Super capacitor-based multi-source collaborative coal mine emergency power supply dispatching optimization method

Abstract

This invention provides a multi-source collaborative emergency power supply scheduling optimization method for coal mines based on supercapacitors, belonging to the field of coal mine safe power supply technology. The method includes acquiring real-time operating parameters of the power supply system to construct a node state feature matrix, decomposing and extracting dynamic features to construct a node priority system, generating power supply path scores and performing global optimization, determining equipment switching strategies and synchronous switching methods, and establishing a supercapacitor transitional power supply mechanism. This invention achieves rapid and reliable switching of coal mine power supply systems, ensures power supply continuity, and improves emergency response capabilities.

Description

Technical Field

[0001] This invention relates to the field of coal mine safe power supply technology, and in particular to a coal mine emergency power supply scheduling optimization method based on multi-source coordination of supercapacitors. Background Technology

[0002] As a high-risk industry, the stability and safety of the power supply system in coal mines directly affect the lives of miners and the economic benefits of the enterprise. Traditional coal mine power supply systems typically employ a primary and backup power supply configuration, switching to the backup power supply in the event of grid failures or disasters to ensure continuous power supply to critical loads. With the increasing intelligence and automation of coal mines, power supply systems face more complex scheduling and optimization challenges. In recent years, supercapacitors, as a novel energy storage technology, have been increasingly applied to emergency power supply in coal mines due to their high power density, long cycle life, and rapid charging and discharging capabilities.

[0003] Currently, coal mine emergency power supply systems have the following defects and shortcomings in multi-source collaborative scheduling: Traditional coal mine power supply scheduling methods lack a collaborative management mechanism for multiple power sources such as main and backup power sources and supercapacitors. It is difficult to ensure the stability of voltage and frequency during power switching, which may lead to momentary power outages during the switching process, affecting the normal operation of critical equipment. Existing coal mine power supply systems mainly rely on static threshold triggering during power switching, lacking in-depth analysis of the dynamic characteristics of nodes. It is difficult to achieve optimized scheduling based on node importance and power supply path reliability, and it is difficult to quickly determine the optimal power supply path in complex power supply networks.

[0004] In addition, in existing technologies, supercapacitors are mainly used as auxiliary power sources in emergency power supply systems for coal mines, failing to fully utilize their rapid response characteristics and lacking precise buffer timing control strategies. This results in an inability to achieve a smooth transition during power switching, increasing the risk of system instability. Summary of the Invention

[0005] This invention provides a method for optimizing emergency power supply scheduling in coal mines using multi-source coordination of supercapacitors, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a method for optimizing emergency power supply scheduling in coal mines based on multi-source coordination using supercapacitors, comprising:

[0007] The real-time operating parameters of the main power supply, backup power supply and supercapacitor energy storage system in the coal mine power supply system are obtained, and the node state feature matrix of the power supply network is constructed using the real-time operating parameters.

[0008] The node state feature matrix is ​​decomposed into trend and periodic terms to extract the dynamic features of the nodes. Based on the dynamic features, the transmission efficiency and reliability indicators of the power supply line are calculated, and a node priority evaluation system is established.

[0009] Each node is classified according to the node priority evaluation system, and a power supply path score is generated by combining feature aggregation operation. Based on the power supply path score, global optimization is performed to output the optimal power supply path scheme.

[0010] Based on the power supply path score, the dependency topology relationship between each node is calculated to construct a power supply equipment switching control strategy and determine the optimal switching path.

[0011] The power supply path score is decomposed and reconstructed to obtain voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, the state matrix of the phase-locked loop and the state equation of the virtual inertia control system are constructed respectively. Through the synergistic effect of the loop filter, the synchronous switching of the power supply is completed.

[0012] A transitional power supply mechanism based on supercapacitors is established. The optimal power supply path scheme is segmented through a backtracking algorithm to determine the buffer timing of each stage and ensure a smooth transition of the power supply state.

[0013] The node state feature matrix is ​​decomposed into trend and periodic terms to extract the dynamic features of the nodes. Based on these dynamic features, the transmission efficiency and reliability indicators of the power supply line are calculated, and a node priority evaluation system is established, including:

[0014] The variational mode decomposition method is used to decompose the state parameters of each node in the node state feature matrix into intrinsic mode functions. The frequency distribution characteristics of the intrinsic mode functions are calculated to obtain the spectral entropy. The intrinsic mode functions with spectral entropy higher than a preset spectral threshold are superimposed to form a trend term, and the rest are superimposed to form a periodic term.

[0015] Change features and fluctuation features are extracted from the trend item and the period item respectively, and the dynamic features of the node are constructed using the change features and the fluctuation features;

[0016] The load variation trend of nodes is determined based on the variation characteristics, and the voltage and current fluctuation of nodes are determined based on the fluctuation characteristics. The transmission efficiency is constructed based on the variation trend and the fluctuation. The dynamic characteristics and the real-time operating parameters are input into a pre-trained reliability assessment model to obtain the reliability index.

[0017] A node priority evaluation system is established based on the ant colony algorithm. The transmission efficiency is set as the forward guidance information, and the reliability index is set as the backward feedback information. The pheromone concentration is updated using the forward guidance information and the backward feedback information, and the priority of nodes is divided based on the pheromone concentration.

[0018] Based on the node priority evaluation system, each node is classified, and a power supply path score is generated by combining feature aggregation operations. Global optimization is then performed based on the power supply path score to output the optimal power supply path scheme, including:

[0019] Obtain the pheromone concentration of each node in the node priority evaluation system, construct a weighted undirected graph based on the pheromone concentration, generate an importance score using degree centrality and betweenness centrality, classify nodes with an importance score greater than a preset key score as key protection nodes, and classify the rest as regular management nodes.

[0020] The dynamic features and adjacency relationships of nodes are input into a graph convolutional layer. After adjacency matrix processing with degree matrix normalization, the output features of the graph convolutional layer are obtained. The pheromone concentration is used as the weight coefficient of the neighboring nodes to construct priority-aware features. The output features of the graph convolutional layer and the priority-aware features are residually connected to form fusion features. The fusion features are input into a multilayer perceptron to generate power supply path scores.

[0021] Based on the power supply path score, a state transition equation is constructed. A forward dynamic programming method is used to traverse the process layer by layer, and the current accumulated power supply path score and the importance score of the node are used as the node state value. In each decision stage, the optimal value of the stage is calculated iteratively by combining the adjacency relationship of the node and the node state value through the Bellman equation. Based on the optimal value of the stage, the optimal power supply path scheme is obtained by backtracking.

[0022] The dependency topology relationships between nodes are calculated based on the power supply path score to construct a power supply equipment switching control strategy and determine the optimal switching path, including:

[0023] Node feature vectors are constructed based on the voltage state and load characteristics of nodes. The correlation between nodes is calculated using power supply path scoring. The strength of the dependency between nodes is determined based on the cosine similarity of the node feature vectors and the correlation.

[0024] Based on the inter-node dependency strength, the sum of the dependency strengths of each node and its neighboring nodes is calculated, and a power supply switching control strategy is constructed. Different strategy levels are divided according to the sum of the dependency strengths. Nodes with a total dependency strength greater than a preset high-intensity threshold adopt a dual-power supply and real-time monitoring strategy; nodes with a total dependency strength between the preset high-intensity threshold and the preset low-intensity threshold adopt a fast switching and periodic monitoring strategy; nodes with a total dependency strength less than the preset low-intensity threshold adopt a basic switching and timed monitoring strategy. For nodes at different strategy levels, corresponding monitoring cycles and switching time parameters are set.

[0025] Based on the inter-node dependency strength and the node feature vectors, a feature propagation matrix is ​​constructed. The switching path evaluation value is calculated by iteratively propagating the node feature vectors on the graph structure. The path with the largest switching path evaluation value is selected as the optimal switching path.

[0026] A switching sequence is generated from high to low based on the total dependence intensity, and the power supply equipment switching operation is performed along the optimal switching path according to the monitoring period and the switching time.

[0027] The power supply path score is decomposed and reconstructed to obtain voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, a phase-locked loop state matrix and a virtual inertia control system state equation are constructed respectively. Through the synergistic effect of the loop filter, synchronous switching of the power supply is completed, including:

[0028] The power supply path score is adaptively decomposed using empirical mode decomposition, and the voltage quality coefficient and power characteristic coefficient are reconstructed through Hilbert transform. The voltage phase deviation is obtained by feature extraction of the voltage quality coefficient, and the power fluctuation is obtained by feature extraction of the power characteristic coefficient.

[0029] A phase-locked loop state matrix is ​​constructed based on the voltage phase deviation, and the center frequency of the phase-locked loop is calculated based on the state matrix; the power fluctuation is input to the loop filter, and the scaling factor of the loop filter is adjusted to obtain the real-time transfer function of the loop filter;

[0030] The state equation of the virtual inertia control system is constructed based on the power fluctuation amount. The virtual moment of inertia and virtual damping coefficient are determined by utilizing the variation characteristics of the power fluctuation amount. The virtual moment of inertia and virtual damping coefficient are substituted into the state equation, and the angular velocity value of the virtual synchronizer is obtained by solving the equation.

[0031] The angular velocity value is combined with the real-time transfer function to calculate the frequency compensation of the phase-locked loop; the frequency compensation is superimposed on the center frequency to obtain the real-time operating frequency of the phase-locked loop; the phase-locking error of the phase-locked loop is determined based on the real-time operating frequency.

[0032] The phase-locked error is simultaneously fed back to the input of the loop filter and the input of the virtual inertia control system to complete the synchronous switching of the power supply.

[0033] A transitional power supply mechanism based on supercapacitors is established. The optimal power supply path scheme is segmented using a backtracking algorithm to determine the buffer timing of each stage, ensuring a smooth transition of the power supply state. This includes:

[0034] A transitional power supply mechanism is constructed, dividing the power supply process into three stages: disconnection of the original path, transitional power supply by supercapacitor, and connection of the new path, and setting switching thresholds for each stage;

[0035] The load power demand curve is collected based on the aforementioned transitional power supply mechanism, and the power compensation amount is calculated in combination with the switching threshold of each stage. The changing characteristics of the load power demand curve are analyzed, a power balance equation is established, and the compensation power output characteristics of the supercapacitor at each stage are determined.

[0036] The power matching degree of the power supply path is evaluated based on the power compensation amount and the compensation power output characteristics, and segmentation criteria are set in combination with energy utilization rate;

[0037] A backtracking algorithm is used to search for segmentation points from the end point of the optimal power supply path backward. The criterion value of each candidate segmentation point is calculated based on the segmentation criterion. When the difference in the criterion values ​​of adjacent candidate segmentation points is less than the preset segmentation threshold, the current candidate segmentation point is marked as the optimal segmentation point. The search is repeated until the optimal power supply path scheme has been traversed.

[0038] The optimal power supply path is divided into power supply segments using the optimal segmentation point. Based on the power supply segments, the buffer timing of the three stages in the transitional power supply mechanism is determined to ensure a smooth transition of the power supply state.

[0039] Also includes:

[0040] The switching delay and response time are calculated based on the power compensation amount of the power supply segment;

[0041] The switching delay and the response time are used to plan the buffer timing of the power supply segment. The switching delay is used as the buffer power supply advance. The buffer exit time is set according to the timing relationship of adjacent power supply segments. The minimum duration of the buffer time window is determined by the response time.

[0042] Power supply switching is performed according to the buffer timing and the buffer exit time. The supercapacitor is activated in advance to provide buffer power supply before the original path is disconnected. During the transition phase, power is supplied according to the compensated power output characteristics and the output power is dynamically adjusted. When the new path is connected, power supply is withdrawn in stages according to the power gradual change rule and the buffer exit time. The switching frequency is adjusted in real time according to the load power change rate to achieve smooth switching of power supply path.

[0043] A second aspect of this invention provides a coal mine emergency power supply dispatch optimization system based on supercapacitor multi-source coordination, comprising:

[0044] The first unit is used to acquire the real-time operating parameters of the main power supply, backup power supply and supercapacitor energy storage system in the coal mine power supply system, and to construct the node state feature matrix of the power supply network using the real-time operating parameters.

[0045] The second unit is used to decompose the node state feature matrix into trend and periodic terms, extract the dynamic features of the nodes, calculate the transmission efficiency and reliability indicators of the power supply line based on the dynamic features, and establish a node priority evaluation system.

[0046] The third unit is used to classify each node according to the node priority evaluation system, combine feature aggregation operation to generate power supply path score, perform global optimization based on the power supply path score, and output the optimal power supply path scheme.

[0047] The fourth unit is used to calculate the dependency topology relationship between each node based on the power supply path score, thereby constructing a power supply equipment switching control strategy and determining the optimal switching path;

[0048] The fifth unit is used to decompose and reconstruct the power supply path score to obtain the voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, the phase-locked loop state matrix and the state equation of the virtual inertia control system are constructed respectively. Through the synergistic effect of the loop filter, the synchronous switching of the power supply is completed.

[0049] The sixth unit is used to establish a transitional power supply mechanism based on supercapacitors. It uses a backtracking algorithm to segment the optimal power supply path scheme, determine the buffer timing of each stage, and ensure a smooth transition of the power supply state.

[0050] A third aspect of the present invention,

[0051] An electronic device is provided, comprising:

[0052] processor;

[0053] Memory used to store processor-executable instructions;

[0054] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0055] Fourth aspect of the embodiments of the present invention,

[0056] A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0057] The beneficial effects of this application are as follows:

[0058] This invention acquires real-time operating parameters of the main power supply, backup power supply, and supercapacitor energy storage system, and constructs a node state feature matrix to achieve comprehensive perception and dynamic feature extraction of the power supply system. It can effectively identify key nodes and weak links in the power supply network and improve the accuracy of reliability assessment of the power supply system.

[0059] This invention generates power supply path scores based on a node priority evaluation system and feature aggregation calculation, and determines the optimal power supply path scheme through global optimization. At the same time, it constructs a power supply equipment switching control strategy, which solves the problems of inflexible power supply path selection and low switching efficiency in emergency situations in traditional methods, and significantly improves the response speed and scheduling efficiency of emergency power supply in coal mines.

[0060] This invention innovatively introduces a supercapacitor-based transitional power supply mechanism, combined with a phase-locked loop state matrix and a virtual inertia control system, to achieve seamless synchronous switching of power supplies and smooth transition of power supply states. This effectively avoids voltage fluctuations and power surges during the switching process, ensuring continuous and stable power supply to critical loads in coal mines and improving the safety and reliability of the entire emergency power supply system. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating the multi-source coordinated emergency power supply scheduling optimization method for supercapacitors in coal mines, as described in an embodiment of the present invention.

[0062] Figure 2 The flowchart shows the power supply path optimization algorithm. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0065] Figure 1 This is a flowchart illustrating the multi-source coordinated emergency power supply scheduling optimization method for supercapacitors in coal mines, as described in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0066] The real-time operating parameters of the main power supply, backup power supply and supercapacitor energy storage system in the coal mine power supply system are obtained, and the node state feature matrix of the power supply network is constructed using the real-time operating parameters.

[0067] The node state feature matrix is ​​decomposed into trend and periodic terms to extract the dynamic features of the nodes. Based on the dynamic features, the transmission efficiency and reliability indicators of the power supply line are calculated, and a node priority evaluation system is established.

[0068] Each node is classified according to the node priority evaluation system, and a power supply path score is generated by combining feature aggregation operation. Based on the power supply path score, global optimization is performed to output the optimal power supply path scheme.

[0069] Based on the power supply path score, the dependency topology relationship between each node is calculated to construct a power supply equipment switching control strategy and determine the optimal switching path.

[0070] The power supply path score is decomposed and reconstructed to obtain voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, the state matrix of the phase-locked loop and the state equation of the virtual inertia control system are constructed respectively. Through the synergistic effect of the loop filter, the synchronous switching of the power supply is completed.

[0071] A transitional power supply mechanism based on supercapacitors is established. The optimal power supply path scheme is segmented through a backtracking algorithm to determine the buffer timing of each stage and ensure a smooth transition of the power supply state.

[0072] In one optional implementation, the node state feature matrix is ​​decomposed into trend and periodic terms to extract the dynamic features of the nodes. Based on the dynamic features, the transmission efficiency and reliability indicators of the power supply line are calculated, and a node priority evaluation system is established, including:

[0073] The variational mode decomposition method is used to decompose the state parameters of each node in the node state feature matrix into intrinsic mode functions. The frequency distribution characteristics of the intrinsic mode functions are calculated to obtain the spectral entropy. The intrinsic mode functions with spectral entropy higher than a preset spectral threshold are superimposed to form a trend term, and the rest are superimposed to form a periodic term.

[0074] Change features and fluctuation features are extracted from the trend item and the period item respectively, and the dynamic features of the node are constructed using the change features and the fluctuation features;

[0075] The load variation trend of nodes is determined based on the variation characteristics, and the voltage and current fluctuation of nodes are determined based on the fluctuation characteristics. The transmission efficiency is constructed based on the variation trend and the fluctuation. The dynamic characteristics and the real-time operating parameters are input into a pre-trained reliability assessment model to obtain the reliability index.

[0076] A node priority evaluation system is established based on the ant colony algorithm. The transmission efficiency is set as the forward guidance information, and the reliability index is set as the backward feedback information. The pheromone concentration is updated using the forward guidance information and the backward feedback information, and the priority of nodes is divided based on the pheromone concentration.

[0077] In one embodiment, the coal mine power supply network collects state characteristic data of multiple nodes, including parameters such as voltage, current, power factor, and load, forming a node state characteristic matrix.

[0078] Variational mode decomposition (VM) is applied to the state parameters of each node in the node state feature matrix. Taking node N001 in a coal mine power supply network as an example, voltage data of this node over 24 hours is collected at a sampling frequency of once per minute, resulting in 1440 sampling points. This voltage data is input into the VM algorithm, with the number of decomposition modes set to 6, resulting in 6 intrinsic mode functions (IMFs) IMF1 to IMF6. The Hilbert transform of each IMF is calculated to obtain its instantaneous frequency characteristics, and then the spectral entropy is calculated. Specifically, the spectral entropy values ​​of IMF1 to IMF6 are calculated as 0.85, 0.76, 0.65, 0.55, 0.42, and 0.38, respectively. A spectral entropy threshold of 0.6 is preset; therefore, the spectral entropies of IMF1, IMF2, and IMF3 are higher than the threshold, and these three IMFs are superimposed to obtain the trend term; the spectral entropies of IMF4, IMF5, and IMF6 are lower than the threshold, and these three IMFs are superimposed to obtain the periodic term.

[0079] The variation characteristics are extracted from the trend term, focusing primarily on its slope and the distribution of peaks and troughs. For node N001, the slope sequence is obtained by calculating the rate of change of the trend term in different time windows: [0.02, 0.05, 0.08, -0.03, -0.06, 0.04]. Peak and trough locations are identified, with three main peaks and two main troughs observed within 24 hours, occurring at 7:00 AM, 12:00 PM, 7:00 PM, 10:00 AM, and 3:00 PM. Fluctuation characteristics are extracted from the periodic term, with a standard deviation of 0.035, a root mean square value of 0.042, a waveform factor of 1.2, and a peak factor of 3.2. Combining the changing characteristics of the trend term and the fluctuation characteristics of the periodic term, the dynamic feature vector of node N001 is constructed as [0.02, 0.05, 0.08, -0.03, -0.06, 0.04, 3, 2, 0.035, 0.042, 1.2, 3.2].

[0080] Based on the variation characteristics of node N001, especially the slope sequence of the trend term, the load variation trend of this node is determined. By analyzing the positive and negative changes in the slope sequence, the patterns of load increase and decrease can be identified. For example, the load of this node increases rapidly from 6:00 to 8:00 AM with a slope of 0.08; and decreases from 2:00 PM to 4:00 PM with a slope of -0.06. Based on the fluctuation characteristics extracted from the periodic term, the voltage and current fluctuations of this node are determined. The voltage standard deviation of this node is 0.035, and the waveform factor is 1.2, indicating that the voltage fluctuations are relatively stable. Based on the above variation trend and fluctuations, a transmission efficiency index is constructed. The calculation formula adopts a weighted combination of line loss rate and voltage qualification rate. For node N001, the calculated line loss rate is 3.2%, the voltage qualification rate is 98.5%, and the weighted transmission efficiency index is 0.92.

[0081] The dynamic characteristics and real-time operating parameters of node N001 (including voltage 220V, current 15A, and power factor 0.95) are input into a pre-trained reliability assessment model. This model is a random forest model trained based on historical failure data, containing 50 decision trees, a maximum depth of 10, and a feature importance threshold of 0.05. The model outputs a reliability index of 0.88 for this node, indicating that the node is operating well and has a low probability of failure.

[0082] A node priority evaluation system was established based on the ant colony algorithm. In the algorithm initialization phase, the number of ants was set to 20, the maximum number of iterations to 100, the pheromone importance parameter α to 1.0, the heuristic factor importance parameter β to 2.0, and the pheromone evaporation coefficient ρ to 0.5. The transmission efficiency of node N001 (0.92) was used as the forward guidance information, and the reliability index (0.88) was used as the backward feedback information. During path selection, ants selected the next node based on the forward guidance information and updated the pheromone concentration based on the backward feedback information after completing the path. After 100 iterations, the pheromone concentration of node N001 reached 0.75, significantly higher than the average pheromone concentration of 0.45 in the network.

[0083] Based on the final pheromone concentration distribution, all nodes are divided into three priority levels according to their pheromone concentration: high priority (pheromone concentration > 0.7), medium priority (0.4 < pheromone concentration ≤ 0.7), and low priority (pheromone concentration ≤ 0.4). Node N001 has a pheromone concentration of 0.75 and is classified as a high-priority node, indicating its important role in the coal mine power supply network and the need to prioritize its power supply quality and reliability. In the entire network, nodes classified as high priority account for 15%, medium priority nodes account for 55%, and low priority nodes account for 30%, forming a reasonable priority distribution.

[0084] Figure 2The flowchart illustrates the power supply path optimization algorithm. In one optional implementation, nodes are graded according to the node priority evaluation system, and a power supply path score is generated by combining feature aggregation operations. Global optimization is then performed based on the power supply path score to output the optimal power supply path scheme, including:

[0085] Obtain the pheromone concentration of each node in the node priority evaluation system, construct a weighted undirected graph based on the pheromone concentration, generate an importance score using degree centrality and betweenness centrality, classify nodes with an importance score greater than a preset key score as key protection nodes, and classify the rest as regular management nodes.

[0086] The dynamic features and adjacency relationships of nodes are input into a graph convolutional layer. After adjacency matrix processing with degree matrix normalization, the output features of the graph convolutional layer are obtained. The pheromone concentration is used as the weight coefficient of the neighboring nodes to construct priority-aware features. The output features of the graph convolutional layer and the priority-aware features are residually connected to form fusion features. The fusion features are input into a multilayer perceptron to generate power supply path scores.

[0087] Based on the power supply path score, a state transition equation is constructed. A forward dynamic programming method is used to traverse the process layer by layer, and the current accumulated power supply path score and the importance score of the node are used as the node state value. In each decision stage, the optimal value of the stage is calculated iteratively by combining the adjacency relationship of the node and the node state value through the Bellman equation. Based on the optimal value of the stage, the optimal power supply path scheme is obtained by backtracking.

[0088] Based on the node priority evaluation system, the pheromone concentration value of each node is obtained. For example, the pheromone concentration of the main shaft hoist power supply node in a coal mine is 0.85, the ventilation system power supply node is 0.78, the coal mining face power supply node is 0.72, the drainage system node is 0.69, the safety monitoring system node is 0.67, while the ordinary lighting system node is only 0.38. A weighted undirected graph G=(V,E,W) is constructed based on these pheromone concentration values, where V represents the set of nodes, E represents the set of edges, and W represents the set of edge weights. For example, the edge weight between the main shaft hoist node and the ventilation system node is 0.85×0.78=0.663, and the edge weight between the main shaft hoist and the safety monitoring system node is 0.85×0.67=0.5695.

[0089] After constructing the weighted undirected graph, the degree centrality and betweenness centrality of each node are calculated. Degree centrality represents the number of directly connected neighbors of a node. For node N1 in a main substation, which connects to 8 downstream equipment nodes, its degree centrality DC(N1) = 8 / 12 = 0.667, where 12 is the maximum number of connections in the network. Betweenness centrality represents the proportion of the number of shortest paths passing through that node to the total number of shortest paths. Node N2 in a distribution box is located at the intersection of multiple power supply paths. There are 90 shortest paths in the network, of which 41 pass through node N2. Therefore, its betweenness centrality BC(N2) = 41 / 90 = 0.456. The degree centrality and betweenness centrality are weighted in a 6:4 ratio to generate a node importance score. Taking node N1 as an example, its importance score IS(N1) = 0.6 × 0.667 + 0.4 × 0.38 = 0.5522, where 0.38 is the betweenness centrality value of N1. The preset critical score is set to 0.65. Nodes with an importance score greater than 0.65 are classified as critical support nodes, such as power supply nodes for the main shaft hoist (IS=0.722), ventilation system (IS=0.689), and safety monitoring system (IS=0.672). The remaining nodes are classified as regular management nodes, such as lighting (IS=0.412) and office area power supply nodes (IS=0.389).

[0090] After node partitioning, the dynamic features and adjacency relationships of the nodes are input into the graph convolutional layer for processing. Dynamic features include parameters such as voltage stability, load factor, and current fluctuation rate. For example, the voltage stability of a node in a main substation is 0.92, the load factor is 0.78, and the current fluctuation rate is 0.15, forming the feature vector [0.92, 0.78, 0.15]. Adjacency relationships are represented by the adjacency matrix A. An element value of 1 indicates a direct connection between nodes, while 0 indicates no connection. Taking a simplified 5-node network as an example, the adjacency matrix is: A = [[0,1,1,0,0], [1,0,1,1,0], [1,1,0,1,1], [0,1,1,0,1], [0,0,1,1,0]].

[0091] To standardize the adjacency matrix, first calculate the degree matrix D. D is a diagonal matrix, where the elements on the diagonal represent the degree of the corresponding node. For the above adjacency matrix, the degree matrix is: D = [[2,0,0,0,0], [0,3,0,0,0], [0,0,4,0,0], [0,0,0,3,0], [0,0,0,0,2]].

[0092] Then calculate the normalized adjacency matrix D. -1 / 2 ·A·D -1 / 2Thus, we get: A_norm = [[0,0.408,0.354,0,0],[0.408,0,0.289,0.333,0],[0.354,0.289,0,0.289,0.354], [0,0.333,0.289,0,0.408], [0,0,0.354,0.408,0]].

[0093] Graph convolutional layers, through a feature propagation mechanism, enable each node to aggregate information from its neighboring nodes. For an input feature matrix X and a weight matrix W, the graph convolution operation is A_norm·X·W. Assume the input feature dimension is 3, the output feature dimension is 64, and the weight matrix W has a dimension of 3×64. Taking node 2 as an example, its initial feature vector is [0.86, 0.72, 0.21]. After passing through the graph convolutional layer, node 2's features are updated to a weighted sum of the features of its neighboring nodes 1, 3, and 4, specifically calculated as follows: X_2_new = 0.408×[0.92,0.78,0.15]+ 0×[0.86,0.72,0.21] + 0.289×[0.79,0.65,0.18]+0.333×[0.83,0.59,0.23] = [0.843,0.672,0.183].

[0094] Then, this vector is multiplied by the weight matrix W to obtain a 64-dimensional output feature. Simultaneously, using the previously obtained pheromone concentration as the weight coefficients of neighboring nodes, a priority-aware feature is constructed. The priority-aware feature is obtained by weighted summation of the features of neighboring nodes. Taking node 2 as an example, the pheromone concentrations of its neighbors 1, 3, and 4 are 0.85, 0.72, and 0.67, respectively. After normalization, the weight coefficients are [0.38, 0.32, 0.30]. The priority-aware feature is calculated as: PF_2 = 0.38×[0.92,0.78,0.15]+ 0.32×[0.79,0.65,0.18] + 0.30×[0.83,0.59,0.23]= [0.853,0.682,0.183].

[0095] This vector is mapped to a 64-dimensional space through a fully connected layer to obtain a priority-aware feature vector. The output feature GCN_2 of the graph convolutional layer and the priority-aware feature PF_2 are fused through a residual connection to form a fused feature FF_2 = ReLU(GCN_2 + PF_2). Assuming that GCN_2 and PF_2 are both 64-dimensional vectors, after element-wise addition and ReLU activation, the fused feature FF_2 is obtained. These fused features are then input into a multilayer perceptron containing two hidden layers with 64 and 32 neurons respectively, using the ReLU activation function. The final output layer uses the Sigmoid activation function to generate a power supply path score between 0 and 1.

[0096] Taking the path evaluation from node 1 to node 5 as an example, the path contains the node sequence [1,3,5], and the fused features of each node on the path are FF_1, FF_3, and FF_5, respectively. The calculation process of the multilayer perceptron is as follows: H1 = ReLU(W1·concat(FF_1,FF_3,FF_5) + b1), H2 = ReLU(W2·H1 + b2), Score = Sigmoid(W3·H2 + b3).

[0097] Here, `concat` represents the feature concatenation operation. `W1`, `W2`, `W3` and `b1`, `b2`, `b3` are the weight matrices and bias vectors of each layer, respectively. Specifically, in the calculation, assuming the concatenated feature vector is 192-dimensional, `W1` is a 192×64-dimensional matrix, and `b1` is a 64-dimensional vector, we calculate `H1` as a 64-dimensional vector. Then, using `W2` (64×32) and `b2`, we calculate `H2` as a 32-dimensional vector. Finally, using `W3` (32×1) and `b3`, we calculate the scalar score. For example, the path score from a main shaft hoist to the backup power supply is calculated to be 0.92, indicating high power supply efficiency and reliability; while the path score to a remote distribution box is only 0.45.

[0098] Based on these power supply path scores, a state transition equation is constructed to solve for the optimal power supply path. The state transition equation is V(i,t) = max{V(j,t-1) + α·Score(j,i) + (1-α)·IS(i)}, where V(i,t) represents the state value of node i in stage t, Score(j,i) represents the path score from node j to node i, IS(i) represents the importance score of node i, and α is a weight coefficient, set to 0.7.

[0099] A forward dynamic programming method is used to traverse the network nodes layer by layer. Taking the path planning from the main power source S to the load node L as an example, the nodes involved are S, A, B, C, D, and L. Nodes A and B are located in the first layer, and C and D are located in the second layer. The initial state is V(S,0)=0. In the first stage, V(A,1) and V(B,1) are calculated: V(A,1) = 0.835, V(B,1) = 0.727. In the second stage, V(C,2) and V(D,2) are calculated. C can be reached from either A or B, and D can be reached from either A or B: V(C,2)_fromA = 1.401, V(C,2)_fromB = 1.28. The maximum value V(C,2) = max{1.401, 1.28} = 1.401 is taken, and the path from A to C is selected.

[0100] Similarly, calculate V(D,2): V(D,2)_fromA = 1.36, V(D,2)_fromB = 1.322, take the maximum value V(D,2) = max{1.36, 1.322} = 1.36, and choose the path from A to D.

[0101] Finally, calculate V(L,3). L can be reached from C or D: V(L,3)_fromC 2.037, V(L,3)_fromD 1.938. Take the maximum value V(L,3) = max{2.037, 1.938} = 2.037, and choose the path from C to L.

[0102] The optimal path S→A→C→L was obtained through backtracking, with a total score of 0.88×0.81×0.86=0.613. This scheme can guide the operation and scheduling of coal mine power supply systems in practical applications. For example, the optimal path from the main power source to the main shaft hoist is: Main Power Source → Main Substation → No. 1 Distribution Room → Main Shaft Substation → Main Shaft Hoist Power Supply System, ensuring reliable power supply to critical equipment and improving the efficiency and safety of the overall power supply system.

[0103] In one optional implementation, the dependency topology relationships between nodes are calculated based on the power supply path score to construct a power supply equipment switching control strategy and determine the optimal switching path, including:

[0104] Node feature vectors are constructed based on the voltage state and load characteristics of nodes. The correlation between nodes is calculated using power supply path scoring. The strength of the dependency between nodes is determined based on the cosine similarity of the node feature vectors and the correlation.

[0105] Based on the inter-node dependency strength, the sum of the dependency strengths of each node and its neighboring nodes is calculated, and a power supply switching control strategy is constructed. Different strategy levels are divided according to the sum of the dependency strengths. Nodes with a total dependency strength greater than a preset high-intensity threshold adopt a dual-power supply and real-time monitoring strategy; nodes with a total dependency strength between the preset high-intensity threshold and the preset low-intensity threshold adopt a fast switching and periodic monitoring strategy; nodes with a total dependency strength less than the preset low-intensity threshold adopt a basic switching and timed monitoring strategy. For nodes at different strategy levels, corresponding monitoring cycles and switching time parameters are set.

[0106] Based on the inter-node dependency strength and the node feature vectors, a feature propagation matrix is ​​constructed. The switching path evaluation value is calculated by iteratively propagating the node feature vectors on the graph structure. The path with the largest switching path evaluation value is selected as the optimal switching path.

[0107] A switching sequence is generated from high to low based on the total dependence intensity, and the power supply equipment switching operation is performed along the optimal switching path according to the monitoring period and the switching time.

[0108] In coal mine power supply systems, node feature vectors are constructed based on the voltage state and load characteristics of nodes. The voltage state includes voltage amplitude, phase angle, and fluctuation rate, while the load characteristics include load type, power factor, and load change rate. Taking node N1 of a coal mine main substation as an example, its voltage amplitude is 10.2kV, phase angle is 2.8°, voltage fluctuation rate is 0.3%, load type is inductive load, power factor is 0.92, and load change rate is 5% / h. Based on this, the node feature vector FV1=[10.2,2.8, 0.3, 1, 0.92, 5] is constructed. Similarly, for node N2 of a substation in a mining area, its eigenvector is FV2=[6.1,3.2, 0.5, 1, 0.88, 8]; for node N3 of a ventilation system, its eigenvector is FV3=[1.12, 4.1,0.8, 2, 0.85, 15]; and for node N4 of a lighting system, its eigenvector is FV4=[0.38, 1.5, 0.2, 3,0.95, 2]. Load types 1, 2, and 3 represent power loads primarily composed of electric motors, combined loads primarily composed of transformers, and resistive loads primarily composed of lighting, respectively.

[0109] Using the previously calculated power path scores, the correlation between nodes is calculated. For nodes N1 and N2, assuming the power path score from N1 to N2 is 0.85 and the power path score from N2 to N1 is 0.82, the correlation CR12 between N1 and N2 is calculated as the average of the two, 0.835. Similarly, the correlation CR23 between N2 and N3 is 0.76, and the correlation CR34 between N3 and N4 is 0.65.

[0110] The strength of dependency between nodes is determined based on the cosine similarity and correlation of their feature vectors. Cosine similarity is calculated using the cosine of the angle between two vectors; a value closer to 1 indicates greater similarity. For nodes N1 and N2, the cosine similarity CS12 of their feature vectors is calculated. First, the dot product of the two vectors FV1·FV2 = 10.2×6.1 + 2.8×3.2 + 0.3×0.5 + 1×1 + 0.92×0.88 + 5×8 = 111.606; then, the magnitude of each vector |FV1| = (10.2 2 +2.8 2 + 0.3 2 +1 2 + 0.92 2 +5 2 ) 1 / 2 = 11.871, |FV2| = (6.1 2 + 3.2 2 + 0.5 2 +1 2 + 0.88 2 +8 2 ) 1 / 2 = 10.863; Finally, the cosine similarity CS12 was calculated as 111.606 / (11.871×10.863) = 0.866.

[0111] The inter-node dependency strength DS12 is calculated using a weighted sum of cosine similarity and association degree: DS12 = 0.6×CS12 + 0.4×CR12 = 0.6×0.866 + 0.4×0.835 = 0.854. Similarly, the dependency strengths of other node pairs are calculated: DS23 = 0.6×0.752 + 0.4×0.76 = 0.755, DS34 = 0.6×0.682 + 0.4×0.65 = 0.669.

[0112] Based on the dependency strength between nodes, the sum of the dependency strengths of each node and its neighboring nodes is calculated. Node N2 is adjacent to nodes N1 and N3, and its total dependency strength DT2 = DS12 + DS23 = 0.854 + 0.755 = 1.609. Node N3 is adjacent to nodes N2 and N4, and its total dependency strength DT3 = DS23 + DS34 = 0.755 + 0.669 = 1.424. The sum of the dependency strengths of other nodes is calculated in this manner.

[0113] Based on the total dependency intensity, different strategy levels are defined, and a power supply equipment switching control strategy is constructed. A preset high-intensity threshold of 1.5 and a preset low-intensity threshold of 1.0 are set. For node N2, its total dependency intensity DT2 = 1.609 > 1.5, a dual-power supply and real-time monitoring strategy is adopted. For node N3, its total dependency intensity DT3 = 1.424, which is between 1.0 and 1.5, a fast switching and periodic monitoring strategy is adopted. For nodes with a total dependency intensity less than 1.0, such as a lighting node N4, a basic switching and timed monitoring strategy is adopted.

[0114] For nodes with different strategy levels, corresponding monitoring cycle and switching time parameters are set. For node N2, which adopts a dual power supply and real-time monitoring strategy, the monitoring cycle is set to 5 seconds and the switching time is set to 0.2 seconds. For node N3, which adopts a fast switching and periodic monitoring strategy, the monitoring cycle is set to 30 seconds and the switching time is set to 0.5 seconds. For node N4, which adopts a basic switching and timed monitoring strategy, the monitoring cycle is set to 5 minutes and the switching time is set to 1 second.

[0115] A feature propagation matrix is ​​constructed based on the inter-node dependency strength and node eigenvectors. The element Pij in the feature propagation matrix P represents the feature propagation weight from node i to node j, calculated as Pij = DSij / ∑DSik, where k iterates through all neighboring nodes of node i. Taking node N2 as an example, it is adjacent to nodes N1 and N3, with dependency strengths DS21 = 0.854 and DS23 = 0.755, respectively. The feature propagation weights are P21 = 0.854 / (0.854+0.755) = 0.531 and P23 = 0.755 / (0.854+0.755) = 0.469.

[0116] The evaluation value of the switching path is calculated by iteratively propagating the node feature vectors on the graph structure. Initially, the feature vector of each node is its original feature vector. After the first iteration, the feature vector of node N2 is updated to FV2_new = P21×FV1 + P23×FV3 = 0.531×[10.2, 2.8, 0.3, 1, 0.92, 5] + 0.469×[1.12, 4.1,0.8, 2, 0.85, 15]= [5.95, 3.41, 0.53, 1.47, 0.89, 9.69]. The feature vectors of other nodes are updated similarly.

[0117] After the iterative propagation process is repeated multiple times (e.g., 3 times), the evaluation value of all possible switching paths from the source node to the target node is calculated. The path evaluation value is obtained by weighting the similarity of the updated feature vectors of each node on the path with the path length. Taking the switching path from the main power source S to the load node L as an example, assume there are two paths: S→A→B→L and S→C→L. For the path S→A→B→L, the path evaluation value PE1 = (CSA + CAB + CBL) / 3 - 0.1×3 = (0.88 + 0.92 + 0.85) / 3 - 0.1×3 = 0.583, where CSA, CAB, and CBL are the cosine similarity of the feature vectors of adjacent nodes, 3 is the path length, and 0.1 is the length penalty coefficient. For the path S→C→L, the path evaluation value PE2 = (CSC + CCL) / 2 - 0.1×2 = (0.82 + 0.79) / 2 - 0.1×2 = 0.605. Comparing the evaluation values ​​of the two paths, the path S→C→L, with PE2 = 0.605 > PE1 = 0.583, is selected as the optimal switching path.

[0118] A switching sequence is generated based on the total dependency strength, from high to low. According to the monitoring cycle and switching time, the power supply equipment switching operation is performed along the optimal switching path. Taking a sudden failure in a coal mine as an example, it is necessary to switch from the main power supply to the backup power supply. Based on the previously calculated total dependency strength, the node switching sequence is N2→N3→N4→... For high-priority node N2, monitoring is performed every 5 seconds, and switching is completed within 0.2 seconds when an anomaly is detected; for medium-priority node N3, monitoring is performed every 30 seconds, and switching is completed within 0.5 seconds when an anomaly is detected; for low-priority node N4, monitoring is performed every 5 minutes, and switching is completed within 1 second when an anomaly is detected.

[0119] In a specific application scenario, the power supply node of a coal mine's main shaft hoisting system employs a dual-power supply and real-time monitoring strategy, with a monitoring cycle of 5 seconds and a switching time of 0.2 seconds. When a main power supply failure occurs and voltage fluctuations exceeding a threshold are detected, the switching procedure is immediately initiated. Based on the pre-calculated optimal switching path, the power supply is switched from the 10kV main power line to the 10kV backup line. The entire process is completed within 0.2 seconds, ensuring uninterrupted operation of the main shaft hoisting system. Simultaneously, following a preset switching sequence, the power supply to other equipment such as the ventilation system and drainage system is switched sequentially, ensuring safe production in the coal mine.

[0120] In one optional implementation, the power supply path score is decomposed and reconstructed to obtain voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, a phase-locked loop state matrix and a state equation of a virtual inertia control system are constructed, respectively. Through the synergistic effect of the loop filter, synchronous switching of the power supply is achieved, including:

[0121] The power supply path score is adaptively decomposed using empirical mode decomposition, and the voltage quality coefficient and power characteristic coefficient are reconstructed through Hilbert transform. The voltage phase deviation is obtained by feature extraction of the voltage quality coefficient, and the power fluctuation is obtained by feature extraction of the power characteristic coefficient.

[0122] A phase-locked loop state matrix is ​​constructed based on the voltage phase deviation, and the center frequency of the phase-locked loop is calculated based on the state matrix; the power fluctuation is input to the loop filter, and the scaling factor of the loop filter is adjusted to obtain the real-time transfer function of the loop filter;

[0123] The state equation of the virtual inertia control system is constructed based on the power fluctuation amount. The virtual moment of inertia and virtual damping coefficient are determined by utilizing the variation characteristics of the power fluctuation amount. The virtual moment of inertia and virtual damping coefficient are substituted into the state equation, and the angular velocity value of the virtual synchronizer is obtained by solving the equation.

[0124] The angular velocity value is combined with the real-time transfer function to calculate the frequency compensation of the phase-locked loop; the frequency compensation is superimposed on the center frequency to obtain the real-time operating frequency of the phase-locked loop; the phase-locking error of the phase-locked loop is determined based on the real-time operating frequency.

[0125] The phase-locked error is simultaneously fed back to the input of the loop filter and the input of the virtual inertia control system to complete the synchronous switching of the power supply.

[0126] Based on the power supply path scores obtained from the aforementioned calculations, an adaptive decomposition is performed using the empirical mode decomposition method. Taking the power supply path score sequence of a coal mine substation as an example, the score data is [0.92, 0.89, 0.85, 0.88, 0.91, 0.87, 0.84, 0.86, 0.89, 0.92]. After empirical mode decomposition, three intrinsic mode functions (IMF1, IMF2, IMF3) and a residual term (res) are obtained. IMF1 reflects high-frequency variation characteristics, with values ​​of [0.03, -0.02, -0.04, 0.01, 0.03, -0.02, -0.03, 0.00, 0.01, 0.02]; IMF2 reflects mid-frequency variation characteristics, with values ​​of [0.01, 0.02, 0.00, -0.01, -0.01, 0.00, 0.01, 0.01, 0.00, -0.01]; IMF3 reflects low-frequency variation characteristics, with values ​​of [-0.01, -0.01, -0.01, 0.00, 0.00, 0.00, -0.01, -0.01, 0.00, 0.01]; the residual term res is [0.89, 0.90, 0.90, 0.88, 0.89, 0.89, 0.87, 0.86, 0.88, 0.90].

[0127] The intrinsic mode functions obtained from the decomposition are reconstructed using Hilbert transform to obtain the voltage quality coefficient and power characteristic coefficient. The voltage quality coefficient VQC is obtained by reconstructing IMF1 and IMF2 using Hilbert transform, with values ​​of [0.031, 0.028, 0.040, 0.014, 0.032, 0.020, 0.032, 0.010, 0.010, 0.022]. The power characteristic coefficient PCC is obtained by reconstructing IMF3 and the residual term res using Hilbert transform, with values ​​of [0.880, 0.890, 0.890, 0.880, 0.890, 0.890, 0.860, 0.850, 0.880, 0.910].

[0128] Feature extraction is performed on the voltage quality coefficient (VQC) to obtain the voltage phase deviation. Feature extraction is achieved by calculating the cumulative sum of the VQC sequence and multiplying it by a phase conversion factor. Assuming a phase conversion factor of 5.0, the calculated voltage phase deviation (DPV) is [0.155, 0.295, 0.495, 0.565, 0.725, 0.825, 0.985, 1.035, 1.085, 1.195] degrees. Feature extraction is performed on the power characteristic coefficient (PCC) to obtain the power fluctuation. Feature extraction is achieved by calculating the difference in the PCC sequence and multiplying it by a power fluctuation conversion factor. Assuming a power fluctuation conversion factor of 10.0, the calculated power fluctuation (DPF) is [0.0, 1.0, 0.0, -1.0, 1.0, 0.0, -3.0, -1.0, 3.0, 3.0] kilowatts.

[0129] The phase-locked loop (PLL) state matrix M is constructed based on the voltage phase deviation (DPV). The PLL state matrix M is a second-order matrix, and its elements are calculated from the adjacent values ​​of the DPV sequence. The first row of matrix M contains elements M11 and M12, which are the weighted sums of the last two values ​​of the DPV sequence: M11 = 0.7 × 1.195 + 0.3 × 1.085 = 1.162, M12 = 0.6 × 1.195 + 0.4 × 1.085 = 1.151. The second row of matrix M contains elements M21 and M22, with M21 = 1.0 and M22 = 0.0, respectively. The center frequency of the PLL is calculated from the state matrix M. The center frequency fc is calculated from the eigenvalues ​​of M: fc = 50 + (M11 + M22) / 2 = 50 + (1.162 + 0.0) / 2 = 50.581 Hz.

[0130] The power fluctuation factor (DPF) is input to the loop filter. The proportional gain of the loop filter is adjusted. The loop filter is a proportional-integral (PI) filter, and its proportional gain Kp is dynamically adjusted based on the mean and standard deviation of the DPF. The mean DPF is calculated as mean_DPF = (0.0 + 1.0 + 0.0 + (-1.0) + 1.0 + 0.0 + (-3.0) + (-1.0) + 3.0 + 3.0) / 10 = 0.3 kW, and the standard deviation std_DPF = 1.767 kW. The proportional gain Kp is calculated as Kp = 0.5 + 0.1 × std_DPF / mean_DPF = 0.5 + 0.1 × 1.767 / 0.3 = 1.089. The integral gain Ki is set to 1 / 10 of Kp, i.e., Ki = 0.1089. The real-time transfer function of the loop filter is obtained as Hlp(s) = 1.089 + 0.1089 / s, where s is a complex frequency domain variable.

[0131] The state equation of the virtual inertia control system is constructed based on the power fluctuation factor (DPF). The virtual inertia control system simulates the mechanical characteristics of a synchronous generator and achieves stable control of the grid frequency by adjusting the virtual moment of inertia and virtual damping coefficient. The matrix form of the state equation is X' = AX + BU, where X is the state vector, including angular velocity deviation and power angle deviation; A is the system matrix; B is the input matrix; and U is the input vector, i.e., the power fluctuation factor (DPF).

[0132] The virtual moment of inertia J and virtual damping coefficient D are determined using the variation characteristics of the power fluctuation DPF. The virtual moment of inertia J is calculated by the maximum fluctuation amplitude of DPF, J = 5.0 + 0.5×max(abs(DPF)) = 5.0 + 0.5×3.0 = 6.5 kg·m². The virtual damping coefficient D is calculated by the average rate of change of DPF, the average rate of change is (abs(DPF[1]-DPF[0]) + ... + abs(DPF[9]-DPF[8])) / 9 = (1.0 + 1.0 + 1.0 + 2.0 + 1.0 + 3.0 + 2.0 + 4.0 + 0.0) / 9 = 1.667 kW / s, and the virtual damping coefficient D = 3.0 + 0.8×1.667 = 4.334 N·m·s / radian.

[0133] Substituting the virtual moment of inertia J and the virtual damping coefficient D into the state equation, the angular velocity of the virtual synchronizer is obtained by solving. In the state equation, the elements of system matrix A are a11 = -D / J = -4.334 / 6.5 = -0.667, a12 = -K / J, where K is the synchronization torque coefficient with a value of 5.0, therefore a12 = -5.0 / 6.5 = -0.769; a21 = 1.0, a22 = 0.0. The elements of input matrix B are b1 = 1 / J = 1 / 6.5 = 0.154, b2 = 0.0.

[0134] Taking the 10th value of DPF, 3.0 kW, as an example, the angular velocity deviation calculated using the state equation is 0.272 radians / second, which translates to a frequency deviation of 0.272 / (2×π) = 0.043 Hz. The angular velocity value of the virtual synchronizer is ω = 2×π×50 + 0.272 = 314.159 + 0.272 = 314.431 radians / second, corresponding to a frequency of 50.043 Hz.

[0135] The frequency compensation of the phase-locked loop is calculated by combining the angular velocity value ω with the real-time transfer function Hlp(s). The frequency compensation Δf is obtained by inputting the angular velocity deviation into the loop filter. Taking an angular velocity deviation of 0.272 radians / second as an example, the corresponding frequency deviation is 0.043 Hz. The frequency compensation Δf after loop filter processing is: Δf = Kp×0.043 + Ki×∫0.043dt = 1.089×0.043 + 0.1089×0.043×t|_0 1 = 0.047 + 0.005 = 0.052 Hz.

[0136] The frequency compensation Δf is superimposed on the center frequency fc to obtain the real-time operating frequency of the phase-locked loop, f = fc + Δf = 50.581 + 0.052 = 50.633 Hz. Based on the real-time operating frequency f, the phase-locked loop error ε is determined. The phase-locked error ε is the difference between the real-time operating frequency f and the target frequency ftarget (usually 50 Hz), ε = f - ftarget = 50.633 - 50.0 = 0.633 Hz.

[0137] The phase-locked error ε is simultaneously fed back to the input of the loop filter and the input of the virtual inertia control system to complete the synchronous switching of the power supply. In practical applications, the phase-locked error is converted into a corresponding phase adjustment signal by a voltage-controlled oscillator, driving the switching equipment to achieve synchronous switching. Taking a coal mine substation as an example, during the switching process between the main power supply and the backup power supply, a phase-locked error of 0.633 Hz was detected, and the calculated phase adjustment was 22.8 degrees. By precisely controlling the switching time, the switching is ensured to be completed when the phase difference is minimal, thus minimizing the phase impact at the moment of switching and ensuring a smooth transition of the power supply system.

[0138] In one optional implementation, a transitional power supply mechanism based on supercapacitors is established. The optimal power supply path scheme is segmented using a backtracking algorithm to determine the buffer timing of each stage, ensuring a smooth transition of the power supply state. This includes:

[0139] A transitional power supply mechanism is constructed, dividing the power supply process into three stages: disconnection of the original path, transitional power supply by supercapacitor, and connection of the new path, and setting switching thresholds for each stage;

[0140] The load power demand curve is collected based on the aforementioned transitional power supply mechanism, and the power compensation amount is calculated in combination with the switching threshold of each stage. The changing characteristics of the load power demand curve are analyzed, a power balance equation is established, and the compensation power output characteristics of the supercapacitor at each stage are determined.

[0141] The power matching degree of the power supply path is evaluated based on the power compensation amount and the compensation power output characteristics, and segmentation criteria are set in combination with energy utilization rate;

[0142] A backtracking algorithm is used to search for segmentation points from the end point of the optimal power supply path backward. The criterion value of each candidate segmentation point is calculated based on the segmentation criterion. When the difference in the criterion values ​​of adjacent candidate segmentation points is less than the preset segmentation threshold, the current candidate segmentation point is marked as the optimal segmentation point. The search is repeated until the optimal power supply path scheme has been traversed.

[0143] The optimal power supply path is divided into power supply segments using the optimal segmentation point. Based on the power supply segments, the buffer timing of the three stages in the transitional power supply mechanism is determined to ensure a smooth transition of the power supply state.

[0144] Taking the power supply system of a coal mining face as an example, the voltage drop threshold during the original path disconnection phase is set to 90% of the rated voltage, that is, the disconnection operation is triggered when the voltage drops below 90% of the rated value; the discharge depth threshold during the supercapacitor transition power supply phase is set to 80%, that is, the new path connection begins when the supercapacitor discharges to 80%; the voltage stability threshold during the new path connection phase is set to 95% of the rated voltage, that is, the switching is completed when the new path voltage stabilizes above 95% of the rated value.

[0145] Based on the constructed transitional power supply mechanism, load power demand curves are collected. Taking the main ventilation fan of a coal mine as an example, its power demand curve changes periodically over 24 hours. The data sampling points are [380, 410, 450, 420, 390, 385, 420, 480, 460, 430, 420, 410] kW, with a sampling interval of 2 hours. Power compensation is calculated based on the switching thresholds at each stage. The power compensation represents the power required by the supercapacitor during the switching process. During the original path disconnection stage, the power compensation equals the load power demand; during the supercapacitor transition power supply stage, the power compensation is the difference between the load power demand and the initial power of the new path; during the new path connection stage, the power compensation gradually decreases to zero. Taking the sampling point at 480 kW as an example, if a power switch occurs at this time, the power compensation amount during the original path disconnection stage is 480 kW, the power compensation amount during the supercapacitor transition power supply stage is initially 480 kW, and then gradually decreases according to the new path access situation. The power compensation amount during the new path access stage gradually decreases from 120 kW to 0 kW.

[0146] Analyze the changing characteristics of the load power demand curve and establish a power balance equation. The power balance equation is expressed as the sum of the load power demand PL, the original path power PO, the supercapacitor power PC, and the new path power PN, i.e., PL = PO + PC + PN. The equation exhibits different characteristics in the three stages of power supply switching: In the original path disconnection stage (t∈[0,T1]), PO gradually decreases from its initial value PL to 0, PC gradually increases from 0 to PL, and PN remains 0. Therefore, PL = PO(t) + PC(t), where PO(t) = PL × (1 - t / T1) and PC(t) = PL × t / T1; In the supercapacitor transition power supply stage (t∈[T1,T1+T2]), PO remains 0, PC remains PL, and PN remains 0. Therefore, PL = PC = PL; In the new path connection stage (t∈[T1+T2,T1+T2+T3]), PO remains 0, PC gradually decreases from PL to 0, and PN gradually increases from 0 to PL. Therefore, PL = PC(t) + PN(t), where PC(t) = PL × (1 - (t-T1-T2) / T3) and PN(t) = PL × (t-T1-T2) / T3.

[0147] Taking a coal mine main drainage pump as an example, its rated power is 320 kilowatts. During the power supply switching process, the duration of the original path disconnection phase is T1=0.15 seconds, the duration of the supercapacitor transition power supply phase is T2=0.3 seconds, and the duration of the new path connection phase is T3=0.2 seconds. During the original path disconnection phase, at t=0.075 seconds, PO(0.075) = 320 × (1 - 0.075 / 0.15) = 160 kW, PC(0.075) = 320 × 0.075 / 0.15 = 160 kW. The power balance equation is verified as: PL = 160 + 160 + 0 = 320 kW. During the supercapacitor transition power supply phase, at t=0.3 seconds, PO = 0 kW, PC = 320 kW, PN = 0 kW. The power balance equation is verified as: PL = 0 + 320 + 0 = 320 kW. During the new path connection phase, at t=0.55 seconds, PO = 0 kW, PC(0.55) = 320 × (1 - (0.55-0.15-0.3) / 0.2) = 160 kW, PN(0.55) = 320 × (0.55-0.15-0.3) / 0.2 = 160 kW, power balance equation verification: PL = 0 + 160 + 160 = 320 kW.

[0148] Based on the power balance equation, the compensation power output characteristics of the supercapacitor at each stage are determined. During the original path disconnection stage, the supercapacitor compensation power exhibits a linear increasing characteristic, rapidly rising from 0 to full load power. During the supercapacitor transition power supply stage, the compensation power remains at full load level. During the new path connection stage, the compensation power exhibits a linear decreasing characteristic until it drops to 0. For loads with large power fluctuations, such as coal mining machines, their power demand may change during the switching process. In this case, the power balance equation needs to consider the load power change factor. Assuming that the load power changes from PL1 to PL2 during the switching process, the power balance equation is corrected to PL(t) = PO(t) + PC(t) + PN(t), where PL(t) = PL1 + (PL2 -PL1) × t / (T1+T2+T3).

[0149] The power matching degree of the power supply path is evaluated based on the power compensation amount and the output characteristics of the compensated power. The calculation method is the degree to which the ratio of the two approaches 1, ranging from 0 to 1; a larger value indicates a higher matching degree. Taking a coal mining face as an example, the power demand at different times is [250, 320, 280, 230] kW, and the corresponding power supply path provides power of [260, 305, 290, 240] kW. The calculated power matching degrees are 1-|260-250| / 250=0.96, 1-|305-320| / 320=0.95, 1-|290-280| / 280=0.96, and 1-|240-230| / 230=0.96, respectively. Segmented criteria are set in conjunction with energy utilization rate. Energy utilization rate represents the energy utilization efficiency of the supercapacitor energy storage system during the switching process, and is calculated as the ratio of output energy to stored energy. Taking a certain type of supercapacitor module as an example, its stored energy is 2000 kJ, and its output energy during the switching process is 1850 kJ, with an energy utilization rate of 1850 / 2000 = 0.925. The segmentation criterion J is calculated by weighting the power matching degree PM and the energy utilization rate EU, J = 0.6 × PM + 0.4 × EU. For the four time periods of the above-mentioned coal mining face, the segmentation criterion values ​​are 0.6 × 0.96 + 0.4 × 0.925 = 0.946, 0.6 × 0.95 + 0.4 × 0.925 = 0.94, 0.6 × 0.96 + 0.4 × 0.925 = 0.946, and 0.6 × 0.96 + 0.4 × 0.925 = 0.946, respectively.

[0150] A backtracking algorithm is used to search for segmentation points from the endpoint of the optimal power supply path backwards. Taking the optimal path S→A→B→C→D→E→F→L of a coal mine power supply system as an example, the search starts from the endpoint L. The criterion value for each candidate segmentation point is calculated based on the previously set segmentation criteria. Assume the criterion values ​​for nodes A, B, C, D, E, and F are 0.92, 0.94, 0.91, 0.93, 0.95, and 0.92, respectively. When the difference in criterion values ​​between adjacent candidate segmentation points is less than a preset segmentation threshold, the current candidate segmentation point is marked as the optimal segmentation point. Set the preset segmentation threshold to 0.03. Search forward from the endpoint L, calculate the difference in criterion values ​​between F and L: |0.92-0.90|=0.02<0.03, and mark F as the optimal segmentation point. Continue searching forward, calculate the difference in criterion values ​​between E and F: |0.95-0.92|=0.03=0.03, and mark E as the optimal segmentation point. Calculate the difference in criterion values ​​between D and E: |0.93-0.95|=0.02<0.03. 03. Do not mark D as the optimal segmentation point; calculate the difference in criterion values ​​between C and D: |0.91-0.93|=0.02<0.03, do not mark C as the optimal segmentation point; calculate the difference in criterion values ​​between B and C: |0.94-0.91|=0.03=0.03, mark B as the optimal segmentation point; calculate the difference in criterion values ​​between A and B: |0.92-0.94|=0.02<0.03, do not mark A as the optimal segmentation point. Repeat the search until all optimal power supply path schemes have been traversed, and the determined optimal segmentation points are B, E, and F.

[0151] The optimal power supply path is divided into power supply segments using the optimal segmentation point. Based on the example above, the optimal power supply path S→A→B→C→D→E→F→L is divided into four power supply segments: S→A→B, B→C→D→E, E→F, and F→L. The buffer timing of the three stages in the transitional power supply mechanism is determined based on these power supply segments. The buffer timing refers to the duration of each stage during the switching process. For each power supply segment, the times for the original path disconnection, supercapacitor transition power supply, and new path access are calculated separately. Taking power supply segment B→C→D→E as an example, the original path disconnection stage time T1 is calculated as the number of nodes multiplied by the basic disconnection time: T1 = 3 × 0.05 = 0.15 seconds; the supercapacitor transition power supply stage time T2 is calculated as the path length divided by the system response speed: T2 = 3 / 10 = 0.3 seconds; the new path access stage time T3 is calculated as the path average load rate multiplied by the basic access time. Assuming the path average load rate is 0.85, T3 = 0.85 × 0.25 = 0.21 seconds. Therefore, the total buffer time for this power supply segment is T1+T2+T3=0.15+0.3+0.21=0.66 seconds.

[0152] In practical applications, this transitional power supply mechanism effectively ensures a smooth transition of power supply status. Taking a coal mine drainage system as an example, three sets of supercapacitor energy storage units are deployed along its power supply path, each with a capacity of 500 farads and a rated voltage of 400 volts. When the main power supply fails and a switch to the backup power supply is required, the switching process is controlled according to a pre-calculated buffer sequence. The original path disconnection phase lasts 0.12 seconds, during which the supercapacitor energy storage units provide all the power required by the drainage system, approximately 150 kilowatts. The supercapacitor transition power supply phase lasts 0.25 seconds, during which the supercapacitor output power remains at 150 kilowatts. The new path connection phase lasts 0.18 seconds, during which the supercapacitor output power gradually decreases from 150 kilowatts to 0, while the new path power gradually increases from 0 to 150 kilowatts. Throughout the entire switching process, the speed fluctuation of the drainage system motor does not exceed 3% of the rated speed, and the voltage fluctuation does not exceed 5% of the rated voltage, ensuring the continuity and stability of system operation.

[0153] In one alternative implementation, it further includes:

[0154] The switching delay and response time are calculated based on the power compensation amount of the power supply segment;

[0155] The switching delay and the response time are used to plan the buffer timing of the power supply segment. The switching delay is used as the buffer power supply advance. The buffer exit time is set according to the timing relationship of adjacent power supply segments. The minimum duration of the buffer time window is determined by the response time.

[0156] Power supply switching is performed according to the buffer timing and the buffer exit time. The supercapacitor is activated in advance to provide buffer power supply before the original path is disconnected. During the transition phase, power is supplied according to the compensated power output characteristics and the output power is dynamically adjusted. When the new path is connected, power supply is withdrawn in stages according to the power gradual change rule and the buffer exit time. The switching frequency is adjusted in real time according to the load power change rate to achieve smooth switching of power supply path.

[0157] Based on the power compensation amount of the aforementioned power supply segments, the switching delay and response time are calculated. For a coal mine main drainage pump power supply system, the power compensation amount is 320 kW. The switching delay TD is calculated based on this power compensation amount. The switching delay calculation formula is TD = TB + K1 × PC, where TB is the base delay, set to 0.05 seconds, K1 is the power coefficient, set to 0.0002 seconds / kW, and PC is the power compensation amount. Substituting the data, we get TD = 0.05 + 0.0002 × 320 = 0.114 seconds. The response time TR is calculated using the formula TR = TM + K2 × PC + K3 × VC, where TM is the equipment mechanical response time, set to 0.08 seconds, K2 is the power response coefficient, set to 0.0001 seconds / kW, K3 is the voltage response coefficient, set to 0.001 seconds / volt, and VC is the system rated voltage, which is 380 volts. Substituting the data, we get TR = 0.08 + 0.0001 × 320 + 0.001 × 380 = 0.112 + 0.38 = 0.492 seconds.

[0158] For different types of loads, the switching delay and response time vary. Taking a coal mining face in a coal mine as an example, it includes three types of loads: coal mining machine, conveyor and support system, with power compensation of 250 kW, 180 kW and 120 kW respectively. The switching delay of the coal mining machine is calculated as TD1 = 0.05 + 0.0002 × 250 = 0.1 seconds, and the response time is TR1 = 0.08 + 0.0001 × 250 + 0.001 × 380 = 0.105 + 0.38 = 0.485 seconds; the switching delay of the conveyor is calculated as TD2 = 0.05 + 0.0002 × 180 = 0.086 seconds, and the response time is TR2 = 0.08 + 0.0001 × 180 + 0.001 × 380 = 0.098 + 0.38 = 0.478 seconds; the switching delay of the support system is calculated as TD3 = 0.05 + 0.0002 × 120 = 0.074 seconds, and the response time is TR3 = 0.08 + 0.0001 × 120 + 0.001 × 380 = 0.092 + 0.38 = 0.472 seconds.

[0159] The buffer timing of the power supply segment is planned using the calculated switching delay and response time. The buffer timing includes the buffer power supply advance, buffer exit time, and the minimum duration of the buffer time window. The switching delay is used as the buffer power supply advance, that is, the supercapacitor is started for buffer power supply TD time before the original path is disconnected. Taking the aforementioned main drainage pump as an example, the buffer power supply advance is 0.114 seconds, that is, the supercapacitor is started for power supply 0.114 seconds before the original path is disconnected. The buffer exit time is set according to the timing relationship of adjacent power supply segments. The buffer exit time TE is calculated by the formula TE = K4 × TR + K5 × PD, where K4 is the response time coefficient, set to 1.2, K5 is the power change coefficient, set to 0.0003 seconds / kW, and PD is the absolute value of the power difference between adjacent power supply segments. Assuming the power difference between adjacent power supply sections of the main drainage pump is 50 kilowatts, the buffer exit time TE = 1.2 × 0.492 + 0.0003 × 50 = 0.59 + 0.015 = 0.605 seconds.

[0160] The minimum duration of the buffer window is determined by the response time. The minimum duration TW represents the shortest time the supercapacitor supplies power to ensure smooth power switching. The formula for calculating the minimum duration is TW = max(TR, TD + TE), which is the larger of the sum of the response time, switching delay, and buffer exit time. For the main drain pump, TW = max(0.492, 0.114 + 0.605) = max(0.492, 0.719) = 0.719 seconds. This means the supercapacitor needs to supply power continuously for at least 0.719 seconds to ensure a smooth transition during power switching.

[0161] For systems with multiple power supply segments, the buffer timing for each segment needs to be calculated separately. Taking a coal mine's main ventilation system as an example, its power supply path is divided into three segments: S1, S2, and S3, with power compensation values ​​of 420 kW, 380 kW, and 400 kW, respectively. The calculated switching delays are TD1 = 0.134 seconds, TD2 = 0.126 seconds, and TD3 = 0.13 seconds; the response times are TR1 = 0.502 seconds, TR2 = 0.498 seconds, and TR3 = 0.5 seconds; the buffer exit times are TE1 = 0.627 seconds, TE2 = 0.615 seconds, and TE3 = 0.63 seconds; and the minimum durations are TW1 = 0.761 seconds, TW2 = 0.741 seconds, and TW3 = 0.76 seconds.

[0162] Power supply switching is performed based on the calculated buffer timing and buffer exit time. A supercapacitor is activated in advance to provide buffer power before the original path is disconnected. Taking a coal mine main hoist as an example, its switching delay is 0.15 seconds. After detecting a signal that the original path needs to be disconnected, the control system immediately activates the supercapacitor to provide power, and waits 0.15 seconds before performing the original path disconnection operation.

[0163] During the transition phase, power is supplied according to the compensated power output characteristics, and the output power is dynamically adjusted. The output power of the supercapacitor is not fixed but dynamically adjusted according to load demand and system status. Taking a substation in a mining area as an example, during the transition phase, the load power demand increases from 350 kW to 380 kW, and the output power of the supercapacitor needs to increase accordingly. Setting the power regulation coefficient of the supercapacitor to 1.1, the maximum output power of the supercapacitor is 380 × 1.1 = 418 kW, exceeding the maximum power demand of the load, ensuring a continuous and stable power supply during load power fluctuations.

[0164] When a new power supply path is connected, power is gradually withdrawn according to a power gradation rule and a buffer withdrawal time. The power gradation rule describes how the power of the new path gradually increases and how the power of the supercapacitor gradually decreases. Taking a coal mine drainage system as an example, its buffer withdrawal time is 0.6 seconds and its power is 150 kW. The new path connection process is divided into three stages: the initial stage (accounting for 20% of the buffer withdrawal time), where the power of the new path increases from 0 to 30 kW (accounting for 20% of the total power), and the power of the supercapacitor decreases from 150 kW to 120 kW; the intermediate stage (accounting for 50% of the buffer withdrawal time), where the power of the new path increases from 30 kW to 105 kW (from 20% to 70%), and the power of the supercapacitor decreases from 120 kW to 45 kW; and the final stage (accounting for 30% of the buffer withdrawal time), where the power of the new path increases from 105 kW to 150 kW (from 70% to 100%), and the power of the supercapacitor decreases from 45 kW to 0.

[0165] The switching frequency is adjusted in real time based on the load power change rate to achieve smooth switching of the power supply path. The load power change rate represents the change in load power per unit time. The larger the change rate, the higher the switching frequency needs to be for more precise control of the switching process. The switching frequency adjustment formula is f = f0 × (1 + K6 × |dP / dt|), where f0 is the base switching frequency, set to 100 Hz, K6 is the change rate coefficient, set to 0.02 s / kW, and |dP / dt| is the absolute value of the load power change rate in kW / s. Taking a coal mining machine as an example, if its power changes at a rate of 50 kW / s during the switching process, then the adjusted switching frequency f = 100 × (1 + 0.02 × 50) = 100 × 2 = 200 Hz.

[0166] Compared with traditional switching methods, voltage fluctuations during switching are reduced by 75% and power surges by 80%, significantly improving the stability and reliability of coal mine power supply systems and providing strong support for safe coal mine production.

[0167] A second aspect of this invention provides a multi-source collaborative supercapacitor-based emergency power supply dispatching and optimization system for coal mines, comprising:

[0168] The first unit is used to acquire the real-time operating parameters of the main power supply, backup power supply and supercapacitor energy storage system in the coal mine power supply system, and to construct the node state feature matrix of the power supply network using the real-time operating parameters.

[0169] The second unit is used to decompose the node state feature matrix into trend and periodic terms, extract the dynamic features of the nodes, calculate the transmission efficiency and reliability indicators of the power supply line based on the dynamic features, and establish a node priority evaluation system.

[0170] The third unit is used to classify each node according to the node priority evaluation system, combine feature aggregation operation to generate power supply path score, perform global optimization based on the power supply path score, and output the optimal power supply path scheme.

[0171] The fourth unit is used to calculate the dependency topology relationship between each node based on the power supply path score, thereby constructing a power supply equipment switching control strategy and determining the optimal switching path;

[0172] The fifth unit is used to decompose and reconstruct the power supply path score to obtain the voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, the phase-locked loop state matrix and the state equation of the virtual inertia control system are constructed respectively. Through the synergistic effect of the loop filter, the synchronous switching of the power supply is completed.

[0173] The sixth unit is used to establish a transitional power supply mechanism based on supercapacitors. It uses a backtracking algorithm to segment the optimal power supply path scheme, determine the buffer timing of each stage, and ensure a smooth transition of the power supply state.

[0174] A third aspect of the present invention provides an electronic device, comprising:

[0175] processor;

[0176] Memory used to store processor-executable instructions;

[0177] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0178] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0179] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing emergency power supply dispatch in coal mines based on multi-source coordination using supercapacitors, characterized in that... include: The real-time operating parameters of the main power supply, backup power supply and supercapacitor energy storage system in the coal mine power supply system are obtained, and the node state feature matrix of the power supply network is constructed using the real-time operating parameters. The node state feature matrix is ​​decomposed into trend and periodic terms to extract the dynamic features of the nodes. Based on the dynamic features, the transmission efficiency and reliability indicators of the power supply line are calculated, and a node priority evaluation system is established. Each node is classified according to the node priority evaluation system, and a power supply path score is generated by combining feature aggregation operation. Based on the power supply path score, global optimization is performed to output the optimal power supply path scheme. Based on the power supply path score, the dependency topology relationship between each node is calculated to construct a power supply equipment switching control strategy and determine the optimal switching path. The power supply path score is decomposed and reconstructed to obtain voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, the state matrix of the phase-locked loop and the state equation of the virtual inertia control system are constructed respectively. Through the synergistic effect of the loop filter, the synchronous switching of the power supply is completed. A transitional power supply mechanism based on supercapacitors is established. The optimal power supply path scheme is segmented using a backtracking algorithm to determine the buffer timing of each stage, ensuring a smooth transition of the power supply state. This includes: A transitional power supply mechanism is constructed, dividing the power supply process into three stages: disconnection of the original path, transitional power supply by supercapacitor, and connection of the new path, and setting switching thresholds for each stage; Based on the aforementioned transitional power supply mechanism, the load power demand curve is collected, and the power compensation amount is calculated in combination with the switching threshold of each stage. Analyze the changing characteristics of the load power demand curve, establish a power balance equation, and determine the compensation power output characteristics of the supercapacitor at each stage. The power matching degree of the power supply path is evaluated based on the power compensation amount and the compensation power output characteristics, and segmentation criteria are set in combination with energy utilization rate; A backtracking algorithm is used to search for segmentation points from the end of the optimal power supply path backward, and the criterion value of each candidate segmentation point is calculated based on the segmentation criterion. When the difference in the criterion values ​​of adjacent candidate segment points is less than the preset segmentation threshold, the current candidate segment point is marked as the optimal segment point, and the search is repeated until the optimal power supply path scheme is traversed. The optimal power supply path is divided into power supply segments using the optimal segmentation point. Based on the power supply segments, the buffer timing of the three stages in the transitional power supply mechanism is determined to ensure a smooth transition of the power supply state.

2. The method according to claim 1, characterized in that, The node state feature matrix is ​​decomposed into trend and periodic terms to extract the dynamic features of the nodes. Based on these dynamic features, the transmission efficiency and reliability indicators of the power supply line are calculated, and a node priority evaluation system is established, including: The variational mode decomposition method is used to decompose the state parameters of each node in the node state feature matrix into intrinsic mode functions. The frequency distribution characteristics of the intrinsic mode functions are calculated to obtain the spectral entropy. The intrinsic mode functions with spectral entropy higher than a preset spectral threshold are superimposed to form a trend term, and the rest are superimposed to form a periodic term. Change features and fluctuation features are extracted from the trend item and the period item respectively, and the dynamic features of the node are constructed using the change features and the fluctuation features; The variation trend of node load is determined based on the variation characteristics, the fluctuation of node voltage and current is determined based on the fluctuation characteristics, and the transmission efficiency is constructed based on the variation trend and the fluctuation. The dynamic features and real-time operating parameters are input into a pre-trained reliability assessment model to obtain a reliability index. A node priority evaluation system is established based on the ant colony algorithm, whereby the transmission efficiency is set as forward guidance information and the reliability index is set as backward feedback information. The pheromone concentration is updated using the forward guidance information and the backward feedback information, and the priority of nodes is determined based on the pheromone concentration.

3. The method according to claim 1, characterized in that, Based on the node priority evaluation system, each node is classified, and a power supply path score is generated by combining feature aggregation operations. Global optimization is then performed based on the power supply path score to output the optimal power supply path scheme, including: Obtain the pheromone concentration of each node in the node priority evaluation system, construct a weighted undirected graph based on the pheromone concentration, generate an importance score using degree centrality and betweenness centrality, classify nodes with an importance score greater than a preset key score as key protection nodes, and classify the rest as regular management nodes. The dynamic features and adjacency relationships of nodes are input into the graph convolutional layer. After the adjacency matrix is ​​processed by degree matrix normalization, the output features of the graph convolutional layer are obtained. The pheromone concentration is used as the weight coefficient of the neighboring nodes to construct priority-aware features; The output features of the graph convolutional layer and the priority-aware features are residually connected to form a fused feature, and the fused feature is input into a multilayer perceptron to generate a power supply path score; Based on the power supply path score, a state transition equation is constructed, and a forward dynamic programming method is used to traverse layer by layer, taking the current accumulated power supply path score and the importance score of the node as the node state value. In each decision-making stage, the optimal value for the stage is calculated iteratively using the Bellman equation by combining the adjacency relationship of the nodes and the node state value. Based on the optimal value of the aforementioned stage, the optimal power supply path scheme is obtained by backtracking.

4. The method according to claim 1, characterized in that, The dependency topology relationships between nodes are calculated based on the power supply path score to construct a power supply equipment switching control strategy and determine the optimal switching path, including: Node feature vectors are constructed based on the voltage state and load characteristics of nodes. The correlation between nodes is calculated using power supply path scoring. The strength of the dependency between nodes is determined based on the cosine similarity of the node feature vectors and the correlation. Based on the inter-node dependency strength, the sum of the dependency strengths between each node and its neighboring nodes is calculated, a power supply equipment switching control strategy is constructed, and different strategy levels are divided according to the sum of the dependency strengths. Nodes whose total dependency intensity is greater than the preset high-intensity threshold adopt a dual-power supply and real-time monitoring strategy; nodes between the preset high-intensity threshold and the preset low-intensity threshold adopt a fast switching and periodic monitoring strategy; and nodes less than the preset low-intensity threshold adopt a basic switching and timed monitoring strategy. Set corresponding monitoring cycles and switching time parameters for nodes at different policy levels; Based on the inter-node dependency strength and the node feature vectors, a feature propagation matrix is ​​constructed. The switching path evaluation value is calculated by iteratively propagating the node feature vectors on the graph structure. The path with the largest switching path evaluation value is selected as the optimal switching path. A switching sequence is generated from high to low based on the total dependence intensity, and the power supply equipment switching operation is performed along the optimal switching path according to the monitoring period and the switching time.

5. The method according to claim 1, characterized in that, The power supply path score is decomposed and reconstructed to obtain voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, a phase-locked loop state matrix and a virtual inertia control system state equation are constructed respectively. Through the synergistic effect of the loop filter, synchronous switching of the power supply is completed, including: The power supply path score is adaptively decomposed using empirical mode decomposition, and the voltage quality coefficient and power characteristic coefficient are reconstructed by Hilbert transform. The voltage phase deviation is obtained by feature extraction of the voltage quality coefficient, and the power fluctuation is obtained by feature extraction of the power characteristic coefficient. A phase-locked loop state matrix is ​​constructed based on the voltage phase deviation, and the center frequency of the phase-locked loop is calculated based on the state matrix. The power fluctuation is input to the loop filter, and the scaling factor of the loop filter is adjusted to obtain the real-time transfer function of the loop filter. Based on the power fluctuation, a state equation for a virtual inertia control system is constructed, and the virtual moment of inertia and virtual damping coefficient are determined using the variation characteristics of the power fluctuation. Substituting the virtual moment of inertia and the virtual damping coefficient into the state equation, the angular velocity value of the virtual synchronizer is obtained by solving the equation. The angular velocity value is combined with the real-time transfer function to calculate the frequency compensation of the phase-locked loop. The frequency compensation amount is superimposed on the center frequency to obtain the real-time operating frequency of the phase-locked loop; The phase-locked loop error is determined based on the real-time operating frequency; The phase-locked error is simultaneously fed back to the input of the loop filter and the input of the virtual inertia control system to complete the synchronous switching of the power supply.

6. The method according to claim 1, characterized in that, Also includes: The switching delay and response time are calculated based on the power compensation amount of the power supply segment; The switching delay and the response time are used to plan the buffer timing of the power supply segment. The switching delay is used as the buffer power supply advance. The buffer exit time is set according to the timing relationship of adjacent power supply segments. The minimum duration of the buffer time window is determined by the response time. Power supply switching is performed according to the buffer timing and the buffer exit time. The supercapacitor is activated in advance to provide buffer power supply before the original path is disconnected. During the transition phase, power is supplied according to the compensated power output characteristics and the output power is dynamically adjusted. When the new path is connected, power supply is withdrawn in stages according to the power gradual change rule and the buffer exit time. The switching frequency is adjusted in real time according to the load power change rate to achieve smooth switching of power supply path.

7. A multi-source collaborative emergency power supply dispatching and optimization system for coal mines based on supercapacitors, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to acquire the real-time operating parameters of the main power supply, backup power supply and supercapacitor energy storage system in the coal mine power supply system, and to construct the node state feature matrix of the power supply network using the real-time operating parameters. The second unit is used to decompose the node state feature matrix into trend and periodic terms, extract the dynamic features of the nodes, calculate the transmission efficiency and reliability indicators of the power supply line based on the dynamic features, and establish a node priority evaluation system. The third unit is used to classify each node according to the node priority evaluation system, combine feature aggregation operation to generate power supply path score, perform global optimization based on the power supply path score, and output the optimal power supply path scheme. The fourth unit is used to calculate the dependency topology relationship between each node based on the power supply path score, thereby constructing a power supply equipment switching control strategy and determining the optimal switching path; The fifth unit is used to decompose and reconstruct the power supply path score to obtain the voltage phase deviation and power fluctuation. Based on the voltage phase deviation and power fluctuation, the phase-locked loop state matrix and the state equation of the virtual inertia control system are constructed respectively. Through the synergistic effect of the loop filter, the synchronous switching of the power supply is completed. The sixth unit is used to establish a transitional power supply mechanism based on supercapacitors. It uses a backtracking algorithm to segment the optimal power supply path scheme, determine the buffer timing of each stage, and ensure a smooth transition of the power supply state.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images