A motor control management system

Abstract

The present invention belongs to the technical field of motor management, and discloses a motor control and management system; the present invention comprises: collecting historical motor comprehensive data, pre-processing the historical motor comprehensive data, and obtaining fused feature data; clustering and grouping the motors based on the fused feature data, and obtaining n motor groups; establishing a grid model of the motors, updating the states of the motors in each motor group according to the grid model, and simulating and obtaining the aging curve and failure rate of the corresponding motors within a preset time; determining the optimal management path for each motor group based on the aging curve and failure rate of the motors within the preset time using a path planning algorithm, and executing the optimal management path to perform patrol management on the motors in each motor group; the optimal management path can be planned, so that management resources are reasonably allocated and utilized, ensuring timely maintenance, thereby extending the service life of the motors and reducing operation and maintenance costs.

Description

Technical Field

[0001] The present invention relates to the technical field of motor management, and more particularly, to a motor control management system. Background Art

[0002] Motors are essential, key equipment in power systems and are widely used in power plants, substations, transmission lines, and other sectors. The safe and reliable operation of motors is directly related to the stability and power supply quality of the entire power system. However, due to the complex and changing working environment of motors, their operating status is affected by many factors, which inevitably leads to aging and failure of motors during long-term operation. To promptly detect and prevent motor failures and ensure the safe and stable operation of the power system, effective monitoring and management of motors is required. Traditional motor management methods rely mainly on regular manual inspections and maintenance, which are somewhat delayed and passive. With the continuous expansion of the power system and the rapid growth in the number of motors, traditional management methods have become difficult to meet actual needs.

[0003] Existing motor management methods still have many shortcomings. First, it is impossible to effectively integrate and utilize various motor data, making it difficult to comprehensively assess the comprehensive status of the motor, resulting in low accuracy in status diagnosis and fault prediction. Second, there is a lack of the ability to perform fine-grained modeling and simulation of motors, making it difficult to provide a basis for formulating forward-looking maintenance strategies. Finally, existing motor management and maintenance methods are inefficient and mainly rely on manual experience for planning. They cannot fully consider the degree of motor aging and failure risks, resulting in inaccurate allocation of maintenance resources and high maintenance costs. At the same time, the manual operation mode has a low degree of automation, high work intensity, and is prone to fatigue and mistakes, further affecting management efficiency.

[0004] In view of this, the present invention proposes a motor control management system to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a motor control and management system, comprising: a data acquisition and fusion module, for acquiring historical motor comprehensive data, performing preliminary comprehensive alignment on the historical motor comprehensive data, and obtaining a preliminary fusion data set;

[0006] The preliminary processing module is used to extract the discrete data in the preliminary fusion data set to form a discrete data set, perform discrete fusion processing on the discrete data set to obtain a comprehensive feature vector; extract the continuous data in the preliminary fusion data set according to different data acquisition sources to form several continuous data sets; perform continuous fusion processing on several continuous data sets to obtain fusion features; and splice the fusion features and the comprehensive feature vector to obtain fusion feature data;

[0007] The clustering module clusters and groups the motors based on the fused feature data to obtain n motor groups;

[0008] The model establishment and prediction module is used to establish a grid model of the motor, update the status of the motors in each motor group according to the grid model, and simulate the aging curve and failure rate of the corresponding motors within a preset time;

[0009] The path planning module uses a path planning algorithm to determine the optimal management path for each motor group based on the motor's aging curve and failure rate within a preset time, and executes the optimal management path to inspect the motors in each motor group; each module is connected to each other via wired and / or wireless means.

[0010] Furthermore, the historical motor comprehensive data includes power action data, motor operation data and geographic location data; the power action data includes switch status data and protection action data;

[0011] The motor operation data includes motor temperature, motor humidity, motor vibration, motor oil level, motor voltage, motor current and motor power; the geographic location data is the geographic coordinate information of the motor.

[0012] Furthermore, the historical motor comprehensive data are preliminarily aligned to obtain a preliminary fused data set, including: cleaning the historical motor comprehensive data, unifying the formats of data from different sources in the historical motor comprehensive data, and temporally aligning and spatially aligning data from different sources of the same motor to obtain a preliminary fused data set.

[0013] Furthermore, the method of performing discrete fusion processing on the discrete data sets includes:

[0014] The discrete data set is constructed into a decision table. The decision table consists of a condition attribute set and a decision attribute set. The condition attribute set includes several condition attribute value vectors, and the decision attribute value set includes several decision attribute values. Each object in the decision table corresponds to a condition attribute value vector and a decision attribute value. All objects constitute an object set U.

[0015] For each attribute subset in the conditional attribute set B, the objects in the object set U are grouped according to the value of the attribute subset, and each group forms an equivalence class;

[0016] For any attribute subset X, calculate the upper and lower deterministic sets of X under the conditional attribute value vector in B; the upper deterministic set contains all objects that completely belong to X, and the lower deterministic set contains the equivalence classes of some objects that belong to X;

[0017] Through the cardinality of the upper and lower determination sets, calculate the roughness of X under the conditional attribute value vector in B for the decision attribute value d;

[0018] The calculation formula for roughness is:

[0019] ;in, represents the roughness of X under the conditional attribute value vector in B, is the lower deterministic set of X under the conditional attribute value vector in B, is the upper deterministic set of X under the conditional attribute value vector in B; is the adjustment parameter; is the characteristic function, when If it belongs to X, the value is 1, otherwise it is 0; It is objects The weight of is the time adjustment parameter, is the average timestamp of the data in the attribute subset X, is the current time; is an attenuation coefficient;

[0020] Based on the roughness, calculate the dependence of the decision attribute value d on the conditional attribute value vector in B ;

[0021] Define the importance of a single conditional attribute value vector a to the decision attribute value d as the dependency when a is included minus the dependency when a is not included; sort all conditional attribute value vectors a from largest to smallest according to their importance values to obtain the sorted attributes; set an importance threshold θ and traverse the sorted attributes. If the corresponding importance is greater than θ, a is retained in B; if the importance is less than or equal to θ, a is removed from B, and the remaining attributes in B are the core attribute set; use the core attribute set to construct a new decision table, which is recorded as the simplified decision table;

[0022] The decision rules are extracted from the simplified decision table using the MODLEM rule extraction algorithm. The antecedent of the decision rule is the combination of the condition attribute value vectors, and the consequent is the corresponding decision attribute value.

[0023] Computational decision rules The overall confidence ;in, Express full support for the decision rule A collection of objects, Indicates partial support for the decision rule A collection of objects; and Adjust the parameters for confidence;

[0024] Computational decision rules Comprehensive coverage ;in, Represents decision rules The number of conditional attribute values; is a partial coverage function;

[0025] ;in, To cover the adjustment parameters; is the exponential decay rate parameter;

[0026] Based on decision rules The decision rule is obtained by calculating the comprehensive coverage and comprehensive confidence The comprehensive weight of ;in, is the comprehensive weight parameter;

[0027] Then the fusion feature vector ;in, Decision rules vector; the fusion feature vector Mapped to a numerical comprehensive feature vector.

[0028] Furthermore, the method of continuously fusing a plurality of continuous data sets includes:

[0029] Divide several continuous data sets into training sets and output sets; for each continuous data set in the training set, use regression analysis method to build a local model;

[0030] For each local model , calculate its model weight ;in, Adjust parameters for model weights; For local models The number of continuous data in the corresponding continuous data set; is the amount of data in the preliminary fusion dataset; For local models Error evaluation indicators; and is a positive real number parameter;

[0031] Based on the calculated model weights, the local models are fused to obtain a comprehensive model. The output set is input into the comprehensive model, and the output obtained is the fusion feature. ;

[0032] ;in It is a local model The output value for the input sample.

[0033] Furthermore, the method of clustering the motors into groups includes:

[0034] The fused feature data is organized into a matrix or vector form, recorded as a cluster feature vector; the similarity between each pair of motors is calculated; a weighted undirected graph is constructed, where the nodes represent the motors and the edge weights represent the similarity between the motors;

[0035] Preset the density threshold, traverse each node in the weighted undirected graph, calculate the sum of the weights of the edges in its neighborhood as the density of the corresponding node; mark the nodes with density greater than the density threshold and the nodes in their neighborhood as density connected areas;

[0036] Preset similarity threshold and calculate density connected areas and The regional similarity between ;

[0037] ;in, is the distance weight coefficient, is the geographic weight coefficient; For the nodes Hedi nodes The Euclidean distance between For the Each node corresponds to the geographical location coordinates of the motor; For the Each node corresponds to the geographical location coordinates of the motor; for and The great circle distance between is a positive bandwidth parameter;

[0038] Each density-connected region is regarded as a region node, and a region weighted undirected graph is constructed based on the calculated region similarity. The Kruskal minimum spanning tree algorithm is run on the region weighted undirected graph. The Kruskal minimum spanning tree algorithm connects all nodes into a tree without loops. This tree is recorded as a minimum spanning tree. In the minimum spanning tree, region nodes are connected by region edges, and the weight of the region edge is the region similarity. For each region edge, if its weight is greater than or equal to the similarity threshold, the two density-connected regions corresponding to the region edge are classified into the same cluster. If the weight of the region edge is less than the similarity threshold, the two corresponding density-connected regions are assigned to different clusters. Repeat until all region edges are processed. At this point, the minimum spanning tree is a connected graph, so each cluster corresponds to a connected component on the minimum spanning tree. Each connected component corresponds to an initial cluster.

[0039] For each initial cluster, calculate its internal similarity and external similarity; if the internal similarity is higher than the external similarity, retain the initial cluster; if the external similarity is higher than the internal similarity, split the initial cluster into smaller sub-clusters; repeat until the internal similarity of all initial clusters is higher than the external similarity; obtain n clusters, i.e., n motor groups.

[0040] Furthermore, the grid model is established by:

[0041] Based on the physical structure and components of the motors in each motor group, the motors are discretized into an initial two-dimensional or three-dimensional grid structure, where the grid structure contains a number of grid cells, each of which represents a component or part of the motor. A finite set of discrete states is determined for each grid cell, and the discrete states are represented by binary codes.

[0042] According to the aging degree of the motor, the state of the grid unit is defined as {0: normal state, 1: slightly aged state, 2: moderately aged state, 3: severely aged state, 4: fault state}; the state transition rule of the j-th grid unit is defined and expressed as a Boolean function; the state transition rule is: if the grid unit itself or any adjacent grid unit is in a fault state, then it is transferred to the fault state; otherwise, if the grid unit itself or any adjacent grid unit is in a severely aged state, then it is transferred to the severely aged state; otherwise, if the grid unit itself or any adjacent grid unit is in a moderately aged state, then it is transferred to the moderately aged state; otherwise, it is transferred to the slightly aged state;

[0043] Each grid cell is assigned an initial state, which is set based on historical operating data. The refinement merging index is defined as the weighted sum of the discrete value of the current state of the grid cell and the rate of change of the discrete value of the state in the last n1 time steps.

[0044] Set the refinement threshold and the merging threshold; define the grid cells whose refinement merging index is greater than the refinement threshold as refined grids; define the grid cells whose refinement merging index is less than the merging threshold as merged grids;

[0045] For grid cells defined as refined grids, they are divided into smaller sub-grid cells. The initial state of the sub-grid cells inherits the state of the original grid cells, and the state transition rules between the sub-grid cells are interpolated according to the state transition rules of the original grid cells. For adjacent grid cells defined as merged grids, they are merged into a new grid cell, and the state of the new grid cell is obtained by weighted averaging the state of the original grid cells. The state transition rules of the new grid cell are extrapolated according to the original rules. After the grid cells are divided or merged, the state transition rules of each new grid cell are updated accordingly; and the grid structure is updated to obtain the grid model.

[0046] Furthermore, the state update method includes:

[0047] Define the stable state of the simulation; a stable state is defined as the state of all grid cells not changing for n2 consecutive time steps;

[0048] For the t-th time step, traverse all grid cells. For the j-th grid cell, obtain the state of itself and its adjacent grid cells at the t-1-th time step; calculate the state of the j-th grid cell at the t-th time step according to the state transition rule, and update the state of the j-th grid cell. Repeat and record the state of all grid cells at the t-th time step, which is the overall grid state; determine whether a stable state is reached. If so, terminate the simulation. Otherwise, enter the t+1-th time step, repeat the state update, introduce random probability terms in the state transition rule, and randomly perturb the state of the grid cell after each time step; until a stable state is reached;

[0049] The simulation duration is defined, and the state of the grid cells at each time step in the simulation duration is output as an aging curve and a failure rate. The aging curve is represented by the proportion of grid cells in different states in each time step; the failure rate is represented by the proportion of grid cells in a failure state in each time step.

[0050] Furthermore, the method for determining the optimal management path includes:

[0051] The spatial position of the motor is mapped to the pathfinding environment of ants, that is, each motor is regarded as a pathfinding node. For each motor, different states are assigned different weight scores. Based on the assigned weight scores, a weighted average score is calculated at each time step. The weighted average scores of all time steps are averaged as the comprehensive aging indicator of the motor.

[0052] For each motor, the failure rate values at different time steps are weighted averaged to obtain the motor's failure risk index; the comprehensive aging index and failure risk index of each motor are normalized to the interval [0, 1] to obtain the normalized comprehensive aging index and normalized failure risk index;

[0053] Set an initial pheromone concentration baseline value ; For the pathfinding nodes, and their pheromone concentration ;in, For the The normalized comprehensive aging index of the pathfinding nodes, For the Normalized failure risk index of each pathfinding node; and is the relative weight coefficient;

[0054] Create a preset number of ants for each motor group and randomly assign their starting positions in the environment. Initially, all ants have no walking path. Iterate and update the walking paths of all ants until the preset maximum number of iterations is reached. The shortest management path outputted is the optimal management path for the motor group.

[0055] Furthermore, the method of iteratively updating the walking paths of all ants includes:

[0056] For each ant, the next walking direction is selected according to the walking probability based on the pheromone concentration and heuristic information of the pathfinding nodes around the current pathfinding node. The calculation formula of the walking probability is:

[0057] ;in, It is the ants from the The pathfinding node moves to the The walking probability of a pathfinding node; It is The pathfinding node moves to the The pheromone concentration of each pathfinding node; It is The pathfinding node moves to the Heuristic information for pathfinding nodes; 、 and are the relative importance parameters of pheromone concentration, heuristic information, and motor importance, respectively; For the The pathfinding node moves and The overall importance of motors between pathfinding nodes;

[0058] The ants take the next step according to their walking probability until they have built a complete management path. After all ants have built the management paths, the length of each management path is calculated.

[0059] For each path edge in the management path, the pheromone concentration is updated according to the update formula; the update formula is:

[0060] ;in, is the pheromone volatility coefficient; For the The pathfinding node moves to the The pheromone concentration after the pathfinding node is updated; This iteration is The pathfinding node moves to the The new increment of pheromone concentration of the pathfinding node; if ant k starts from the The pathfinding node moves to the pathfinding nodes, then ;in, is a constant, It's an ant The path length; otherwise .

[0061] The technical effects and advantages of the motor control management system of the present invention are as follows:

[0062] By effectively integrating various operating data, status data and position data of the motor, the comprehensive operating status of the motor is comprehensively evaluated, laying a solid data foundation for status diagnosis and fault prediction; the system establishes a refined motor grid model, which can accurately simulate the aging and fault evolution trends of the motor in the future, providing key support for the formulation of forward-looking maintenance strategies, and introduces an intelligent optimization algorithm that can automatically plan the optimal management and maintenance path based on the aging degree and failure risk of the motor, thereby maximizing management efficiency and reducing operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic diagram of a motor control management system according to the present invention;

[0064] Figure 2 Schematic diagram of a motor control management method of the present invention. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] Example 1

[0067] See also Figure 1 As shown, the motor control management system described in this embodiment includes:

[0068] The data acquisition and fusion module is used to collect historical motor comprehensive data, perform preliminary comprehensive alignment on the historical motor comprehensive data, and obtain a preliminary fusion data set;

[0069] The clustering module clusters and groups the motors based on the fused feature data to obtain n motor groups; n is a positive integer greater than 1;

[0070] The model establishment and prediction module is used to establish a grid model of the motor, update the status of the motors in each motor group according to the grid model, and simulate the aging curve and failure rate of the corresponding motors within a preset time;

[0071] The path planning module uses a path planning algorithm to determine the optimal management path for each motor group based on the motor's aging curve and failure rate within a preset time, and executes the optimal management path to inspect the motors in each motor group. The modules are connected via wired and / or wireless means to achieve data transmission between modules.

[0072] Historical motor comprehensive data includes power action data, motor operation data and geographic location data;

[0073] Power action data includes switch status data and protection action data. Switch status data refers to the open and closed status of various switch motors (such as circuit breakers, disconnectors, etc.); protection action data records the action status of various protection devices in the power system (such as overcurrent protection, distance protection, etc.); motor operation data includes motor temperature, motor humidity, motor vibration, motor oil level, motor voltage, motor current and motor power; geographic location data is the geographic coordinate information of the motor; these data come from different data acquisition sources, including monitoring, motor detection, manual recording, etc.

[0074] The historical motor comprehensive data is cleaned to remove invalid data such as outliers and missing values. The data from different sources in the historical motor comprehensive data are formatted in a unified manner to facilitate subsequent processing. Different types of data from the same motor are aligned in time and space to obtain a preliminary fused data set.

[0075] Different types of data may have different sampling frequencies and need to be unified into the same time dimension through methods such as interpolation or resampling. Different types of data from the same motor may come from different sensor locations and need to be mapped into the same spatial coordinate system to reflect the spatial and positional relationships between them.

[0076] The preliminary processing module is used to extract discrete data from the preliminary fusion data set to form a discrete data set. Discrete data, for example, includes switch status data, that is, the on and off status of various switch motors (such as circuit breakers, disconnectors, etc.); protection action data, that is, recording the action status of various protection devices in the motor system (such as overcurrent protection, distance protection, etc.).

[0077] The discrete data set is subjected to discrete fusion processing to obtain a comprehensive feature vector. Specifically, the discrete data set is constructed into a decision table, which consists of a condition attribute set (various discrete data) and a decision attribute set (comprehensive motor state). The condition attribute set includes several condition attribute value vectors, and the decision attribute value set includes several decision attribute values. Each object (motor) in the decision table corresponds to a condition attribute value vector and a decision attribute value. All objects constitute an object set U.

[0078] For each attribute subset (composed of several conditional attribute value vectors) in the conditional attribute set B, the objects in the object set U are grouped according to the value of the attribute subset, and each group forms an equivalence class;

[0079] For any attribute subset X, calculate the upper and lower determination sets of X under the conditional attribute value vector in B; the upper determination set contains all objects that completely belong to X, and the lower determination set contains the equivalence classes in which some objects belong to X.

[0080] Through the cardinality (number of elements) of the upper and lower determination sets, calculate the roughness of X to the decision attribute value d under the conditional attribute value vector in B, that is, the degree to which it cannot be completely determined by B;

[0081] The calculation formula for roughness is:

[0082] ;in, represents the roughness of X under the conditional attribute value vector in B, is the lower deterministic set of X under the conditional attribute value vector in B, is the upper deterministic set of X under the conditional attribute value vector in B; It is an adjustment parameter used to control the influence of the weighted item, and its value range is usually [0, 1]. is the characteristic function, when If it belongs to X, the value is 1, otherwise it is 0; It is objects The weight of It is a time adjustment parameter used to control the influence of time factors. is the average timestamp of the data in the attribute subset X, is the current time; It is an attenuation coefficient used to control the rate of time decay and is usually a positive value.

[0083] Based on the roughness, calculate the dependence of the decision attribute value d on the conditional attribute value vector in B ;

[0084] The importance of a single conditional attribute value vector a to the decision attribute value d is defined as the dependence when a is included minus the dependence when a is not included; the larger the importance value, the greater the influence of attribute a on determining d.

[0085] Sort all conditional attribute value vectors a from large to small according to their importance values to obtain the sorted attributes; set an importance threshold θ and traverse the sorted attributes. If the corresponding importance is greater than θ, a will be retained in B; if the importance is less than or equal to θ, a will be removed from B and regarded as a redundant attribute. The remaining attributes in B are the core attribute set, and the other removed attributes are regarded as redundant; use the core attribute set to construct a new decision table; record it as the simplified decision table.

[0086] The decision rules are extracted from the simplified decision table using the MODLEM rule extraction algorithm. The antecedent of the decision rule is the combination of the conditional attribute value vectors, and the consequent is the corresponding decision attribute value (the motor's comprehensive state).

[0087] Computational decision rules The overall confidence ;in, Express full support for the decision rule A collection of objects, Indicates partial support for the decision rule A collection of objects; and It is a confidence adjustment parameter used to control the impact of some supported objects and rule coverage objects on confidence.

[0088] Computational decision rules Comprehensive coverage ;in, Represents decision rules The number of conditional attribute values; is a partial coverage function;

[0089] ;in, To cover the adjustment parameters, the value range is [0, 1]; is the exponential decay rate parameter, which is a positive real number;

[0090] Based on decision rules The decision rule is obtained by calculating the comprehensive coverage and comprehensive confidence The comprehensive weight of ;in, is a comprehensive weight parameter used to control the relative importance of confidence and coverage in the comprehensive weight.

[0091] Then the fusion feature vector ;in, Decision rules The vector is a 0-1 vector with a dimension equal to the number of all different conditional attribute values;

[0092] The fused feature vector It is mapped into a numerical comprehensive feature vector, and the comprehensive weight of the decision rule is used as the component value of the comprehensive feature vector.

[0093] Through the above steps, the discrete state data of the motor can be effectively fused to obtain the comprehensive characteristic data of the motor, providing data support for subsequent state assessment, fault diagnosis and optimization decision-making.

[0094] The continuous data in the preliminary fusion data set are extracted according to different data acquisition sources to form several continuous data sets; the several continuous data sets are continuously fused to obtain fusion features; specifically, the several continuous data sets are divided into training sets (part of the continuous data set) and output sets (part of the continuous data set).

[0095] For each continuous data set in the training set, a local model is constructed using regression analysis to describe the pattern of data in the continuous data set; for each local model , calculate its model weight ;in, is the model weight adjustment parameter, with a value range of [0, 1], which controls the relative importance of the two weight items; For local models The number of continuous data in the corresponding continuous data set; is the amount of data (number) in the preliminary fusion dataset; For local models Error evaluation indicators, such as root mean square error, R square, etc. and is a positive real number parameter that controls the shape of the function.

[0096] Based on the calculated model weights, the local models are fused to obtain a comprehensive model. The output set is input into the comprehensive model, and the output obtained is the fusion feature. ;

[0097] ;in It is a local model The output value of the input sample (data sample in the output set); the fusion feature The fused feature data are obtained by concatenating the feature vector with the comprehensive feature vector.

[0098] Ways to cluster motors include:

[0099] The fused feature data is organized into a matrix or vector form, recorded as a cluster feature vector; the similarity between each pair of motors (such as Euclidean distance or cosine similarity) is calculated; and a weighted undirected graph is constructed, in which the nodes represent the motors and the edge weights represent the similarity between the motors.

[0100] A density threshold is preset, and each node in the weighted undirected graph is traversed. The sum of the weights of the edges in its neighborhood (obtained by the preset radius) is calculated as the density of the corresponding node; nodes with a density greater than the density threshold and the nodes in their neighborhood are marked as density-connected areas.

[0101] Preset similarity threshold to determine whether two density connected areas are similar enough to calculate the density connected area and The regional similarity between ;

[0102] ;in, is the distance weight coefficient, is the geographical weight coefficient; it controls the contribution of original distance and geographical distance in similarity calculation; For the nodes Hedi nodes The Euclidean distance between them (calculated based on cluster eigenvectors); For the Each node corresponds to the geographical location coordinates of the motor; For the Each node corresponds to the geographical location coordinates of the motor; for and The great circle distance between A positive bandwidth parameter that controls the smoothness of the geographic location constraint. A larger value indicates a looser constraint on the geographic location.

[0103] Each density-connected region is regarded as a region node, and a region weighted undirected graph is constructed based on the calculated region similarity. The Kruskal minimum spanning tree algorithm is run on the region weighted undirected graph. The Kruskal minimum spanning tree algorithm connects all nodes into a tree without loops, and this tree is recorded as the minimum spanning tree. In the minimum spanning tree, region nodes are connected by region edges, and the weight of the region edge is the region similarity. For each region edge, if its weight (regional similarity) is greater than or equal to the similarity threshold, the two density-connected regions corresponding to the region edge are classified into the same cluster. If the weight (regional similarity) of the region edge is less than the similarity threshold, the two corresponding density-connected regions are assigned to different clusters. Repeat until all region edges are processed. At this point, the minimum spanning tree is a connected graph, so each cluster corresponds to a connected component on the minimum spanning tree; each connected component corresponds to an initial cluster.

[0104] A connected component is a subgraph in the minimum spanning tree where any two nodes in the region are connected by a path and are not connected to any other nodes. Therefore, each connected component represents an initial cluster containing densely connected regions with high similarity.

[0105] For each initial cluster, calculate its internal similarity (the average similarity of all regional node pairs within the cluster) and external similarity (the average similarity of all regional node pairs between two clusters); if the internal similarity is higher than the external similarity, retain the initial cluster; if the external similarity is higher than the internal similarity, split the initial cluster into smaller sub-clusters; repeat until the internal similarity of all initial clusters is higher than the external similarity; obtain n clusters, i.e., n motor groups; give each cluster an easy-to-understand label or description to facilitate subsequent management and decision-making.

[0106] The grid model is established in the following ways:

[0107] Based on the physical structure and components of the motors in each motor group, the motors are discretized into an initial two-dimensional or three-dimensional grid structure. The grid structure contains several grid cells, each of which represents a component or component of the motor. A finite set of discrete states is determined for each grid cell, such as normal, slightly aged, moderately aged, severely aged, and faulty. The discrete states are represented by binary codes, such as 0 for normal and 1 for faulty.

[0108] According to the aging degree of the motor, the state of the grid unit is defined as {0: normal state, 1: slightly aged state, 2: moderately aged state, 3: severely aged state, 4: fault state}; the state transition rule of the j-th grid unit is defined and expressed as a Boolean function; the state transition rule is: if the grid unit itself or any adjacent grid unit is in a fault state, then transfer to the fault state; otherwise, if the grid unit itself or any adjacent grid unit is in a severely aged state, then transfer to the severely aged state; otherwise, if the grid unit itself or any adjacent grid unit is in a moderately aged state, then transfer to the moderately aged state; otherwise, transfer to the slightly aged state.

[0109] Use historical operating data and failure cases to verify the accuracy of state transition rules. Based on the verification results, adjust and optimize the rules to better reflect the actual aging and failure process.

[0110] Each grid cell is assigned an initial state (represented by a discrete value). This initial state is set based on historical operating data. Historical operating data refers to various operating parameter data collected during the motor's past operation, reflecting the motor's operating status at different time points. By analyzing this data, the motor's aging degree is assessed and its health status is determined, thereby assigning a reasonable initial state value to the grid cell.

[0111] The refinement merging index is defined as the weighted sum of the discrete value of the current state of the grid cell and the rate of change of the discrete value of the state in the last n1 time steps;

[0112] Set the refinement threshold and merging threshold. Set reasonable refinement threshold and merging threshold according to the simulation accuracy requirements. If high-precision simulation results are required, appropriately lower the refinement threshold and merging threshold to make the grid cells finer and capture more detailed information.

[0113] The grid cells whose refinement and merging index is greater than the refinement threshold are defined as refined grids; the grid cells whose refinement and merging index is less than the merging threshold are defined as merged grids;

[0114] For grid cells defined as refined grids, they are divided into smaller sub-grid cells. The initial state of the sub-grid cell inherits the state of the original grid cell, and the state transition rules between sub-grid cells are interpolated based on the state transition rules of the original grid cell. Specifically, the state transition rules of the original grid cell are regarded as known data points, and the positions of the sub-grid cells are unknown points. The state transition rules between sub-grid cells are estimated through Kriging interpolation.

[0115] For adjacent grid cells defined as merged grids, they are merged into a new grid cell, and the state of the new grid cell is obtained by weighted averaging the states of the original grid cells. The state transition rule of the new grid cell is obtained by extrapolating the original rule. Specifically, the original state transition rule still maintains a linear relationship at the new scale, and only the parameters of the rule need to be linearly scaled. For example, if the original state transition rule is a linear function, the coefficient of the linear function is scaled accordingly based on the scale ratio of the new and old grid cells.

[0116] After the grid units are divided or merged, the state transition rules of each new grid unit are updated accordingly; the grid structure is updated; and a grid model is obtained.

[0117] Ways to update status include:

[0118] Define the stable state of the simulation; a stable state is defined as the state of all grid cells not changing for n2 consecutive time steps;

[0119] For the t-th time step, traverse all grid cells. For the j-th grid cell, obtain the state of itself and its adjacent grid cells at the t-1-th time step; calculate the state of the j-th grid cell at the t-th time step according to the state transition rule (Boolean function), and update the state of the j-th grid cell. Repeat and record the states of all grid cells at the t-th time step, which is the overall grid state; determine whether a stable state is reached. If so, terminate the simulation. Otherwise, enter the t+1-th time step, repeat the state update, introduce random probability terms in the state transition rule, and randomly perturb the state of the grid cell after each time step; until a stable state is reached.

[0120] The simulation duration is defined, and the state of the grid cells at each time step in the simulation duration is output as an aging curve and failure rate. The aging curve is expressed as the proportion of grid cells in different states (normal state, moderate aging state, etc.) in each time step; the failure rate is expressed as the proportion of grid cells in the failure state in each time step.

[0121] By establishing a grid model, the aging and failure evolution process of the motor over a certain period of time can be effectively simulated, providing an important basis for subsequent management optimization and decision-making.

[0122] Methods for determining the optimal management path include:

[0123] The spatial position of the motor is mapped to the pathfinding environment of ants, that is, each motor is regarded as a pathfinding node. The pheromone concentration of the pathfinding node is defined based on the aging curve and failure rate of the motor within a preset time. Specifically, for each motor, different states are assigned different weight scores. Based on the assigned weight scores, a weighted average score is calculated at each time step. The weighted average scores of all time steps are averaged to obtain the comprehensive aging index of the motor.

[0124] For each motor, the failure rate values at different time steps are weighted averaged to obtain the motor's failure risk index; the comprehensive aging index and failure risk index of each motor are normalized to the interval [0, 1] to obtain the normalized comprehensive aging index and normalized failure risk index.

[0125] Set an initial pheromone concentration baseline value ; For the pathfinding nodes, and their pheromone concentration ;in, For the The normalized comprehensive aging index of the pathfinding nodes, For the Normalized failure risk index of each pathfinding node; and is the relative weight coefficient that controls the relative importance of aging and failure risks in the pheromone setting.

[0126] Create a preset number of ants (possible solutions) for each motor group and randomly assign their starting positions in the environment; initially, all ants have no walking path; iteratively update the walking paths of all ants;

[0127] Specifically, for each ant, the next walking direction (i.e., the next pathfinding node to be visited) is selected according to the walking probability based on the pheromone concentration and heuristic information (such as distance) of the pathfinding nodes around the current pathfinding node. The calculation formula of the walking probability is:

[0128] ;in, It is the ants from the The pathfinding node moves to the The walking probability of a pathfinding node; It is The pathfinding node moves to the The pheromone concentration of each pathfinding node; It is The pathfinding node moves to the The heuristic information of a pathfinding node is usually the inverse of the distance between two pathfinding nodes; 、 and are the relative importance parameters of pheromone concentration, heuristic information, and motor importance, respectively; For the The pathfinding node moves and The comprehensive importance of the motors between the pathfinding nodes; specifically, The pathfinding node moves and The pathfinding nodes correspond to the weighted sum of the importance of the motor type and the importance of the load.

[0129] Type importance is an importance score determined based on the motor type; load importance reflects the importance of the load connected to the motor.

[0130] Ants take the next step based on their walking probability until they have built a complete management path (visiting all pathfinding nodes). After all ants have built a management path, the length of each management path (or other fitness function value) is calculated. For each path edge in the management path (the edge between two pathfinding nodes that the ants walked on), the pheromone concentration is updated according to the update formula.

[0131] The update formula is:

[0132] ;in, is the pheromone volatility coefficient ; For the The pathfinding node moves to the The pheromone concentration after the pathfinding node is updated; This iteration is The pathfinding node moves to the The new increment of pheromone concentration of the pathfinding node is contributed by all ants; if ant k The pathfinding node moves to the pathfinding nodes, then ;in, is a constant, It's an ant The path length; otherwise ;

[0133] Repeat until the preset maximum number of iterations is reached, and the management path with the shortest length output at the end is the optimal management path of the motor group; repeat the above process for all motor groups to obtain the optimal management path for each motor group.

[0134] This embodiment effectively integrates various operating data, status data and position data of the motor to comprehensively evaluate the comprehensive operating status of the motor, laying a solid data foundation for status diagnosis and fault prediction; the system establishes a refined motor grid model, which can accurately simulate the aging and fault evolution trends of the motor in the future, providing key support for the formulation of forward-looking maintenance strategies, and introduces an intelligent optimization algorithm that can automatically plan the optimal management and maintenance path based on the aging degree and fault risk of the motor, thereby maximizing management efficiency and reducing operation and maintenance costs.

[0135] Example 2

[0136] See also Figure 2 As shown, for the parts not described in detail in this embodiment, please refer to the description of embodiment 1. A motor control management method is provided, including:

[0137] Step 1: Collect historical motor comprehensive data, perform preliminary comprehensive alignment on the historical motor comprehensive data, and obtain a preliminary fusion data set;

[0138] Step 2: Extract the discrete data from the preliminary fusion data set to form a discrete data set, perform discrete fusion processing on the discrete data set to obtain a comprehensive feature vector; extract the continuous data from the preliminary fusion data set according to the different data acquisition sources to form several continuous data sets; perform continuous fusion processing on the several continuous data sets to obtain fusion features; splice the fusion features and the comprehensive feature vector to obtain fusion feature data;

[0139] Step 3: Cluster the motors based on the fused feature data to obtain n motor groups;

[0140] Step 4: Establish a grid model of the motor, update the status of the motors in each motor group according to the grid model, and simulate the aging curve and failure rate of the corresponding motor within a preset time;

[0141] Step 5: Based on the aging curve and failure rate of the motor within a preset time, a path planning algorithm is used to determine the optimal management path for each motor group, and the optimal management path is executed to inspect the motors in each motor group.

[0142] Example 3

[0143] This embodiment discloses an electronic motor, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the motor control management method provided above is implemented.

[0144] Since the electronic motor described in this embodiment is used to implement a motor control and management method according to the embodiment of this application, those skilled in the art will be able to understand the specific implementation and various variations of the electronic motor according to the embodiment of this application based on the motor control and management method described in the embodiment of this application. Therefore, how the electronic motor implements the method according to the embodiment of this application will not be described in detail here. As long as those skilled in the art can implement the electronic motor used in the motor control and management method according to the embodiment of this application, it falls within the scope of protection of this application.

[0145] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0146] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for users of ordinary skill in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A motor control management system, characterized in that: include: The data acquisition and fusion module is used to collect historical motor comprehensive data, perform preliminary comprehensive alignment on the historical motor comprehensive data, and obtain a preliminary fusion data set; The preliminary processing module is used to extract the discrete data from the preliminary fusion data set to form a discrete data set, and perform discrete fusion processing on the discrete data set to obtain a comprehensive feature vector; Extract the continuous data in the preliminary fusion data set according to the different data acquisition sources to form several continuous data sets; continuously fuse the several continuous data sets to obtain fusion features; splice the fusion features and the comprehensive feature vector to obtain fusion feature data; The clustering module clusters and groups the motors based on the fused feature data to obtain n motor groups; The model establishment and prediction module is used to establish a grid model of the motor, update the status of the motors in each motor group according to the grid model, and simulate the aging curve and failure rate of the corresponding motors within a preset time; The grid model is established in the following manner: According to the physical structure and components of the motor in each motor group, the motor is discretized into an initial two-dimensional or three-dimensional grid structure, and the grid structure contains a number of grid units, each grid unit represents a component unit or component of the motor; a finite set of discrete states is determined for each grid unit, and the state of the grid unit is defined according to the aging degree of the motor, and the state transition rule of the j-th grid unit is defined; the state transition rule is: if the grid unit itself or any adjacent grid unit is in a fault state, it is transferred to the fault state; otherwise, if the grid unit itself or any adjacent grid unit is in a severe aging state, it is transferred to a severe aging state; otherwise, if the grid unit itself or any adjacent grid unit is in a moderate aging state, it is transferred to a moderate aging state; otherwise, it is transferred to a slightly aging state; an initial state is assigned to each grid unit, and the initial state is set based on historical operating data; the refinement merging index is defined as the weighted sum of the discrete value of the current state of the grid unit and the rate of change of the discrete value of the state in the last n1 time steps; Set the refinement threshold and the merging threshold; define the grid cells whose refinement merging index is greater than the refinement threshold as refined grids; define the grid cells whose refinement merging index is less than the merging threshold as merged grids; For a grid cell defined as a refined grid, it is divided into smaller sub-grid cells. The initial state of the sub-grid cell inherits the state of the original grid cell, and the state transition rules between the sub-grid cells are interpolated based on the state transition rules of the original grid cell. For adjacent grid cells defined as a merged grid, they are merged into a new grid cell. The state of the new grid cell is obtained by weighted average based on the state of the original grid cell. The state transition rules of the new grid cell are extrapolated based on the original rules. After the grid cells are divided or merged, the state transition rules of each new grid cell are updated accordingly; and the grid structure is updated to obtain a grid model. The path planning module uses a path planning algorithm to determine the optimal management path for each motor group based on the motor's aging curve and failure rate within a preset time, and executes the optimal management path to inspect the motors in each motor group; each module is connected to each other via wired and / or wireless means.

2. A motor control management system according to claim 1, characterized in that: The historical motor comprehensive data includes power action data, motor operation data and geographic location data; the power action data includes switch status data and protection action data; The motor operation data includes motor temperature, motor humidity, motor vibration, motor oil level, motor voltage, motor current and motor power; the geographic location data is the geographic coordinate information of the motor.

3. A motor control management system according to claim 2, characterized in that: The preliminary comprehensive alignment of the historical motor comprehensive data to obtain a preliminary fused data set includes: cleaning the historical motor comprehensive data, unifying the formats of data from different sources in the historical motor comprehensive data, and temporally aligning and spatially aligning data from different sources of the same motor to obtain a preliminary fused data set.

4. A motor control management system according to claim 3, characterized in that: The method of continuously fusing a plurality of continuous data sets includes: Divide several continuous data sets into training sets and output sets; for each continuous data set in the training set, use regression analysis method to build a local model; For each local model , calculate its model weight ;in, Adjust parameters for model weights; For local models The number of continuous data in the corresponding continuous data set; is the amount of data in the preliminary fusion dataset; For local models Error evaluation indicators; and is a positive real number parameter; Based on the calculated model weights, the local models are fused to obtain a comprehensive model. The output set is input into the comprehensive model, and the output obtained is the fusion feature. ; ;in It is a local model The output value for the input sample.

5. A motor control management system according to claim 4, characterized in that: The method of clustering the motors into groups includes: The fused feature data is organized into a matrix or vector form, recorded as a cluster feature vector; the similarity between each pair of motors is calculated; a weighted undirected graph is constructed, where the nodes represent the motors and the edge weights represent the similarity between the motors; Preset the density threshold, traverse each node in the weighted undirected graph, calculate the sum of the weights of the edges in its neighborhood as the density of the corresponding node; mark the nodes with density greater than the density threshold and the nodes in their neighborhood as density connected areas; Preset similarity threshold and calculate density connected areas and The regional similarity between ; ;in, is the distance weight coefficient, is the geographic weight coefficient; For the nodes Hedi nodes The Euclidean distance between For the Each node corresponds to the geographical location coordinates of the motor; For the Each node corresponds to the geographical location coordinates of the motor; for and The great circle distance between is a positive bandwidth parameter; Each density-connected region is regarded as a region node, and a region weighted undirected graph is constructed based on the calculated region similarity. The Kruskal minimum spanning tree algorithm is run on the region weighted undirected graph. The Kruskal minimum spanning tree algorithm connects all nodes into a tree without loops. This tree is recorded as a minimum spanning tree. In the minimum spanning tree, region nodes are connected by region edges, and the weight of the region edge is the region similarity. For each region edge, if its weight is greater than or equal to the similarity threshold, the two density-connected regions corresponding to the region edge are classified into the same cluster. If the weight of the region edge is less than the similarity threshold, the two corresponding density-connected regions are assigned to different clusters. Repeat until all region edges are processed. At this point, the minimum spanning tree is a connected graph, so each cluster corresponds to a connected component on the minimum spanning tree. Each connected component corresponds to an initial cluster. For each initial cluster, calculate its internal similarity and external similarity; if the internal similarity is higher than the external similarity, retain the initial cluster; if the external similarity is higher than the internal similarity, split the initial cluster into smaller sub-clusters; repeat until the internal similarity of all initial clusters is higher than the external similarity; obtain n clusters, i.e., n motor groups.

6. A motor control management system according to claim 5, characterized in that: The state update method includes: Define the stable state of the simulation; a stable state is defined as the state of all grid cells not changing for n2 consecutive time steps; For the t-th time step, traverse all grid cells. For the j-th grid cell, obtain the state of itself and its adjacent grid cells at the t-1-th time step; calculate the state of the j-th grid cell at the t-th time step according to the state transition rule, and update the state of the j-th grid cell. Repeat and record the state of all grid cells at the t-th time step, which is the overall grid state; determine whether a stable state is reached. If so, terminate the simulation. Otherwise, enter the t+1-th time step, repeat the state update, introduce random probability terms in the state transition rule, and randomly perturb the state of the grid cell after each time step; until a stable state is reached; The simulation duration is defined, and the state of the grid cells at each time step in the simulation duration is output as an aging curve and a failure rate. The aging curve is represented by the proportion of grid cells in different states in each time step; the failure rate is represented by the proportion of grid cells in a failure state in each time step.

7. The motor control management system according to claim 6, characterized in that: The method for determining the optimal management path includes: The spatial position of the motor is mapped to the pathfinding environment of ants, that is, each motor is regarded as a pathfinding node. For each motor, different states are assigned different weight scores. Based on the assigned weight scores, a weighted average score is calculated at each time step. The weighted average scores of all time steps are averaged as the comprehensive aging indicator of the motor. For each motor, the failure rate values at different time steps are weighted averaged to obtain the motor's failure risk index; the comprehensive aging index and failure risk index of each motor are normalized to the interval [0, 1] to obtain the normalized comprehensive aging index and normalized failure risk index; Set an initial pheromone concentration baseline value ; For the pathfinding nodes, and their pheromone concentration ;in, For the The normalized comprehensive aging index of the pathfinding nodes, For the Normalized failure risk index of each pathfinding node; and is the relative weight coefficient; Create a preset number of ants for each motor group and randomly assign their starting positions in the environment. Initially, all ants have no walking path. Iterate and update the walking paths of all ants until the preset maximum number of iterations is reached. The shortest management path outputted is the optimal management path for the motor group.

8. The motor control management system according to claim 7, characterized in that: The method of iteratively updating the walking paths of all ants includes: For each ant, the next walking direction is selected according to the walking probability based on the pheromone concentration and heuristic information of the pathfinding nodes around the current pathfinding node. The calculation formula of the walking probability is: ;in, It is the ants from the The pathfinding node moves to the The walking probability of a pathfinding node; It is The pathfinding node moves to the The pheromone concentration of each pathfinding node; It is The pathfinding node moves to the Heuristic information for pathfinding nodes; 、 and are the relative importance parameters of pheromone concentration, heuristic information, and motor importance, respectively; For the The pathfinding node moves and The overall importance of motors between pathfinding nodes; The ants take the next step according to their walking probability until they have built a complete management path. After all ants have built the management paths, the length of each management path is calculated. For each path edge in the management path, the pheromone concentration is updated according to the update formula; the update formula is: ;in, is the pheromone volatility coefficient; For the The pathfinding node moves to the The pheromone concentration after the pathfinding node is updated; This iteration is The pathfinding node moves to the The new increment of pheromone concentration of the pathfinding node; if ant k starts from the The pathfinding node moves to the pathfinding nodes, then ;in, is a constant, It's an ant The path length; otherwise .

Images