Harbor equipment green power system risk early warning method and system

Abstract

The invention discloses a port equipment green power system risk early warning method and system, and belongs to the technical field of risk assessment. Comprising the steps of collecting operation data related to new energy in a port equipment green power system; the method comprises the following steps of: constructing a risk map by taking port equipment as nodes and energy transmission and operation coordination relationships among the port equipment as edges, and endowing each node and each edge with characteristics by utilizing acquired operation data; constructing a graph convolutional network model, processing the risk graph data, and marking potential risk nodes; and calculating a risk level for the potential risk node, and implementing a layered alarm system. According to the invention, the limitation that a traditional method is only used for monitoring single equipment or single energy is overcome. And the collaborative operation intensity coefficient and the energy transmission loss coefficient between equipment in the port scene are considered in learning and analyzing the risk map in the graph convolutional network, so that the potential risk nodes can be identified more accurately, and the accuracy and reliability of risk identification are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of risk assessment, and in particular to a risk early warning method and system for a green power system of port equipment. Background Art

[0002] Currently, port equipment power systems are transitioning to greener power systems, widely incorporating new energy sources such as wind, solar, hydrogen, and electricity. However, significant deficiencies exist in risk early warning. Energy supply monitoring is flawed. Wind and solar energy are significantly affected by natural conditions, and existing monitoring methods fail to accurately assess the risks posed by their unstable supply. In the event of severe weather, the impact of energy supply disruptions on equipment operations cannot be promptly assessed. Furthermore, risk assessment of equipment coordination is lagging. Port operations require the coordination of multiple types of equipment, such as quayside container cranes, port tugboats, and automated guided vehicles (AGVs), which involve complex energy transmission and operational coordination. Existing dispatching and monitoring systems often focus on a single piece of equipment or a specific link, failing to fully consider overall coordination. A single piece of equipment failure can easily trigger a chain reaction, impacting operational processes and port efficiency. Furthermore, risk monitoring methods are relatively outdated, relying primarily on manual inspections and simple sensors. Manual inspections are inefficient, subjective, and time-consuming, making it difficult to detect potential faults in real time. Simple sensors, however, provide limited parameters, making comprehensive analysis of equipment status difficult. At the same time, existing technologies are unable to efficiently process and analyze the massive amounts of multi-source, heterogeneous data generated by the system, making it difficult to extract valuable information for risk warning. Furthermore, existing risk warning models are mostly static models based on historical data, making them difficult to adapt to complex and changing operating environments. Algorithms typically only consider a single factor, ignoring the interactions among multiple factors. This leads to inaccurate warnings, an inability to provide effective decision-making support for port operations management, and a serious impact on the safe and stable operation of the system. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention constructs the port equipment and the energy transmission and operation coordination relationships between them into a risk map, and uses a graph convolutional network to learn and analyze the risk map, accurately identifying potential risk nodes, and improving the accuracy and reliability of risk identification.

[0004] In order to achieve the above-mentioned object, the present invention provides a risk early warning method for a green power system of port equipment, comprising the following steps:

[0005] (1) Collect and pre-process the operating data related to new energy in the green power system of port equipment;

[0006] (2) Taking port equipment as nodes and the energy transmission and operation coordination relationships between port equipment as edges, a risk map is constructed, and the collected operation data is used to assign features to each node and edge;

[0007] (3) Constructing a graph convolutional network model to process the risk map data and mark potential risk nodes;

[0008] (4) Calculate the risk level for the potential risk nodes and implement a hierarchical alarm system.

[0009] Furthermore, the preprocessing includes:

[0010] (1.1) Remove outliers and erroneous data from the collected operating data related to new energy in the green power system of port equipment;

[0011] (1.2) Perform consistency check on data from different data sources;

[0012] (1.3) Standardize and normalize all types of data.

[0013] Furthermore, the step (2) includes:

[0014] (2.1) Taking port equipment as nodes and the energy transmission, operation coordination and physical connection relationships between port equipment as edges, a risk map is constructed;

[0015] G b =(V b ,E b )

[0016] Where: V b is the node set, E b is the edge set;

[0017] (2.2) For time series operating data, a sliding window is used for feature extraction. The extracted features include mean, variance, maximum, minimum, trend and cycle.

[0018] (2.3) For non-time series operating data, use its original attributes or perform normalization and standardization as the extracted features;

[0019] (2.4) Assign values ​​to each node and edge based on the features extracted in steps (2.2) and (2.3).

[0020] Furthermore, the step (3) is specifically as follows:

[0021] (3.1) Build a graph convolutional network model and construct an embedded representation of the node based on its feature information. The embedded representation of node i in the l+1 layer is:

[0022]

[0023] Where: N i Represents node v i The set of neighbor nodes of is node v i The embedding representation of the neighbor nodes at layer l; W l is the learnable weight matrix of layer l; b l is the bias term; σ is the activation function; α i,j is the intensity coefficient of collaborative operation between equipments; β i,j is the energy transmission loss coefficient; j is the neighbor node of node i; and Nodes i and j are The degree of In order to introduce the self-loop and symmetrically normalized adjacency matrix based on the energy transmission and operation coordination relationship between port equipment, A is the adjacency matrix;

[0024] (3.2) Inputting the risk map data into the trained graph convolutional network model to obtain the embedded representation of each node, and calculating the distance between the embedded representation of each node and the embedded representation of the normal behavior pattern. If the distance exceeds a set threshold, the node is marked as a potential risk node;

[0025]

[0026] Where: τ is the set threshold; Indicates that after multi-layer graph convolution operations, node v i In the last layer, i.e. the embedding representation of layer L; H′ represents the embedding representation of the normal behavior pattern, i.e. the mean of the embedding representations of all normal nodes.

[0027] Furthermore, the embedding representation of the normal behavior pattern is specifically:

[0028] Use historical fault-free graph data to input the graph convolutional network, obtain the embedded representation of all nodes after multi-layer propagation, and calculate its mean as the embedded representation of the normal behavior pattern;

[0029] As new graph data is input into the graph convolutional network, the embedded representations of non-potential risk nodes are obtained, and the mean of the embedded representations of all nodes obtained from the original historical fault-free graph data is recalculated, or the mean is calculated using a sliding window as the embedded representation of the updated normal behavior pattern.

[0030] Furthermore, the step (4) is specifically as follows:

[0031] (4.1) For the potential risk node marked in step (3), calculate the mean μ and standard deviation s of the running data related to the node and obtain the z-score of each sample point;

[0032] z i =(x i -μ) / s

[0033] If |z i |≥M, then it is determined to be an abnormal point Where M is the set judgment threshold;

[0034] (4.2) Use the dynamic time warping algorithm to calculate the similarity S between the time series running data of the potential risk node and the time series running data of the node when it is operating normally. DTW (k t ,k n ), where: k t is the time series running data of potential risk nodes, k n This is the time series operation data of the node when it is operating normally;

[0035] (4.3) Calculate the anomaly score ξ;

[0036]

[0037] Where: ω1 and ω2 are coefficients set based on experience, T represents the length of the time window or the number of data points, and u is the weight adjustment parameter;

[0038] (4.4) Divide the risk level intervals according to the predefined risk threshold sequence and implement a tiered alarm system.

[0039] The present invention also provides a port equipment green power system risk early warning system, comprising:

[0040] Data acquisition module, used to collect and pre-process the operating data related to new energy in the green power system of port equipment;

[0041] A graph structure construction module is used to construct a risk map using port equipment as nodes and the energy transmission and operation coordination relationships between port equipment as edges. The module also uses the collected operation data to assign features to each node and edge.

[0042] A risk identification module is used to build a graph convolutional network model, process the risk map data, and mark potential risk nodes;

[0043] The risk warning module is used to calculate the risk level for the potential risk nodes and implement a hierarchical warning system.

[0044] The present invention further provides an electronic device, comprising:

[0045] at least one processor; and

[0046] a memory communicatively connected to the at least one processor; wherein,

[0047] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the risk early warning method for the green power system of port equipment as described above.

[0048] The present invention also provides a computer-readable storage medium, which stores computer instructions. The computer instructions are used to enable a processor to implement the above-mentioned port equipment green power system risk warning method when executed.

[0049] Beneficial effects of the present invention:

[0050] 1. This invention constructs a risk map of port equipment and the energy transmission and operation coordination relationships between them, which can more comprehensively and intuitively reflect the overall situation of the green power system of port equipment, overcoming the limitations of traditional methods that only monitor a single device or a single energy source.

[0051] 2. The present invention uses graph convolutional networks to learn and analyze risk maps, and takes into account the collaborative operation intensity coefficient and energy transmission loss coefficient between equipment in port scenarios. It can more accurately identify potential risk nodes and improve the accuracy and reliability of risk identification.

[0052] 3. After identifying potential risk nodes, the present invention comprehensively considers multiple factors, calculates the statistical characteristics of key parameters and the dynamic time warping algorithm to obtain anomaly scores, and determines the risk level based on the score for layered alarms, providing port staff with clear and unambiguous risk information, helping them to take corresponding measures in a timely manner, reduce risk losses, and ensure the safe and stable operation of the green power system of port equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a risk warning method for a green power system of port equipment according to an embodiment of the present invention.

[0054] Figure 2 This is a schematic diagram of the structure of the green power system risk warning system for port equipment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The present invention will be further explained below with reference to the accompanying drawings and embodiments.

[0056] like Figure 1 As shown, the present invention provides a risk early warning method for a green power system of port equipment, comprising the following steps:

[0057] S101. Collect and pre-process the operating data related to new energy in the green power system of port equipment.

[0058] Collect operational data on new energy sources (including wind, solar, hydrogen, and electricity) in port equipment's green power systems. Preprocess the data by removing outliers and erroneous data, performing consistency checks on data from different data sources, and finally standardizing and normalizing all types of data.

[0059] S102. Port equipment is used as nodes, and the energy transmission and operation coordination relationship between port equipment is used as edges to construct a risk map, and the collected operation data is used to assign characteristics to each node and edge.

[0060] A risk map is constructed by taking port equipment (shore container cranes, port tugboats, AGVs, energy storage container trains, etc.) as nodes and energy transmission, operational coordination relationships, and physical connections between port equipment as edges.

[0061] G b =(V b ,E b )

[0062] Where: V b is the node set, E b is the edge set.

[0063] Based on the collected operational data, each node and edge is assigned features. Node features include the device's rated power, capacity, operating status (e.g., normal, faulty, under maintenance), energy consumption, operating efficiency, and environmental parameters. Edge features include energy transmission rate, collaborative operation frequency, transmission loss coefficient, and collaborative operation intensity coefficient.

[0064] For time-series operational data (such as the operating status of port equipment, energy consumption, and energy transmission rate), a sliding window is used for feature extraction. The extracted features include mean, variance, maximum, minimum, trend, and period. For non-time-series operational data (typically static attributes of port equipment, such as rated power, capacity, type, and environmental parameters), the original attributes are used or normalized and standardized to form the new feature vector of the node.

[0065] S103: Construct a graph convolutional network model, process the risk map data, and mark potential risk nodes.

[0066] Construct a graph convolutional network model and construct an embedded representation of the node based on the node's feature information. Construct an adjacency matrix A based on the energy transmission, operation collaboration and other relationships between equipment. If equipment i and equipment j have the above neighbor relationship, then A ij =1, otherwise A ij = 0. Introducing self-loops and performing symmetric normalization, the adjacency matrix A becomes The degree matrix is in Get the normalized matrix The feature representation of each node is updated using the neighbor node information of each node. The mathematical expression is as follows:

[0067]

[0068] Specifically, the embedding representation of node i at layer l+1 is:

[0069]

[0070] in: is node v i Embedding representation at layer l; N i Represents node v i The set of neighbor nodes of |N i | represents the number of neighbor nodes; W l is the learnable weight matrix of layer l; b l is the bias term; σ is the activation function (ReLU or tanh is used); α i,j β is the inter-equipment collaborative operation intensity coefficient, which reflects the closeness of the operation collaboration between equipment node i and its neighboring equipment node j. The larger the coefficient value, the more frequent and close the collaborative operation between the two. i,j is the energy transmission loss coefficient, which reflects the energy transmission loss between equipment node i and its neighboring equipment node j. Its value range is between 0 and 1. The closer it is to 0, the greater the energy transmission loss. j is the neighboring node of node i. and Nodes i and j are The degree of medium.

[0071] The risk map data is input into the trained graph convolutional network model to obtain the embedded representation of each node. The distance between the embedded representation of each node and the embedded representation of the normal behavior pattern is calculated. If the distance exceeds the set threshold, the node is marked as a potential risk node.

[0072]

[0073] Where: τ is the set threshold; Indicates that after multi-layer graph convolution operations, node v i In the last layer, i.e., the embedding representation of the L layer; H′ represents the embedding representation of the normal behavior pattern, which is the mean of the embedding representations of all normal nodes.

[0074] During the model training process, historical fault-free graph data is used to input the graph convolutional network. After multi-layer propagation, the embedded representation of all nodes is obtained, and its mean is calculated as the embedded representation of the normal behavior pattern.

[0075] In actual use, as new graph data is input into the graph convolutional network, the embedded representation of non-potential risk nodes obtained will be recalculated with the embedded representation of all nodes obtained from the original historical fault-free graph data, or the mean will be calculated using a sliding window as the embedded representation of the updated normal behavior pattern, which will be used to judge the potential risk nodes of subsequent graph data.

[0076] S104. Calculate the risk level for the potential risk node and implement a hierarchical alarm system.

[0077] (1) For the potential risk node marked in step (3), calculate the mean μ and standard deviation s of the key operating data of the node and obtain the z-score of each sample point.

[0078] z i =(x i -μ) / s

[0079] If |z i |≥M, then it is determined to be an outlier point k i * ; Where M is the set judgment threshold.

[0080] (2) Use the dynamic time warping algorithm to calculate the similarity S between the time series running data of the potential risk node and the time series running data of the node when it is operating normally DTW (k t ,k n ), where: k t is the time series running data of potential risk nodes, k n This is the time series running data of the node when it is operating normally.

[0081] Based on the basic principles of the dynamic time warping algorithm, the specific process is as follows:

[0082] Assume that the current parameter sequence k t For X = {x1, x2, ..., x m}, parameter sequence k during normal operation n Y={y1,y2,...,y n}.

[0083] Construct an m*n distance matrix D, where D(i, j) represents x i and y j The distance between them is calculated using the Euclidean distance:

[0084]

[0085] Where: i = 1, 2, ..., m, j = 1, 2, ..., n.

[0086] Find a path W = {w1, w2, ..., w k} makes k t and k n The similarity is the highest, that is, k t and k n After one stretch and one shorten, the distance between the two time series is the shortest.

[0087] Considering that the elements in the path W need to meet the conditions of boundary constraints, continuity and monotonicity constraints, the optimal path can be solved by dynamic programming. Define the cumulative matrix R = {r(i, j))} m，n To record the shortest path, that is:

[0088]

[0089] (3) Calculate the anomaly score ξ:

[0090]

[0091] Where: ω1 and ω2 are coefficients set based on experience, T represents the length of the time window or the number of data points, and u is the weight adjustment parameter.

[0092] (4) Divide the risk level intervals according to the predefined risk threshold sequence and implement a hierarchical alarm system.

[0093] The division of risk levels should be based on actual conditions. Different ports have different business characteristics and equipment tolerance, so the division standards may also vary. In some large and busy ports, due to the large volume of business and low risk tolerance, the division of risk levels may be more detailed; in small ports, the division of risk levels is relatively broad. For example:

[0094] If ξ is within the range of [ξ1,ξ2], it is judged as a mild risk and the system outputs signal 1.

[0095] If ξ is in the interval [ξ2,ξ3], it is judged as moderate risk and the system outputs signal 2.

[0096] If ξ≥ξ3, it is judged as a severe risk and the system outputs signal 3.

[0097] Among them: ξ1, ξ2, ξ3 are predefined risk thresholds.

[0098] The embodiment of the present invention further provides a port equipment green power system risk early warning system, comprising:

[0099] The data acquisition module 201 is used to collect and pre-process the operating data related to the new energy in the green power system of the port equipment;

[0100] A graph structure construction module 202 is used to construct a risk graph using port equipment as nodes and energy transmission and operation coordination relationships between port equipment as edges, and to assign features to each node and edge using the collected operation data;

[0101] The risk identification module 203 is used to build a graph convolutional network model, process the risk map data, and mark potential risk nodes;

[0102] The risk warning module 204 is used to calculate the risk level for potential risk nodes and implement a hierarchical warning system.

[0103] An embodiment of the present invention also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the port equipment green power system risk warning method as described above.

[0104] An embodiment of the present invention further provides a computer-readable storage medium storing computer instructions, which are used to enable a processor to implement the above-mentioned port equipment green power system risk warning method when executed.

[0105] 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 and improvements made within the principles and spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A risk early warning method for green power system of port equipment, characterized in that: The steps include: (1) Collect and pre-process the operating data related to new energy in the green power system of port equipment; (2) Taking port equipment as nodes and the energy transmission and operation coordination relationships between port equipment as edges, a risk map is constructed, and the collected operation data is used to assign features to each node and edge; (3) Constructing a graph convolutional network model to process the risk map data and mark potential risk nodes; (4) Calculate the risk level for the potential risk nodes and implement a hierarchical alarm system.

2. The risk early warning method for green power system of port equipment according to claim 1 is characterized in that: The pretreatment includes: (1.1) Remove outliers and erroneous data from the collected operating data related to new energy in the green power system of port equipment; (1.2) Perform consistency check on data from different data sources; (1.3) Standardize and normalize all types of data.

3. The risk early warning method for green power system of port equipment according to claim 1 is characterized in that: The step (2) comprises: (2.1) Taking port equipment as nodes and the energy transmission, operation coordination and physical connection relationships between port equipment as edges, a risk map is constructed; G b =(V b ,E b ) Where: V b is the node set, E b is the edge set; (2.2) For the time series running data, a sliding window is used for feature extraction. The extracted features include mean, variance, maximum, minimum, trend and cycle; (2.3) For non-time series operating data, use its original attributes or perform normalization and standardization as the extracted features; (2.4) Assign values ​​to each node and edge based on the features extracted in steps (2.2) and (2.3).

4. The risk early warning method for green power system of port equipment according to claim 1 is characterized in that: The step (3) is specifically as follows: (3.1) Build a graph convolutional network model and construct an embedded representation of the node based on its feature information. The embedded representation of node i in the l+1 layer is: Where: N i Represents node v i The set of neighbor nodes of is node v i The embedding representation of the neighbor nodes at layer l; W l is the learnable weight matrix of layer l; b l is the bias term; σ is the activation function; α i,j is the intensity coefficient of collaborative operation between equipment; β i,j is the energy transmission loss coefficient; j is the neighbor node of node i; and Nodes i and j are The degree of In order to introduce the self-loop and symmetrically normalized adjacency matrix based on the energy transmission and operation coordination relationship between port equipment, A is the adjacency matrix; (3.2) Inputting the risk map data into the trained graph convolutional network model to obtain the embedded representation of each node, and calculating the distance between the embedded representation of each node and the embedded representation of the normal behavior pattern. If the distance exceeds a set threshold, the node is marked as a potential risk node; Where: τ is the set threshold; Indicates that after multi-layer graph convolution operations, node v i In the last layer, i.e. the embedding representation of layer L; H′ represents the embedding representation of the normal behavior pattern, i.e. the mean of the embedding representations of all normal nodes.

5. The risk early warning method for green power system of port equipment according to claim 4 is characterized in that: The embedding representation of the normal behavior pattern is specifically: Use historical fault-free graph data to input the graph convolutional network, obtain the embedded representation of all nodes after multi-layer propagation, and calculate its mean as the embedded representation of the normal behavior pattern; As new graph data is input into the graph convolutional network, the embedded representations of non-potential risk nodes are obtained, and the mean of the embedded representations of all nodes obtained from the original historical fault-free graph data is recalculated, or the mean is calculated using a sliding window as the embedded representation of the updated normal behavior pattern.

6. The risk early warning method for green power system of port equipment according to claim 1 is characterized in that: The step (4) is specifically as follows: (4.1) For the potential risk node marked in step (3), calculate the mean μ and standard deviation s of the running data related to the node and obtain the z-score of each sample point; z i =(x i -μ) / s If |z i |≥M, then it is determined to be an abnormal point Where M is the set judgment threshold; (4.2) Use the dynamic time warping algorithm to calculate the similarity S between the time series running data of the potential risk node and the time series running data of the node when it is operating normally. DTW (k t ,k n ), where: k t is the time series running data of potential risk nodes, k n This is the time series operation data of the node when it is operating normally; (4.3) Calculate the anomaly score ξ; Where: ω1 and ω2 are coefficients set based on experience, T represents the length of the time window or the number of data points, and u is the weight adjustment parameter; (4.4) Divide the risk level intervals according to the predefined risk threshold sequence and implement a tiered alarm system.

7. A port equipment green power system risk warning system, characterized by: include: Data acquisition module, used to collect and pre-process the operating data related to new energy in the green power system of port equipment; A graph structure construction module is used to construct a risk map using port equipment as nodes and the energy transmission and operation coordination relationships between port equipment as edges. The module also uses the collected operation data to assign features to each node and edge. A risk identification module is used to build a graph convolutional network model, process the risk map data, and mark potential risk nodes; The risk warning module is used to calculate the risk level for the potential risk nodes and implement a hierarchical warning system.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the port equipment green power system risk warning method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the port equipment green power system risk warning method according to any one of claims 1 to 6 when executed.

Images