A method and system for intelligent verification of the correlation between end-point load and power source based on multi-source data fusion

Abstract

This application provides a method and system for intelligent verification of the correlation between end-point load power supply based on multi-source data fusion. The method includes: acquiring preprocessed real-time energy consumption data and mechanical vibration signals of electrical equipment sent by edge computing nodes; using a deep learning model to perform multi-source data fusion on the real-time energy consumption data and the mechanical vibration signals to identify the dynamic correlation between the electrical equipment and the power supply; adjusting the abnormal threshold rules of the electrical equipment according to the dynamic correlation; and triggering a corresponding alarm signal when abnormal energy consumption or power disconnection of the electrical equipment is detected according to the abnormal threshold rules of automated logistics equipment, and pushing the alarm signal to the management terminal through an IoT platform. The technical solution provided in this application optimizes the abnormal threshold rules of electrical equipment in real time through multi-source data fusion and deep learning, and triggers intelligent alarms when abnormalities occur, thereby improving power supply safety and energy efficiency management.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to an intelligent verification method and system for the correlation between end-load power sources based on multi-source data fusion. Background Technology

[0002] In industrial IoT scenarios, power supply management for electrical equipment (such as automated logistics equipment) requires real-time sensing of equipment operating status and dynamic optimization of power output to match load changes. Due to the complex and variable operating conditions of electrical equipment (such as start-stop shocks and sudden load changes), existing technologies need to address issues such as real-time fusion analysis of multi-source heterogeneous signals, dynamic correlation modeling, and rapid response to abnormal threshold rules to avoid equipment downtime or energy waste caused by power supply lag or overload.

[0003] To address the aforementioned issues, existing solutions generate power supply commands using a finite state machine by predefined static combination rules of energy consumption thresholds and vibration amplitudes. This approach relies on expert experience and a fixed rule base, enabling signal acquisition and rule matching to a certain extent.

[0004] However, the static combination rules in the existing scheme are difficult to adapt to the nonlinear characteristics of the dynamic relationship between electrical equipment and power supply, resulting in rigid lag and error accumulation problems in the adjustment of abnormal threshold rules. Furthermore, the multi-source signals are simply superimposed at the threshold level, which fails to uncover deep coupling relationships and causes misjudgment of the critical status of the equipment. Summary of the Invention

[0005] This application provides an intelligent verification method for the correlation between end-load power sources based on multi-source data fusion, which solves the problems in the prior art, such as rigid lag and error accumulation caused by the inability of static combination rules to adapt to nonlinear characteristics, as well as the lack of deep coupling relationship and misjudgment of state caused by the shallow logic superposition of multi-source signals.

[0006] Firstly, this application provides an intelligent verification method for the correlation between end-point load power sources based on multi-source data fusion, including:

[0007] Acquire preprocessed real-time energy consumption data and mechanical vibration signals of electrical equipment sent by edge computing nodes;

[0008] A deep learning model is used to perform multi-source data fusion on the real-time energy consumption data and the mechanical vibration signal in order to identify the dynamic correlation between the electrical equipment and the power supply.

[0009] Based on the dynamic correlation, adjust the abnormal threshold rules of the electrical equipment;

[0010] When abnormal energy consumption or power failure of the electrical equipment is detected according to the adjusted abnormal threshold rules, the corresponding alarm signal is triggered and pushed to the management terminal through the Internet of Things platform.

[0011] Optionally, the step of using a deep learning model to perform multi-source data fusion of the real-time energy consumption data and the mechanical vibration signal to identify the dynamic correlation between the electrical equipment and the power supply includes:

[0012] The real-time energy consumption data and mechanical vibration signal of the electrical equipment are divided into continuous energy consumption segments and continuous vibration segments according to time windows, so as to construct the energy consumption feature vector and vibration feature vector corresponding to each time window.

[0013] The energy consumption feature vectors and vibration feature vectors corresponding to all time windows are input into a deep learning model, and a joint feature vector is generated by the multimodal fusion module in the deep learning model using a cross-modal interaction mechanism.

[0014] Based on the joint feature vector, the dynamic weight allocation network in the deep learning model is used to identify the changes in the correlation strength between electrical equipment and power supply, and output a matrix sequence, which is composed of the dynamic correlation coefficient matrix corresponding to all time windows.

[0015] Extract the temporal variation pattern of the dynamic correlation coefficient matrix from the matrix sequence, and establish a dynamic correlation relationship based on the temporal variation pattern.

[0016] Optionally, the step of identifying the change in the correlation strength between electrical equipment and power supply based on the joint feature vector through a dynamic weight allocation network in a deep learning model, and outputting a matrix sequence, includes:

[0017] The joint feature vectors corresponding to all time windows are arranged in chronological order to form a feature sequence. The feature sequence is traversed through a sliding window to calculate the correlation decay factor between adjacent time windows in order to generate a time-series weight vector.

[0018] The joint feature vector is input into the cross-modal correlation analysis module of the dynamic weight allocation network to calculate the cross-entropy fluctuation value between the energy consumption feature vector and the vibration feature vector, and generate the inter-modal weight vector.

[0019] The intermodal weight vector and the temporal weight vector are bidirectionally coupled to output a dynamic weight coefficient vector;

[0020] The joint feature vector is non-linearly scaled based on the dynamic weight coefficient vector to generate a weighted feature vector;

[0021] Cross-channel feature compression is performed on the weighted feature vector to obtain a dynamic correlation coefficient matrix. The dynamic correlation coefficient matrices corresponding to all time windows are aggregated along the time axis to generate a matrix sequence.

[0022] Optionally, adjusting the abnormal threshold rule of the electrical equipment according to the dynamic correlation includes:

[0023] Extract the set of feature parameters between the electrical equipment and the power supply in the dynamic correlation relationship under the state of the electrical equipment. The set of feature parameters includes energy consumption characteristics, vibration characteristics, dynamic weighting coefficients and dynamic correlation coefficients.

[0024] The feature parameter set is input into a pre-constructed anomaly threshold rule adjustment model, and anomaly threshold rule adjustment parameters are generated through the parameter mapping function in the anomaly threshold rule adjustment model.

[0025] The abnormal threshold rules of the electrical equipment are adjusted according to the abnormal threshold rules.

[0026] Optionally, when abnormal energy consumption or power failure of the electrical equipment is detected according to the adjusted abnormal threshold rules, triggering a corresponding alarm signal and pushing it to the management terminal through the IoT platform includes:

[0027] An energy consumption fluctuation curve is generated based on the energy consumption feature vector, and a vibration spectrum segment is generated based on the vibration feature vector.

[0028] Extract the mean offset rate of the energy consumption fluctuation curve and the main frequency band energy proportion of the vibration spectrum segment;

[0029] Based on the adjusted abnormal threshold rules, the mean deviation rate and the main frequency band energy ratio are judged to determine abnormal energy consumption events or power supply failure events.

[0030] For the aforementioned energy consumption anomaly or power outage event, generate corresponding alarm signals according to the anomaly level;

[0031] Through the data channel encapsulation protocol of the IoT platform, the alarm type, electrical equipment number, abnormal timestamp and associated data fragments corresponding to the alarm signal are packaged into a lightweight transmission message and pushed to the management terminal.

[0032] Optionally, the step of inputting the energy consumption feature vector and the vibration feature vector corresponding to all time windows into a deep learning model, and generating a joint feature vector through a multimodal fusion module in the deep learning model using a cross-modal interaction mechanism, includes:

[0033] The energy consumption feature vectors and vibration feature vectors of all time windows are divided into start-up and shutdown phases, steady-state operation phases and load change phases according to the equipment operation stage.

[0034] Based on the power change slope of the energy consumption characteristic vector and the main frequency band energy distribution of the vibration characteristic vector in each stage, the cross-modal dynamic interaction weight is calculated through the cross-modal interaction mechanism.

[0035] Based on the cross-modal dynamic interactive weights, vibration feature nonlinear enhancement and energy consumption feature superposition are performed in the start-stop phase to generate the start-stop phase vector; dual-modal projection residual vector is generated in the steady-state operation phase; and energy consumption piecewise transformation and vibration window convolution are performed in the load mutation phase to generate the load mutation vector.

[0036] The start-stop phase vector, the dual-modal projection residual vector, and the load mutation vector are concatenated in time window order to generate a joint feature vector.

[0037] Optionally, the step of calculating the cross-modal dynamic interaction weights based on the power change slope of the energy consumption characteristic vector and the dominant frequency band energy distribution of the vibration characteristic vector within each stage through a cross-modal interaction mechanism includes:

[0038] The peak value of the power change slope of the energy consumption characteristic vector during the start-stop phase and the peak value of the main frequency band energy distribution of the vibration characteristic vector are input into the vibration dominance factor calculation module in the multi-modal fusion module to generate the vibration dominance factor.

[0039] The stability coefficient of the power change slope of the energy consumption characteristic vector during the steady-state operation phase and the dispersion of the main frequency band energy distribution of the vibration characteristic vector are input into the steady-state balance factor calculation module in the multi-mode fusion module to generate the steady-state balance factor.

[0040] The deviation amplitude of the main frequency band energy distribution of the vibration characteristic vector during the load change phase is input into the energy consumption weight gain calculation module in the multimodal fusion module to generate the energy consumption weight gain.

[0041] The vibration dominant factor, the steady-state equilibrium factor, and the energy consumption weight gain are dynamically coupled across stages to generate cross-modal dynamic interactive weights.

[0042] Secondly, this application provides an intelligent verification system for the correlation between end-point load power sources based on multi-source data fusion, including:

[0043] The acquisition module is used to acquire preprocessed real-time energy consumption data and mechanical vibration signals of the electrical equipment sent by the edge computing node;

[0044] The identification module is used to perform multi-source data fusion of the real-time energy consumption data and the mechanical vibration signal using a deep learning model, so as to identify the dynamic correlation between the electrical equipment and the power supply.

[0045] An adjustment module is used to adjust the abnormal threshold rules of the electrical equipment according to the dynamic correlation.

[0046] The detection module is used to trigger a corresponding alarm signal when the power consumption of the electrical equipment is abnormal or the power supply is disconnected according to the adjusted abnormal threshold rules, and push the alarm signal to the management terminal through the Internet of Things platform.

[0047] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are to be invoked and executed by the processing component to implement an intelligent verification method for end-load power supply correlation based on multi-source data fusion as described in any of the first aspects.

[0048] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements an intelligent verification method for the correlation between end-load power sources based on multi-source data fusion as described in any of the first aspects.

[0049] This application provides an intelligent verification method for the correlation between end-point load power supply based on multi-source data fusion. The method includes: acquiring preprocessed real-time energy consumption data and mechanical vibration signals of electrical equipment sent by edge computing nodes; using a deep learning model to perform multi-source data fusion on the real-time energy consumption data and the mechanical vibration signals to identify the dynamic correlation between the electrical equipment and the power supply; adjusting the abnormal threshold rules of the electrical equipment according to the dynamic correlation; and triggering a corresponding alarm signal when abnormal energy consumption or power supply disconnection of the electrical equipment is detected according to the adjusted abnormal threshold rules, and pushing it to the management terminal through the Internet of Things platform.

[0050] This application reduces the computational load on central nodes by preprocessing and synchronously transmitting multi-source heterogeneous data through edge computing nodes. Simultaneously, it leverages the complementary characteristics of multi-source data in the time-frequency domain to improve the spatiotemporal resolution of modeling the relationship between electrical equipment and power supply. Based on dynamic correlation features extracted through deep learning fusion, it automatically corrects the sensitivity bias of traditional fixed threshold rules to scenarios such as equipment aging and load fluctuations, achieving dynamic alignment between anomaly detection conditions and the actual operating status of equipment. Through a dual-dimensional collaborative judgment mechanism of energy consumption anomalies and power disconnection, it eliminates alarm noise caused by false alarms from single sensors. Combined with the multi-protocol adaptation capabilities of the IoT platform, it ensures reliable delivery and traceability of alarm signals in complex network environments. Based on the nonlinear correlation characteristics between vibration signals and energy consumption data, it can identify the gradual increase in power contact impedance caused by equipment mechanical wear, providing early warning of implicit coupling faults before traditional current overload protection is triggered. Furthermore, this application divides real-time energy consumption data and mechanical vibration signals of electrical equipment into continuous segments according to time windows, constructs energy consumption feature vectors and vibration feature vectors respectively, and inputs them into the multi-modal fusion module of a deep learning model to generate a joint feature vector. Based on sliding window analysis, the correlation attenuation factor and cross-modal cross-entropy fluctuation value of adjacent time windows are analyzed, and a dynamic weight coefficient vector is generated through bidirectional coupling. After nonlinear scaling and cross-channel compression of the joint feature vector, a dynamic correlation coefficient matrix sequence is output. Finally, the temporal variation law is extracted from the matrix sequence to establish the dynamic correlation relationship between the equipment and the power supply. Through time window segmentation and cross-modal interactive fusion of multi-modal data, the fine-grained perception capability of the correlation strength change between electrical equipment and the power supply is improved. Combined with the dynamic coupling mechanism of temporal weight and inter-modal weight, the nonlinear influence of equipment operating state drift on the correlation relationship is adaptively captured. By using cross-channel feature compression and matrix sequence aggregation, single-modal noise interference can be suppressed and the temporal law of correlation strength evolution can be extracted, realizing the early and accurate identification of implicit coupling faults of equipment mechanical wear and power supply contact impedance changes.

[0051] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating an intelligent verification method for the correlation between end-load power sources based on multi-source data fusion, provided in an embodiment of this application;

[0054] Figure 2 A schematic diagram of the structure of an intelligent verification system for the correlation between end-load power sources based on multi-source data fusion, provided in an embodiment of this application;

[0055] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0056] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0057] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 11, 12, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0058] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] Figure 1 A flowchart for intelligent verification of end-point load power supply correlation based on multi-source data fusion is provided for embodiments of this application, such as... Figure 1 As shown, the method includes:

[0060] S11. Obtain the preprocessed real-time energy consumption data and mechanical vibration signals of the power-consuming equipment sent by the edge computing node.

[0061] The edge computing nodes possess preprocessing capabilities and periodically transmit preprocessed real-time energy consumption data and mechanical vibration signals from electrical equipment to the cloud monitoring platform via the network. The preprocessed real-time energy consumption data refers to standardized power time-series data generated after filtering, noise reduction, and piecewise integration of the original current and voltage signals by the edge computing nodes. This data includes active power, reactive power, and power factor during equipment operation. The preprocessed real-time energy consumption data may also include RMS current and instantaneous power fluctuation rate. The mechanical vibration signal refers to the spectral characteristic sequence generated after time-frequency domain decomposition of the equipment casing vibration acceleration signal collected by piezoelectric sensors by the edge nodes. This sequence includes fundamental frequency components, harmonic components, and resonant peak energy distribution. Analyzing the mechanical vibration signal yields vibration data (such as spectral energy and vibration intensity).

[0062] For example, during the operation of a CNC machine tool in a metal processing workshop, an edge computing node collects real-time motor current signals via a current sensor at a sampling rate of 1kHz, sets a window length of 50ms, and uses a moving average filter to eliminate high-frequency noise, generating pre-processed energy consumption data (e.g., including an RMS current of 25A and an instantaneous power fluctuation rate of 0.3). Simultaneously, a triaxial vibration sensor collects mechanical vibration signals at a sampling rate of 5kHz. The mechanical vibration signals are transformed to extract the spectral energy distribution (e.g., 92% of the energy at the dominant frequency of 300Hz) and the time-domain root mean square value (e.g., 4.2 m / s²), completing the pre-processing operation of the mechanical vibration signals. The edge node encapsulates the timestamp-aligned energy consumption data (RMS current and instantaneous power fluctuation rate) and the mechanical vibration signals into a JavaScript Object Notation (JSON) format data packet via a message queue or telemetry transmission protocol, and pushes it to the cloud monitoring platform every 500ms for subsequent multimodal fusion and anomaly detection analysis.

[0063] S12. Use a deep learning model to perform multi-source data fusion on real-time energy consumption data and mechanical vibration signals in order to identify the dynamic correlation between electrical equipment and power supply.

[0064] Multi-source data fusion refers to the nonlinear correlation modeling of the time-domain trend characteristics of real-time energy consumption data with the frequency-domain structural characteristics of mechanical vibration signals through a cross-modal feature interaction mechanism. Dynamic correlation refers to the time-varying coupling strength between equipment load changes and power supply contact impedance fluctuations; this dynamic correlation can be quantified as a dynamic correlation coefficient matrix.

[0065] For example, during high-precision milling on a CNC machine tool, real-time energy consumption data and mechanical vibration signals from the spindle motor are collected. Using a multimodal fusion module of a deep learning model, the energy consumption feature vector and vibration feature vector are aligned according to a time window and input into a cross-modal attention network. The energy consumption feature is used as the query vector, and the vibration feature as the key vector. The attention weight (e.g., 0.75) of the mechanical vibration signal during the current peak period (e.g., at the 15th second) is calculated, generating a fused 12-dimensional joint feature vector. A dynamic weight allocation network is used to analyze the joint features. Under normal operating conditions with a baseline of 0.85, the correlation strength coefficient between the power supply and the equipment in the current time window is output as 0.62. A linear decay trend of 0.1 decrease per second is detected as the correlation strength decreases with sudden changes in spindle load, indicating a "break in the dynamic correlation between the power supply and the equipment," triggering a real-time alarm and adjusting the power supply output to suppress the deterioration of machining accuracy.

[0066] S13. Adjust the abnormal threshold rules for electrical equipment based on dynamic correlation. The abnormal threshold rules refer to the critical conditions for determining abnormal energy consumption or power supply failure of electrical equipment, including power fluctuation amplitude thresholds, vibration energy mutation thresholds, and the correlation strength thresholds between the two.

[0067] For example, during the finishing stage of a CNC machine tool, the system detects that the real-time correlation strength between the equipment and the power supply is 0.68 (the normal benchmark is 0.85) based on a dynamic correlation model, and determines that it is currently in a low-load sensitive state. According to the correlation strength decrease ratio (20%), it is calculated linearly: 30A × (1-20%) = 24A. Therefore, the current abnormal threshold is dynamically adjusted from the default 30A to 24A, and the vibration abnormal threshold is reduced from 5.0 m / s² to 4.0 m / s². When the real-time monitoring shows that the instantaneous value of the spindle motor current is 25A, which exceeds the adjusted threshold of 24A and the vibration intensity is 4.3 m / s², an "overload abnormality" alarm is immediately triggered, which identifies the potential power coupling failure risk 8 seconds earlier than the fixed threshold rule, thus avoiding the surface accuracy of the machined workpiece being out of tolerance.

[0068] S14. When abnormal energy consumption or power failure of electrical equipment is detected according to the adjusted abnormal threshold rules, a corresponding alarm signal is triggered and pushed to the management terminal through the IoT platform. The alarm signal is a structured data packet containing anomaly type code, correlation strength deviation, and suggested handling measures.

[0069] For example, during continuous machining on a CNC machine tool, the identification system adjusts the current abnormality threshold to 24A and the vibration threshold to 4.0 m / s² based on dynamic correlation. It then monitors in real time that the instantaneous value of the spindle motor current reaches 25A and the X-axis vibration intensity is 4.3 m / s², triggering an "overload abnormality" alarm signal. At the same time, it detects that the power supply voltage drops sharply to 360V, which is lower than the predicted value of 380V by the correlation model and lasts for 3 seconds, thus determining it as a power failure event. The alarm information (including equipment identification, abnormality type, real-time value, and timestamp) is pushed to the management terminal in JSON format through an IoT platform (such as Alibaba Cloud IoT). The large screen in the workshop control center can simultaneously display a red warning pop-up and activate the audible and visual alarm. Furthermore, maintenance personnel can also view the alarm details through the terminal and immediately perform an emergency shutdown to prevent the spindle motor from overheating and being damaged.

[0070] Here is a specific example:

[0071] In the coupled fault monitoring of a CNC machine tool machining workshop, the implementation process of a certain CNC machine tool workshop solution is as follows: edge nodes collect the machine tool spindle motor current at a sampling rate of 2kHz and the spindle bearing vibration signal at a sampling rate of 12kHz, generating a data packet every 500ms. The deep learning model detected that the energy consumption feature vector of a certain machine tool increased by 15% during the cutting stage, but the energy of the 6.4kHz resonance peak in the vibration spectrum decreased by 40%. The model output dynamic correlation coefficient matrix showed that the correlation strength between the two dropped to 30% of the baseline. The identification system determined it as "increased power contact impedance" and dynamically relaxed the power fluctuation threshold from ±10% to ±15% to avoid false alarms. When the correlation strength is continuously lower than 20% of the baseline for more than 5 minutes, a "power contact abnormality" alarm is triggered and pushed to the operation and maintenance terminal through the 5G network, prompting to check the motor wiring terminals.

[0072] By executing S11~S14, this embodiment of the application achieves fine-grained modeling of the dynamic coupling relationship between electrical equipment and power supply through heterogeneous data preprocessing and real-time feature extraction of edge computing nodes, combined with cross-domain correlation analysis of multimodal deep learning models; the adaptive threshold adjustment mechanism based on the dynamic correlation coefficient matrix effectively suppresses false alarms caused by load fluctuations and environmental noise; and through a dual-dimensional alarm judgment and protocol adaptive push strategy, it achieves early and accurate warning of implicit coupling faults and improves the efficiency of handling response in complex industrial scenarios.

[0073] In one possible embodiment, S12, using a deep learning model to perform multi-source data fusion on real-time energy consumption data and mechanical vibration signals to identify the dynamic correlation between electrical equipment and power supply, including:

[0074] Step 121: Divide the real-time energy consumption data and mechanical vibration signal of the electrical equipment into continuous energy consumption segments and continuous vibration segments according to time windows, so as to construct the energy consumption feature vector and vibration feature vector corresponding to each time window.

[0075] In this context, a time window is a data slice of fixed length (e.g., 10 seconds) used to divide continuous time-series signals. An energy consumption continuous segment refers to the energy consumption time-series data of a device within a specific time window, such as current or power value sequences. A vibration continuous segment is the vibration acceleration time-series data within the same time window, such as vibration waveforms along the X, Y, or Z axes. An energy consumption feature vector is a set of statistical features extracted from an energy consumption segment, such as mean, variance, and extreme values. A vibration feature vector is a set of frequency domain or time domain features extracted from a vibration segment, such as spectral energy and waveform factors.

[0076] For example, during continuous machining on a CNC machine tool, current data of the spindle motor is collected in real time at a sampling rate of 1kHz and vibration signals are collected in real time at a sampling rate of 5kHz. Data segments are divided with a time window of 5 seconds and a sliding step of 2.5 seconds. For each current window, the root mean square value (e.g., 18A), variance (e.g., 0.2), and peak factor (e.g., 3.5) are calculated to construct a 3-dimensional energy consumption feature vector. Fast Fourier transform analysis is performed on the vibration window to extract the energy proportion (e.g., 85%), time domain kurtosis (e.g., 4.2), and waveform factor (e.g., 5.0) of the 200Hz frequency band, generating a 3-dimensional vibration feature vector. Timestamp alignment ensures a one-to-one correspondence between energy consumption features and vibration features, ultimately forming a 6-dimensional joint feature vector, which is transmitted to the cloud monitoring platform via the Transmission Control Protocol (TCP) for subsequent multimodal fusion.

[0077] Step 122: Input the energy consumption feature vectors and vibration feature vectors corresponding to all time windows into the deep learning model. The multimodal fusion module in the deep learning model uses a cross-modal interaction mechanism to generate a joint feature vector. This cross-modal interaction mechanism achieves dynamic interaction of information from different modalities through feature cross-attention. The joint feature vector is a higher-order feature vector after fusion, containing the semantic association between energy consumption and vibration; therefore, the joint feature vector includes both energy consumption and vibration features.

[0078] For example, in real-time monitoring of CNC machine tools, the generated energy consumption feature vector (root mean square current 20A, variance 0.15, peak factor 3.8) and vibration feature vector (energy proportion of 88% in the 250Hz frequency band, temporal kurtosis 4.5, waveform factor 5.2) are aligned by time window and input into a deep learning model. Through the cross-modal attention mechanism of the multimodal fusion module, the energy consumption feature is used as the query vector and the vibration feature is used as the key-value vector. During the current peak period (e.g., the 6th second), the attention weight of the vibration feature (e.g., 0.85) is calculated to generate a fused 8-dimensional joint feature vector (e.g., dimensions 1-3 are energy consumption features, 4-6 are vibration features, and 7-8 are cross-modal attention features). The joint feature vector is compressed by a fully connected layer and output to a dynamic weight allocation network for subsequent correlation strength modeling, realizing deep collaborative perception of energy consumption and vibration signals.

[0079] Step 123: Based on the joint feature vector, identify the changes in the correlation strength between electrical equipment and power supply through the dynamic weight allocation network in the deep learning model, and output a matrix sequence. The matrix sequence consists of dynamic correlation coefficient matrices corresponding to all time windows. The dynamic weight allocation network is a neural network structure that dynamically adjusts the weights of each feature according to the input features. The dynamic correlation coefficient matrix represents the correlation strength between electrical equipment and power supply in different time windows; for example, rows represent time, and columns represent correlation dimensions.

[0080] For example, based on the generated 8-dimensional joint feature vector (including features such as the root mean square value of energy consumption of 20A and the energy proportion of vibration at 250Hz of 88%), the dynamic weight allocation network analyzes the joint features of each time window. Specifically, the energy consumption feature weight is calculated to be 0.6 and the vibration feature weight is 0.4 at the current moment through a gating mechanism. After weighted fusion, the correlation strength coefficient between the electrical equipment and the power supply is output (such as the current-voltage correlation coefficient of 0.75 and the vibration-voltage correlation coefficient of 0.52). A dynamic correlation coefficient matrix is ​​generated every 5 seconds. The rows in the matrix represent the time window, and the columns represent the current / vibration dimension. During the high-speed machining stage of the spindle (10-15 seconds), the current-voltage correlation coefficient decreases from 0.75 to 0.48, and the vibration-voltage correlation coefficient decreases from 0.52 to 0.31, forming a matrix sequence for subsequent time series pattern analysis.

[0081] Step 124: Extract the temporal variation pattern of the dynamic correlation coefficient matrix from the matrix sequence, and establish a dynamic correlation relationship based on the temporal variation pattern. The temporal variation pattern refers to the evolution pattern of the correlation strength over time, such as linear decay or periodic fluctuations. The dynamic correlation relationship is a mathematical model (e.g., a piecewise function) that quantifies the correlation between the device and the power supply state.

[0082] For example, during a continuous 24-hour machining process, based on the output dynamic correlation coefficient matrix sequence (such as the current and voltage coefficient matrix, vibration and voltage coefficient matrix), a moving average filter (window length 10) was applied to the correlation coefficients for each time window (5-second interval). It was found that the current-voltage coefficient linearly decreased from an initial 0.75 to 0.45, while a periodic fluctuation of ±0.12 was detected every 45 minutes. A sudden drop in the coefficient to 0.3 was located during the spindle tool change period in the 8th hour using a mutation point detection algorithm. A dynamic correlation model was then established based on equipment process parameters (such as tool wear coefficient and power curve).

[0083] ;

[0084] R(t) represents the quantified dynamic correlation between multiple parameters of the electrical equipment over time t, such as the real-time correlation between current and voltage, vibration and voltage. The basic attenuation term is 0.75-0.004t, indicating a linear decrease in correlation strength over time, reflecting long-term trends such as equipment aging and load accumulation. The initial value of 0.75 represents the correlation coefficient under healthy conditions, with an attenuation rate of 0.004 / second. The dynamic correlation is captured by a mathematical model to observe the evolution of correlation strength over time (e.g., long-term attenuation, periodic fluctuations, sudden anomalies), used to assess the equipment's operating status. When the current and voltage coefficients deviate from the model's predicted values ​​by more than 20% in real time (e.g., actual value 0.38 vs. predicted value 0.48), it is determined to be an abnormality in power supply dynamic coupling, triggering an early warning 15 minutes earlier than the traditional fixed threshold rule, guiding operators to adjust the power supply output to match the processing load.

[0085] Here is a specific example:

[0086] In the machining process of a CNC machine tool, current data of the spindle motor is collected in real time at a sampling rate of 1kHz and vibration signals are collected in real time at a sampling rate of 5kHz. First, the two types of data are divided into continuous segments with a 10-second time window (sliding step of 7 seconds). The root mean square value of the current (25A), power fluctuation rate (0.3), and peak current percentage (12%) are extracted to construct a 3D energy consumption feature vector. Simultaneously, the energy percentage (92%) and waveform factor (5.1) of the 300Hz frequency band are calculated based on the vibration signal to form a 4D vibration feature vector. Subsequently, the multimodal fusion module of a deep learning model is used to... Energy consumption features are used as query vectors, and vibration features are used as key vectors for cross-modal attention calculation. During the current peak period (8th second), the weight of vibration features is increased to 0.8, generating a 12-dimensional joint feature vector. A dynamic weight allocation network (such as a gated recurrent unit) determines that the vibration features account for 65% of the correlation strength based on the joint feature vector, and outputs a correlation matrix. This correlation matrix shows that the correlation strength between the electrical equipment and the power supply decreases from an initial 0.9 to 0.4 after a load surge. Finally, through moving average filtering and abrupt change detection, it is found that the correlation strength decreases by 10% periodically every 30 minutes, and a model is fitted. When the actual correlation strength is less than 15% of the predicted value, an "abnormal power load" alarm is immediately triggered and pushed to the management terminal to achieve real-time blocking of processing abnormalities.

[0087] By executing steps 121-124, this embodiment of the application achieves accurate modeling and real-time monitoring of the dynamic correlation between electrical equipment and power supply: energy consumption and vibration data are segmented and multi-dimensional features are extracted to ensure the timeliness and richness of data representation; the implicit correlation between energy consumption and vibration is jointly analyzed through a multi-modal fusion mechanism to enhance the collaborative sensing capability of abnormal signals; the correlation strength between equipment and power supply is quantified using a dynamic weight allocation network to generate a time-series correlation matrix to capture dynamic change trends; a dynamic model is established based on the time-series patterns of the correlation matrix to adapt to complex operating conditions such as equipment aging and load fluctuations, ultimately improving the sensitivity and reliability of anomaly detection, reducing false alarms and missed alarms, and providing data-driven basis for power supply stability optimization.

[0088] In one possible embodiment, step 123, based on the joint feature vector, identifies the changes in the correlation strength between the electrical equipment and the power supply through a dynamic weight allocation network in a deep learning model, and outputs a matrix sequence, including:

[0089] Step a1: Arrange the joint feature vectors corresponding to all time windows in chronological order to form a feature sequence. Traverse the feature sequence through a sliding window and calculate the correlation decay factor between adjacent time windows to generate a time-series weight vector.

[0090] Here, the feature sequence is a set of joint feature vectors arranged in chronological order. The correlation decay factor is an indicator that quantifies the feature similarity between adjacent time windows; a smaller value indicates faster correlation decay. The temporal weight vector is a vector that assigns weights to each time window; a higher weight indicates a greater influence of that window on the current state. The correlation decay factor between adjacent time windows is calculated. The following formula can be used: ;in It is the joint feature vector of the t-th time window; It is the joint feature vector of the (t+1)th time window.

[0091] For example, in the real-time monitoring of CNC machine tools, the joint feature vectors of three consecutive time windows (each window is 5 seconds) are [0.8, 1.2, 0.5] for window 1, [0.7, 1.1, 0.6] for window 2, and [0.6, 1.0, 0.7] for window 3, arranged in chronological order as a feature sequence; the correlation attenuation factor between adjacent windows is calculated by traversing through a sliding window: the cosine similarity between window 1 and window 2 is 0.99, and the attenuation factor... =1-0.995=0.005, time series weight =e -0.005 ≈0.995; the cosine similarity between window 2 and window 3 is also 0.995, weight The final time-series weight vector [0.995, 0.995] is generated, which indicates that the features of adjacent windows are highly continuous and the weight is close to 1, reflecting that the device state is stable and without fluctuations.

[0092] Step a2: Input the joint feature vector into the cross-modal correlation analysis module of the dynamic weight allocation network to calculate the cross-entropy fluctuation value between the energy consumption feature vector and the vibration feature vector, and generate the inter-modal weight vector.

[0093] The inter-modal weight vector is a vector that assigns weights to different modes (energy consumption / vibration); a higher weight indicates a greater contribution of that mode to the correlation strength. The cross-entropy fluctuation value is an indicator of the difference in the distribution characteristics of energy consumption and vibration; a higher value indicates a lower correlation between modes. The cross-entropy calculation formula is:

[0094] ,in, The probability distribution representing energy consumption characteristics. The probability distribution representing the vibration characteristics. This represents the cross-entropy fluctuation value, where n represents the total number of time windows after discretization of continuous feature values. i For the first i A time window.

[0095] For example, within a certain time window, the energy consumption feature vector of a CNC machine tool is [root mean square current 20A, variance 0.15, peak factor 3.8], and the vibration feature vector is [energy proportion of 250Hz band 88%, kurtosis 4.5, waveform factor 5.2]. The energy consumption feature is converted into a probability distribution [0.999, 0.0003, 0.0007], and the vibration feature is converted into [0.999, 0.0001, 0.0009]. The cross-entropy fluctuation values ​​of the two distributions are then calculated. According to the formula The generated intermodal weight vector [0.999, 0.001] indicates that the vibration characteristics within the current window contribute as much as 99.9% to the dynamic correlation strength, while the energy consumption characteristics only account for 0.1%, reflecting that the equipment anomaly is mainly dominated by the vibration signal.

[0096] Step a3: Bidirectionally couple the intermodal weight vector with the temporal weight vector to output a dynamic weight coefficient vector; where bidirectional coupling is the operation of fusing temporal weights and modal weights to ensure that both influence the final weight allocation.

[0097] For example, within a certain time window, the inter-modal weight vector is This indicates that vibration characteristics dominate, and the time-series weight vector is... This reflects the high continuity of features between adjacent windows; weight alignment is achieved through bidirectional coupling. For example, the modal weight of 0.53 and the temporal weight of 0.86 in the same time window are matched and weighted and summed according to a preset ratio (temporal weight accounts for 60%, modal weight accounts for 40%) to calculate the dynamic weight coefficient of the first window. :

[0098] Assuming the dynamic weight coefficient vector of the current window sequence is [0.728, 0.5, 0.7], calculate the L2 norm: Normalization based on the L2 norm yields a dynamic weight coefficient vector [0.65, 0.45, 0.62]. In this dynamic weight coefficient vector, the current window weight of 0.65 is higher than that of other windows (0.45, 0.62), indicating that the vibration characteristics and high temporal continuity of this period jointly dominate the modeling of the correlation strength between equipment and power supply, providing a dynamic weight basis for subsequent feature scaling.

[0099] Step a4: Perform nonlinear scaling on the joint feature vector based on the dynamic weight coefficient vector to generate a weighted feature vector; where nonlinear scaling is based on nonlinear adjustment of the feature vector based on the weights.

[0100] Exemplarily, in a certain time window, the dynamic weight coefficient vector is [0.65, 0.45, 0.62], and the joint feature vector is [0.8 (cross-modal feature 1), 1.2 (cross-modal feature 2), 0.5 (cross-modal feature 3)]; the dynamic weight coefficients are non-linearly mapped through a function to obtain the scaling coefficients. , and then each dimension of the joint feature vector is multiplied by the scaling coefficient to generate a weighted feature vector: [0.8×0.656≈0.525, 1.2×0.656≈0.787, 0.5×0.656≈0.328]. After weighting, the overall eigenvalue is compressed to 65.6% of the original value, retaining the relative ratio between features and suppressing low-weight feature noise, providing a redundant-reduced input for subsequent cross-channel compression.

[0101] Step a5: Perform cross-channel feature compression on the weighted feature vector to obtain a dynamic correlation coefficient matrix, and aggregate the dynamic correlation coefficient matrices corresponding to all time windows along the time axis dimension to generate a matrix sequence.

[0102] Among them, cross-channel feature compression is to reduce the high-dimensional weighted features to low-dimensional weighted features. The dynamic correlation coefficient matrix represents the multi-dimensional correlation strength between the electrical equipment and the power supply in each time window. The compression process is to map the N-dimensional weighted features to M-dimensional weighted features (M < N) through a fully connected layer. Calculate the attention weights between the M-dimensional weighted features to generate a dynamic correlation coefficient matrix. Perform an outer product or covariance operation on the low-dimensional weighted features to generate an M×M matrix. Among them, the M×M dynamic correlation coefficient matrix, for example:

[0103] ;

[0104] Exemplarily, in the continuous machining of a numerically controlled machine tool, the weighted feature vector of a certain time window is [0.525 (cross-modal feature 1), 0.787 (cross-modal feature 2), 0.328 (cross-modal feature 3)]. It is compressed into 2-dimensional dynamic correlation coefficients (current-voltage correlation coefficient 0.62, vibration-voltage correlation coefficient 0.48) through a fully connected layer to form the correlation coefficient matrix of the current window 0.62, 0.48; aggregate the correlation coefficient matrices of 5 consecutive windows along the time axis (the current-voltage coefficients of windows 1 to 5 are 0.62, 0.58, 0.55, 0.52, 0.48 respectively, and the vibration-voltage coefficients are 0.48, 0.45, 0.42, 0.40, 0.38) to generate a matrix sequence [0.62, 0.48], [0.58, 0.4" +

[0105] Here is a specific example:

[0106] For the power supply correlation analysis of CNC machine tools, the joint feature sequence contains 100 time windows (5 seconds / window). The cosine similarity between adjacent windows is calculated (e.g., the similarity between windows 1 and 2 is 0.85, the attenuation factor is 0.15, and the time weight is...). Energy consumption characteristic distribution of window 1 Vibration characteristic distribution Cross-entropy H=0.89, modal weights The time-series weight of 0.86 and the modal weight of 0.53 are coupled to obtain... and will Normalized to 0.75. Scaling factor. The joint feature vector (e.g., [0.8, 1.2]) was scaled to [0.54, 0.82]. After compression, the current-voltage correlation strength was 0.62 and the vibration-voltage correlation strength was 0.48, which were added to the matrix sequence. During continuous monitoring, the matrix sequence showed that the current-voltage correlation strength linearly decreased from 0.75 to 0.45, triggering a power supply stability alarm.

[0107] By executing a1~a5, this embodiment of the application combines the attenuation of adjacent window associations and the distribution differences between modes to dynamically adjust feature weights, thereby improving the robustness of association strength modeling. Cross-entropy fluctuation values ​​quantify the implicit correlation between energy consumption and vibration, avoiding misjudgment based on a single mode. The correlation coefficient matrix sequence intuitively reflects the dynamic evolution of the device-power supply state, supporting rapid decision-making.

[0108] In one possible embodiment, S13, adjusting the abnormal threshold rules of electrical equipment according to dynamic correlation, includes:

[0109] Step 131: Extract the set of characteristic parameters between the electrical equipment and the power supply in the dynamic correlation relationship, based on the state of the electrical equipment. The set of characteristic parameters includes energy consumption characteristics, vibration characteristics, dynamic weighting coefficients, and dynamic correlation coefficients. The set of characteristic parameters is a set of real-time state parameters extracted from the dynamic correlation model between the equipment and the power supply, used to characterize the correlation strength between the equipment's operating state and the power supply interaction.

[0110] For example, during continuous machining on a CNC machine tool, when the equipment is under high load, the feature parameter set for the current time window (5-second interval) is extracted from the dynamic correlation relationship: energy consumption features include an average current of 25A and a power fluctuation rate of 0.3; vibration features include an energy proportion of 90% in the 300Hz frequency band and a root mean square value in the time domain of 5.2 m / s²; the dynamic weighting coefficient is 0.68; and the dynamic correlation coefficient is a current-voltage correlation strength of 0.55. The above parameters are constructed in sequence into a feature vector of 25, 0.3, 90, 5.2, 0.68, 0.55, which is used for subsequent dynamic adjustment of threshold rules to accurately characterize the real-time interaction status between the equipment and the power supply under sudden load changes.

[0111] Step 132: Input the feature parameter set into the pre-built anomaly threshold rule adjustment model, and generate anomaly threshold rule adjustment parameters through the parameter mapping function in the anomaly threshold rule adjustment model. The anomaly threshold rule adjustment model can be a pre-trained neural network model used to map feature parameters to threshold adjustment amounts. The parameter mapping function is a mathematical function within the model that converts input features into adjustment parameters.

[0112] The parameter mapping function can be expressed using the following formula:

[0113] ;

[0114] in, X The feature parameter set is a multi-dimensional vector, including mean current, power fluctuation, 300Hz vibration energy percentage, vibration intensity, dynamic weighting coefficient, current-voltage correlation coefficient, etc. f Mapping ( X ) is the parameter mapping function. To output the activation function, This is the weight matrix from the hidden layer to the output layer. This is the output layer bias vector. It is a weight matrix. It's a bias. It is an activation function.

[0115] For example, during the high-precision machining stage of CNC machine tools, the average current in the feature parameter set is 25A, the power fluctuation rate is 0.3, the proportion of 300Hz vibration energy is 90%, the vibration intensity is 5.2, the dynamic weight coefficient is 0.68, and the current-voltage correlation coefficient is 0.55. After normalization, the feature parameter set is input into a pre-constructed abnormal threshold rule adjustment model. The model performs a linear transformation on the weight matrix and bias through a fully connected layer, and then processes it through a modified linear unit activation function. The output current threshold adjustment parameter is -6A, the difference between the original threshold of 30A and the current threshold adjustment parameter of 6A is 24A, the vibration threshold adjustment parameter is +15%, and the original threshold is 4.0 m / s²×(1+15%)=4.6 m / s². The equipment monitoring rules are updated in real time to match the current load change conditions.

[0116] Step 133: Adjust parameters according to the abnormal threshold rules. Adjust the abnormal threshold rules of the electrical equipment. The abnormal threshold rules are preset abnormal judgment thresholds (such as current threshold 30A, vibration threshold 4.0 m / s²).

[0117] Here is a specific example:

[0118] In the dynamic threshold adjustment of CNC machine tools, during CNC machining, the current window's feature parameters are extracted, including the average current of 22A and power fluctuation rate of 0.25; vibration characteristics: 250Hz energy percentage of 88% and time-domain root mean square value of 4.5 m / s²; dynamic weighting coefficient of 0.72; and dynamic correlation coefficient of 0.58 (current and voltage correlation strength). The input adjustment model, after calculation by the fully connected layer, outputs adjustment parameters: current threshold adjustment parameter is -5A, and vibration threshold adjustment parameter is +15%. The original current threshold of 30A is updated to 25A; the original vibration threshold of 4.0 m / s² is updated to 4.6 m / s². Real-time monitoring shows that when the current reaches 26A (exceeding the new threshold of 25A) and the vibration intensity reaches 4.7 m / s² (exceeding 4.6 m / s²), a "composite anomaly" alarm is triggered and pushed to the management terminal.

[0119] By executing steps 131-133, the embodiments of this application adjust the threshold based on the real-time status of the equipment (e.g., average current of 22A and correlation strength of 0.58) to avoid false alarms or missed alarms caused by fixed thresholds. By integrating energy consumption, vibration, weight and correlation strength parameters (e.g., increasing the vibration threshold by 15%), the robustness of anomaly detection is enhanced. The threshold adjustment is synchronized with the equipment operating conditions (e.g., updated every 5 seconds) to adapt to the dynamic changes in complex industrial scenarios.

[0120] In one possible embodiment, S14, when abnormal energy consumption or power failure of electrical equipment is detected according to the adjusted abnormal threshold rules, a corresponding alarm signal is triggered and pushed to the management terminal through the IoT platform, including:

[0121] Step 141: Generate an energy consumption fluctuation curve based on the energy consumption feature vector, and generate a vibration spectrum segment based on the vibration feature vector. The energy consumption fluctuation curve is a time-series curve with time on the horizontal axis and energy consumption feature values ​​(such as average current) on the vertical axis, reflecting the dynamic changes in equipment energy consumption. The vibration spectrum segment is a local segment (such as the 100-500Hz frequency band) of the frequency domain energy distribution in the vibration feature vector.

[0122] Step 142: Extract the mean offset rate of the energy consumption fluctuation curve and the proportion of the dominant frequency band energy in the vibration spectrum segment. The mean offset rate is the relative deviation rate between the current window's average energy consumption and the historical average (e.g., +15%). The dominant frequency band energy proportion is the percentage of the dominant frequency band energy in the vibration spectrum segment relative to the total energy.

[0123] Step 143: Based on the adjusted anomaly threshold rules, perform anomaly determination on the mean offset rate and the main frequency band energy ratio to determine an energy consumption anomaly event or a power outage event. The anomaly threshold rules are dynamically adjusted determination thresholds (e.g., mean offset rate threshold ±20%, main frequency band energy ratio threshold ≥80%). In this embodiment, the anomaly determination process is described as follows: If the mean offset rate exceeds the allowable offset threshold of the current anomaly threshold rules, and the main frequency band energy ratio is lower than the associated threshold, it is determined to be an energy consumption anomaly; if the main frequency band energy of the vibration spectrum suddenly drops to zero within three consecutive time windows, and energy consumption data is simultaneously lost, it is determined to be a power outage.

[0124] Step 144: For energy consumption anomalies or power outages, generate corresponding alarm signals according to the anomaly level. The alarm signals include: Level 1 alarm: the duration of the energy consumption anomaly is less than a preset safety window, triggering a low-priority alarm; Level 2 alarm: the energy consumption anomaly continues and the energy attenuation rate of the main vibration frequency band exceeds a preset safety threshold, triggering a high-priority alarm.

[0125] Step 145: Through the data channel encapsulation protocol of the IoT platform, package the alarm type, electrical equipment number, abnormal timestamp and associated data fragments corresponding to the alarm signal into a lightweight transmission message and push it to the management terminal.

[0126] The lightweight transmission message is an alarm data packet encapsulated in binary or compressed format. The equipment number is a unique identifier for the equipment within the system, used to accurately locate the equipment's identity and physical location. The anomaly timestamp is a precise record of the time the anomaly occurred, typically using an international standard time format. The associated data fragment is the original data or feature data fragment directly related to the anomaly that triggered the alarm.

[0127] Here is a specific example:

[0128] During high-precision milling on a CNC machine tool, the identification system detected that the mean offset rate of the energy consumption fluctuation curve in the current time window (timestamp 2024-03-15 09:25:30) was +28% (historical mean 20A, current mean 25.6A), and the energy proportion of the main frequency band (200-400Hz) of the vibration spectrum segment suddenly dropped to 72% (normal threshold ≥80%). Based on the dynamically adjusted abnormal threshold rules (mean offset rate threshold ±20%, main frequency energy threshold 80%), it was determined to be a composite abnormal event (abnormal energy consumption and abnormal vibration), triggering a level-two alarm. The associated data segment packet contains: instantaneous current value, main frequency energy distribution, dynamic correlation coefficient, and the data window for the previous 10 seconds. This embodiment of the application can encapsulate the alarm information into a lightweight message using the Message Queuing Telemetry Transport (MQTT) protocol of the Internet of Things platform.

[0129] By executing steps 141-145, this embodiment of the application transforms the energy consumption feature vector into an energy consumption fluctuation curve and the vibration feature vector into a spectrum segment, intuitively presenting the evolution of the equipment status; based on the adjusted threshold rules, combined with the mean offset rate (+28%) and the main frequency energy ratio (72%), it accurately identifies composite anomalies and avoids misjudgment of a single signal; alarm messages are pushed to the management terminal within 500ms via the MQTT protocol, triggering a shutdown command, while associated data segments (current waveform, energy trend) are synchronously retained, supporting root cause localization; the time from anomaly detection to alarm response is reduced by 90% compared to traditional solutions, the equipment fault interruption time is reduced by 15 minutes, and maintenance costs are reduced by 40%.

[0130] In one possible embodiment, step 122 involves inputting the energy consumption feature vectors and vibration feature vectors corresponding to all time windows into a deep learning model, and generating a joint feature vector using a cross-modal interaction mechanism through the multimodal fusion module in the deep learning model, including:

[0131] Step b1: Divide the energy consumption characteristic vectors and vibration characteristic vectors of all time windows into start-up and shutdown phases, steady-state operation phases, and load change phases according to the equipment operation stages.

[0132] The start-up and shutdown phases are transient processes during equipment startup or shutdown, characterized by rapid power increases / decreases and dispersed vibration spectrum energy. The steady-state operation phase occurs when the equipment is in a stable working state, characterized by small power fluctuations and concentrated vibration frequency band energy. The load change phase is a sudden change in load caused by changes in processing tasks, characterized by sharp increases / decreases in power and frequency shifts in vibration energy.

[0133] Step b2: Based on the power change slope of the energy consumption characteristic vector and the main frequency band energy distribution of the vibration characteristic vector in each stage, calculate the cross-modal dynamic interaction weight through the cross-modal interaction mechanism.

[0134] Among them, the power change slope is the rate of change of energy consumption characteristics (such as average current) per unit time (e.g., +8A / s). The dominant frequency band energy distribution is determined by the frequency band with the highest energy proportion in the vibration spectrum and its energy change trend. The cross-modal dynamic interaction weight quantifies the correlation strength between energy consumption and vibration signal in a specific stage.

[0135] Step b3: Based on cross-modal dynamic interactive weights, nonlinear enhancement of vibration features and superposition of energy consumption features are performed during the start-up and shutdown phases to generate a start-up and shutdown phase vector; a bimodal projection residual vector is generated during the steady-state operation phase; and a piecewise energy consumption transformation and vibration window convolution are performed during the load abrupt change phase to generate a load abrupt change vector. Specifically, nonlinear enhancement of vibration features involves applying a weight-related nonlinear transformation to the vibration features. The bimodal projection residual vector is the projection difference vector between energy consumption and vibration features in a low-dimensional space. The piecewise energy consumption transformation involves piecewise linear fitting of the power data during the load abrupt change phase to extract inflection point parameters.

[0136] Step b4: Concatenate the start-stop phase vector, the dual-modal projection residual vector, and the load mutation vector in the order of the time window to generate a joint feature vector.

[0137] The start-up / shutdown phase vector is an optimized feature vector generated during the equipment's start-up and shutdown phases, formed by superimposing nonlinear enhancements to vibration characteristics and energy consumption characteristics. The dual-modal projection residual vector, calculated during steady-state operation by projecting energy consumption and vibration characteristics into a low-dimensional space, represents the degree of difference between the two modal characteristics under steady-state conditions. The load mutation vector, an optimized feature vector generated during the load mutation phase, includes inflection point parameters from the piecewise energy consumption transformation and smoothing features obtained after convolution with the vibration window.

[0138] Here is a specific example:

[0139] CNC machine tool startup phase (0-10 seconds): power slope 8A / s, vibration main frequency band (200-400Hz) energy distribution dispersed (standard deviation) =12%), marked as start-stop phase; Steady-state phase (10-50 seconds): Power fluctuation rate 0.5%, main frequency energy accounts for 88%; Load change phase (50-55 seconds): Power slope 15A / s (due to switching processing tasks), marked as change phase. Start-stop phase interaction weight calculation:

[0140] Q=MLP(8)=[0.6,0.4], K=V=MLP(75%,12%)=[0.5,0.3], the vibration characteristics are enhanced from [75%, 12%] to [75%×e].0.7 ≈151%, 12%×e 0.7 ≈24%], concatenated with energy consumption features, becomes [20A, 0.3, 151%, 24%]; energy consumption projection is [0.8, 0.2], vibration projection is [0.7, 0.3], residual vector The inflection point time was determined to be 52 seconds by segmented fitting of energy consumption, and the main frequency energy was smoothed from 85% to 82% after vibration convolution.

[0141] By executing steps b1 to b4, this embodiment of the application suppresses transient noise interference by superimposing vibration characteristic nonlinear enhancement (e.g., increasing the main frequency energy from 85% to 151%) with energy consumption, thereby improving the detection accuracy of start-stop anomalies by 25% (e.g., increasing the bearing jamming recognition rate from 70% to 95%); the dual-modal projection residual vector (e.g., residual 0.14) quantifies the synergistic difference between energy consumption and vibration, reducing the false alarm rate of steady-state anomalies by 40% (e.g., reducing the false judgment rate from 15% to 9%); and the piecewise transformation of energy consumption (e.g., the slope difference of the inflection point is 15A / s) and convolution with the vibration window (reducing the standard deviation of the main frequency energy from 12% to 5%) effectively distinguishes between normal load switching and power coupling failure, improving the classification accuracy of sudden events by 30%. Based on cross-modal interaction weights (e.g., a weight of 0.7 for the start-stop phase), differentiated fusion of energy consumption and vibration is achieved, increasing the amount of key feature information by 50% (e.g., reducing the variance of vibration dominant frequency energy by 60%). Dynamic allocation of feature importance through an attention mechanism reduces invalid feature interference by 35% (e.g., reducing the proportion of noise frequency energy from 20% to 8%). The joint feature vector formed by concatenating phased features (e.g., 20A, 0.3, 151) fully preserves the state evolution information of the entire equipment lifecycle, supports end-to-end model training, and improves the global anomaly detection F1-score by 18% (from 0.82 to 0.97). The differentiated expression of start-stop, steady-state, and abrupt change phases in the joint features reduces root cause localization time by 50% (e.g., reducing power fault localization from 30 minutes to 15 minutes).

[0142] In one possible embodiment, step b2, calculating the cross-modal dynamic interaction weights based on the power change slope of the energy consumption characteristic vector and the dominant frequency band energy distribution of the vibration characteristic vector within each stage through a cross-modal interaction mechanism, includes:

[0143] Step b21: Input the peak value of the power change slope of the energy consumption characteristic vector during the start-stop phase and the peak value of the main frequency band energy distribution of the vibration characteristic vector into the vibration dominance factor calculation module in the multi-modal fusion module to generate the vibration dominance factor.

[0144] Among them, the peak power change slope is the absolute value of the maximum rate of change of energy consumption characteristics (such as current) during the start-up and shutdown phase. The peak energy of the dominant frequency band is the energy value of the frequency band with the highest energy proportion in the vibration spectrum during the start-up and shutdown phase. The vibration dominance factor quantifies the degree to which the vibration signal dominates anomaly detection during the start-up and shutdown phase; the larger the value, the more critical the vibration characteristic.

[0145] Step b22: Input the stability coefficient of the power change slope of the energy consumption characteristic vector during the steady-state operation phase and the dispersion of the main frequency band energy distribution of the vibration characteristic vector into the steady-state balance factor calculation module in the multi-mode fusion module to generate the steady-state balance factor.

[0146] Among them, the stability coefficient is the reciprocal of the standard deviation of the energy consumption slope in the steady-state stage. The dominant frequency energy dispersion is the variance of the dominant frequency energy proportion in the steady-state stage. The steady-state balance factor measures the balance between energy consumption and vibration characteristics in steady state; a larger value indicates stronger synergy between the two modes.

[0147] Step b23: Input the deviation amplitude of the main frequency band energy distribution of the vibration characteristic vector during the load abrupt change stage into the energy consumption weight gain calculation module in the multi-modal fusion module to generate the energy consumption weight gain. Here, the main frequency energy deviation amplitude is the difference between the current main frequency energy distribution and the historical normal distribution. The energy consumption weight gain is the weight enhancement coefficient of the energy consumption characteristics during load abrupt changes; a larger value indicates that energy consumption anomalies should be given priority attention.

[0148] Step b24: Dynamically couple the vibration dominance factor, steady-state equilibrium factor, and energy consumption weight gain across stages to generate cross-modal dynamic interaction weights. The cross-stage dynamic coupling involves weighting and fusing multimodal factors according to stage importance.

[0149] Here is a specific example:

[0150] During the CNC machine tool startup phase (0-10 seconds), the system detected a power consumption increase rate of 10 amps per second (the normal startup rate is 6 amps per second), while the energy proportion of the dominant frequency band (200-400Hz) of the vibration signal reached as high as 90%. The vibration dominance factor calculation module combined the power increase rate with the dominant frequency energy for analysis: higher than the benchmark value indicates a large startup load; high energy concentration in a specific frequency band reflects initial friction of mechanical components. Entering the steady-state operation phase (10-50 seconds), the system analyzed that the power consumption fluctuations were extremely small (fluctuation standard deviation only 0.1), and the vibration dominant frequency energy distribution was highly stable (energy proportion variance 0.03). The steady-state balance factor calculation module evaluated the following parameters: energy consumption stability: extremely low fluctuations indicate stable power output; vibration consistency: concentrated dominant frequency energy distribution with no abnormal frequency shift. When the machine tool performs heavy cutting tasks (load abrupt change at 50-55 seconds), the difference between the vibration dominant frequency energy distribution and the historical normal state reached 0.4. Vibration energy deviation indicates an abnormally dispersed distribution of dominant frequency energy, which may mask the true fault; a sudden increase in energy consumption, with the signal current surging from 20A to 28A, is directly related to the power supply load. The system is weighted according to the importance of each stage: start-up / shutdown phase 60%, steady state 30%, and sudden change 10%. The combined factors for these three stages are: start-up / shutdown contribution: 0.82 vibration dominance factor × 60% = 0.49; steady state contribution: 9.7 steady state balance factor × 30% = 2.91; sudden change contribution: 1.22 energy consumption gain × 10% = 0.12.

[0151] By executing steps b21-b24, this embodiment of the application accurately identifies transient faults such as bearing jamming through the vibration dominance factor (0.82) during the start-stop phase, reducing the false alarm rate by 40%; the steady-state balance factor (9.71) suppresses misjudgments due to small power fluctuations, reducing invalid work orders by 60%; and the energy consumption gain (1.22) during the load change phase provides an early warning of power overload 5 minutes in advance. Combined with staged parallel computing, the inference time is reduced from 60ms to 25ms, and the model memory usage is reduced by 45%. In practical applications, the overall detection accuracy is improved to 94%, the fault interruption time is shortened to 10 seconds, and the maintenance cost is reduced by 50%, providing high real-time and robust anomaly monitoring protection for complex operating conditions of industrial equipment.

[0152] It should be noted that the above-mentioned technical solutions improve the safety and reliability of power systems through intelligent and automated means, meeting the needs of modern power management, and can be extended to multiple fields such as power, transportation, and healthcare. Furthermore, this method is applicable to important locations such as hospitals, data centers, airports, and subways, improving the efficiency and accuracy of verifying the power supply relationships of end-loads in these locations while reducing labor costs and error rates.

[0153] Figure 2A schematic diagram of the structure of an intelligent verification system for end-point load power supply correlation based on multi-source data fusion provided in this application embodiment is shown below. Figure 2 As shown, the system includes:

[0154] The acquisition module 21 is used to acquire preprocessed real-time energy consumption data and mechanical vibration signals of the electrical equipment sent by the edge computing node.

[0155] The identification module 22 is used to perform multi-source data fusion of real-time energy consumption data and mechanical vibration signals using a deep learning model in order to identify the dynamic correlation between electrical equipment and power supply.

[0156] Adjustment module 23 is used to adjust the abnormal threshold rules of electrical equipment according to dynamic correlation.

[0157] The detection module 24 is used to trigger the corresponding alarm signal when the power consumption of the electrical equipment is abnormal or the power supply is disconnected according to the adjusted abnormal threshold rules, and push it to the management terminal through the Internet of Things platform.

[0158] Figure 2 The aforementioned intelligent verification system for end-point load power supply correlation based on multi-source data fusion can perform... Figure 1 The implementation principle and technical effects of the intelligent verification method for end-load power supply correlation based on multi-source data fusion described in the illustrated embodiment will not be repeated here. The specific operation methods of each module and unit in the intelligent verification system for end-load power supply correlation based on multi-source data fusion in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0159] In one possible design, Figure 2 The intelligent verification system for end-point load power supply correlation based on multi-source data fusion, as shown in the embodiment, can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32.

[0160] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0161] The processing component 32 is used to: acquire preprocessed real-time energy consumption data and mechanical vibration signals of the electrical equipment sent by the edge computing node; perform multi-source data fusion on the real-time energy consumption data and mechanical vibration signals using a deep learning model to identify the dynamic correlation between the electrical equipment and the power supply; adjust the abnormal threshold rules of the electrical equipment according to the dynamic correlation; when abnormal energy consumption or power failure of the electrical equipment is detected according to the adjusted abnormal threshold rules, a corresponding alarm signal is triggered and pushed to the management terminal through the IoT platform.

[0162] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0163] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as Random Access Memory (RAM), Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0164] Of course, computing devices may also include other components, such as input / output interfaces, display components, and communication components. Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc. Communication components are configured to facilitate wired or wireless communication between the computing device and other devices. The computing device can be a physical device or an elastic computing host provided by a cloud computing platform; in this case, the computing device can refer to a cloud server, and the aforementioned processing components and storage components can be basic server resources rented or purchased from the cloud computing platform.

[0165] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown is an intelligent verification method for the correlation between end-load power sources based on multi-source data fusion.

[0166] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0167] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0168] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for intelligent verification of the correlation between end-point load power sources based on multi-source data fusion, characterized in that, include: Acquire preprocessed real-time energy consumption data and mechanical vibration signals of electrical equipment sent by edge computing nodes; A deep learning model is used to perform multi-source data fusion on the real-time energy consumption data and the mechanical vibration signal in order to identify the dynamic correlation between the electrical equipment and the power supply. Based on the dynamic correlation, adjust the abnormal threshold rules of the electrical equipment; When abnormal energy consumption or power failure of the electrical equipment is detected according to the adjusted abnormal threshold rules, the corresponding alarm signal is triggered and pushed to the management terminal through the Internet of Things platform; The method of using a deep learning model to perform multi-source data fusion of the real-time energy consumption data and the mechanical vibration signal to identify the dynamic correlation between the electrical equipment and the power supply includes: The real-time energy consumption data and mechanical vibration signal of the electrical equipment are divided into continuous energy consumption segments and continuous vibration segments according to time windows, so as to construct the energy consumption feature vector and vibration feature vector corresponding to each time window. The energy consumption feature vectors and vibration feature vectors corresponding to all time windows are input into a deep learning model, and a joint feature vector is generated by the multimodal fusion module in the deep learning model using a cross-modal interaction mechanism. Based on the joint feature vector, the dynamic weight allocation network in the deep learning model is used to identify the changes in the correlation strength between electrical equipment and power supply, and output a matrix sequence, which is composed of the dynamic correlation coefficient matrix corresponding to all time windows. Extract the temporal variation pattern of the dynamic correlation coefficient matrix from the matrix sequence, and establish a dynamic correlation relationship based on the temporal variation pattern.

2. The method according to claim 1, characterized in that, Based on the joint feature vector, the dynamic weight allocation network in the deep learning model identifies the changes in the correlation strength between electrical equipment and power supply, and outputs a matrix sequence, including: The joint feature vectors corresponding to all time windows are arranged in chronological order to form a feature sequence. The feature sequence is traversed through a sliding window to calculate the correlation decay factor between adjacent time windows in order to generate a time-series weight vector. The joint feature vector is input into the cross-modal correlation analysis module of the dynamic weight allocation network to calculate the cross-entropy fluctuation value between the energy consumption feature vector and the vibration feature vector, and generate the inter-modal weight vector. The intermodal weight vector and the temporal weight vector are bidirectionally coupled to output a dynamic weight coefficient vector; The joint feature vector is non-linearly scaled based on the dynamic weight coefficient vector to generate a weighted feature vector; Cross-channel feature compression is performed on the weighted feature vector to obtain a dynamic correlation coefficient matrix. The dynamic correlation coefficient matrices corresponding to all time windows are aggregated along the time axis to generate a matrix sequence.

3. The method according to claim 1, characterized in that, The step of adjusting the abnormal threshold rules of the electrical equipment according to the dynamic correlation includes: Extract the set of feature parameters between the electrical equipment and the power supply in the dynamic correlation relationship under the state of the electrical equipment. The set of feature parameters includes energy consumption characteristics, vibration characteristics, dynamic weighting coefficients and dynamic correlation coefficients. The feature parameter set is input into a pre-constructed anomaly threshold rule adjustment model, and anomaly threshold rule adjustment parameters are generated through the parameter mapping function in the anomaly threshold rule adjustment model. The abnormal threshold rules of the electrical equipment are adjusted according to the abnormal threshold rules.

4. The method according to claim 1, characterized in that, When abnormal energy consumption or power failure of the electrical equipment is detected according to the adjusted abnormal threshold rules, a corresponding alarm signal is triggered and pushed to the management terminal through the IoT platform, including: An energy consumption fluctuation curve is generated based on the energy consumption feature vector, and a vibration spectrum segment is generated based on the vibration feature vector. Extract the mean offset rate of the energy consumption fluctuation curve and the main frequency band energy proportion of the vibration spectrum segment; Based on the adjusted abnormal threshold rules, the mean deviation rate and the main frequency band energy ratio are judged to determine abnormal energy consumption events or power supply failure events. For the aforementioned energy consumption anomaly or power outage event, generate corresponding alarm signals according to the anomaly level; Through the data channel encapsulation protocol of the IoT platform, the alarm type, electrical equipment number, abnormal timestamp and associated data fragments corresponding to the alarm signal are packaged into a lightweight transmission message and pushed to the management terminal.

5. The method according to claim 1, characterized in that, The step of inputting the energy consumption feature vector and the vibration feature vector corresponding to all time windows into a deep learning model, and generating a joint feature vector through the multimodal fusion module in the deep learning model using a cross-modal interaction mechanism includes: The energy consumption feature vectors and vibration feature vectors of all time windows are divided into start-up and shutdown phases, steady-state operation phases and load change phases according to the equipment operation stage. Based on the power change slope of the energy consumption characteristic vector and the main frequency band energy distribution of the vibration characteristic vector in each stage, the cross-modal dynamic interaction weight is calculated through the cross-modal interaction mechanism. Based on the cross-modal dynamic interactive weights, vibration feature nonlinear enhancement and energy consumption feature superposition are performed in the start-stop phase to generate the start-stop phase vector; dual-modal projection residual vector is generated in the steady-state operation phase; and energy consumption piecewise transformation and vibration window convolution are performed in the load mutation phase to generate the load mutation vector. The start-stop phase vector, the dual-modal projection residual vector, and the load mutation vector are concatenated in time window order to generate a joint feature vector.

6. The method according to claim 5, characterized in that, The calculation of cross-modal dynamic interaction weights based on the power change slope of the energy consumption characteristic vector and the dominant frequency band energy distribution of the vibration characteristic vector within each stage, through a cross-modal interaction mechanism, includes: The peak value of the power change slope of the energy consumption characteristic vector during the start-stop phase and the peak value of the main frequency band energy distribution of the vibration characteristic vector are input into the vibration dominance factor calculation module in the multi-modal fusion module to generate the vibration dominance factor. The stability coefficient of the power change slope of the energy consumption characteristic vector during the steady-state operation phase and the dispersion of the main frequency band energy distribution of the vibration characteristic vector are input into the steady-state balance factor calculation module in the multi-mode fusion module to generate the steady-state balance factor. The deviation amplitude of the main frequency band energy distribution of the vibration characteristic vector during the load change phase is input into the energy consumption weight gain calculation module in the multimodal fusion module to generate the energy consumption weight gain. The vibration dominant factor, the steady-state equilibrium factor, and the energy consumption weight gain are dynamically coupled across stages to generate cross-modal dynamic interactive weights.

7. An intelligent verification system for the correlation between end-point load power sources based on multi-source data fusion, characterized in that, A method for performing intelligent verification of end-load power supply correlation based on multi-source data fusion as described in any one of claims 1 to 6, comprising: The acquisition module is used to acquire preprocessed real-time energy consumption data and mechanical vibration signals of the electrical equipment sent by the edge computing node; The identification module is used to perform multi-source data fusion of the real-time energy consumption data and the mechanical vibration signal using a deep learning model, so as to identify the dynamic correlation between the electrical equipment and the power supply. An adjustment module is used to adjust the abnormal threshold rules of the electrical equipment according to the dynamic correlation. The detection module is used to trigger a corresponding alarm signal when the power consumption of the electrical equipment is abnormal or the power supply is disconnected according to the adjusted abnormal threshold rules, and push the alarm signal to the management terminal through the Internet of Things platform.

8. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the intelligent verification method for end-load power supply correlation based on multi-source data fusion as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements an intelligent verification method for the correlation between end-load power sources based on multi-source data fusion, as described in any one of claims 1 to 6.

Images