-time monitoring is characterized in that: The centralized control unit (5) is used to receive the abnormal time point within the preset time period transmitted by the abnormality detection unit (4), and obtain a control command list corresponding to the abnormal time point within the preset time period, and obtain a standard status data list corresponding to the abnormal time point within the preset time period from the distributed storage network, obtain an average value list of historical standard status data corresponding to the control command list, obtain an abnormality level corresponding to the control command list corresponding to the abnormal time point within the preset time period based on the difference between the average value list of historical standard status data and the standard status data list, generate a corresponding abnormal alarm signal based on the abnormal level, and transmit the abnormal alarm signal to the abnormal alarm unit (7); The abnormal alarm unit (7) is used to receive the abnormal alarm signal transmitted by the centralized control unit (5), and to issue an abnormal alarm based on the abnormal alarm signal using the sound and light alarm device.

4. According to claim 1, the intelligent monitoring system of the centralized control center based on real-time monitoring is characterized in that: The state graph includes multiple state nodes and multiple state connection edges, wherein the state nodes correspond to the standard state data at each time point, and the state connection edges correspond to the correlation between two state nodes, wherein the correlation between the state nodes is the correlation of the time points corresponding to the state nodes, and wherein the correlation of the time points is determined by a preset time window.

5. According to claim 1, the intelligent monitoring system of the centralized control center based on real-time monitoring is characterized in that: The real-time monitoring unit (1) comprises a unit monitoring module (101), a photovoltaic monitoring module (102), an energy storage monitoring module (103) and a meteorological monitoring module (104); The wind turbine monitoring module (101) is used to monitor the status of multiple wind turbines in the target new energy station to obtain status data of the multiple wind turbines; The photovoltaic monitoring module (102) is used to monitor the status of multiple photovoltaic components in the target new energy station to obtain status data of the multiple photovoltaic components; The energy storage monitoring module (103) is used to monitor the status of multiple energy storage devices in the target new energy station to obtain status data of the multiple energy storage devices; The meteorological monitoring module (104) is used to perform meteorological monitoring on the target new energy station to obtain meteorological data of the target new energy station.

6. The intelligent monitoring system of a centralized control center based on real-time monitoring according to claim 1 is characterized in that: The abnormal data identification model adopts an improved BP neural network, and the BP neural network is optimized based on the firefly algorithm, wherein the BP neural network is optimized based on the firefly algorithm, including: A group of firefly individuals are randomly generated, and the position of each firefly individual represents the combination of each weight and bias vector of the BP neural network; For each combination of weights and bias values ​​represented by an individual firefly, the output of the BP neural network is calculated by forward propagation, and the brightness is calculated according to the error between the output of the BP neural network and the actual target, wherein the brightness is a mean square error; Updating the light intensity and attractiveness of the individual fireflies, and updating the position of the individual fireflies based on the light intensity and attractiveness of the individual fireflies; Determine whether the training error reaches the convergence value or the number of iterations reaches the maximum value. If the training error reaches the convergence value and the number of iterations reaches the maximum value, the iteration ends and the weight and bias value combination of the neural network is output, otherwise the number of iterations is increased by 1.

7. The intelligent monitoring system of a centralized control center based on real-time monitoring according to claim 6 is characterized in that: The update formula of the light intensity of the firefly individual is as follows: ； in, and Indicates The first iteration and the The first iteration The light intensity of a firefly individual, represents the specific gravity of fluorescein, represents the light intensity absorption coefficient, Indicates The position of the individual fireflies at the iteration, Indicates The brightness corresponding to the position of the individual firefly at the iteration; The update formula of the attractiveness of the firefly individual is as follows: ； in, represents the attractiveness of individual fireflies, represents the maximum attraction, Indicates Firefly individuals and The distance between individual fireflies, and , and Indicates Firefly individuals and The location of individual fireflies; The update formula of the position of the firefly individual is as follows: ； in, and Indicates The first iteration and the The first iteration The position of individual fireflies, Indicates The first iteration The position of individual fireflies, Indicates The random factor of firefly individuals, Represents the step size scaling factor.

8. The intelligent monitoring system of a centralized control center based on real-time monitoring according to claim 2 is characterized in that: The anomaly detection model adopts a graph neural network, and the anomaly detection model includes a graph convolutional network, a convolutional network, a hybrid module and an anomaly judgment module. The graph convolutional network is used to perform convolution operations on the feature matrix and the adjacency matrix of the state graph of the preset time period to obtain the time correlation characteristics of the standard state data. The convolutional network is used to perform convolution operations on the state feature matrix of the standard state data of the preset time period to obtain the periodic characteristics of the standard state data. The hybrid module is used to perform feature fusion on the time correlation characteristics and periodic characteristics of the standard state data to obtain an abnormal probability result. The abnormal judgment module is used to judge whether the preset standard state data is abnormal according to the abnormal probability result.

9. The intelligent monitoring system of a centralized control center based on real-time monitoring according to claim 8 is characterized in that: The graph convolutional network includes an input layer, a first graph convolutional layer, a second graph convolutional layer, a first fully connected layer and an output layer in sequence. The convolutional network includes a first convolutional layer, a spatial attention layer, a second convolutional layer, a pooling layer and a second fully connected layer in sequence. The spatial attention layer performs average pooling and maximum pooling on the input features to obtain an average pooling vector and a maximum pooling vector. The average pooling vector and the maximum pooling vector are processed by a convolutional layer and a Sigmoid activation function to obtain a spatial attention output weight. The spatial attention output weight is multiplied by the same-position element with the input feature to obtain a spatial attention vector. The mixing module splices the time correlation feature and the periodic feature to obtain a spliced ​​vector feature. The fully connected layer based on one neuron extracts the time series feature of the spliced ​​vector feature, and the abnormal probability of the time series feature is output by the Sigmoid activation function.

10. The intelligent monitoring system of a centralized control center based on real-time monitoring according to claim 9, characterized in that: The expression of the first graph convolutional layer is as follows: ； in, represents the feature matrix of the first graph convolutional layer, represents the activation function, The adjacency matrix representing the state graph for a preset period of time, The characteristic matrix representing the state spectrum of a preset period, represents the weight matrix of the first graph convolutional layer, represents the bias vector of the first graph convolutional layer; The expression of the second graph convolutional layer is as follows: ； in, represents the feature matrix of the second graph convolutional layer, represents the weight matrix of the second graph convolutional layer, Represents the bias vector of the second graph convolutional layer; The expression of the first fully connected layer is as follows: ； in, represents the feature matrix of the first fully connected layer, represents the weight matrix of the first fully connected layer, represents the bias vector of the first fully connected layer; The calculation formula of the abnormal probability is as follows: ； in, represents the predicted label probability, represents the time-dependent characteristics, Represents periodic characteristics, represents the weight matrix of the hybrid module, represents the bias vector of the hybrid module, Represents the concatenated vector feature.

Images