Substation main transformer equipment operation state monitoring system based on voiceprint recognition

Abstract

The invention relates to the technical field of equipment operation state monitoring, in particular to a transformer substation main transformer equipment operation state monitoring system based on voiceprint recognition. Comprising a voiceprint collection module, a data preprocessing module, a voiceprint analysis module, a state evaluation module, a cooperative monitoring module, an interaction early warning module, a block chain evidence storage module and an operation and maintenance optimization module. According to the invention, the directional acoustic sensor collects voiceprint signals of key components, synchronously collects multi-dimensional operation parameters, and combines a wavelet threshold noise reduction algorithm to filter environmental noise and timestamp synchronization algorithm associated data, thereby achieving the effects of eliminating environmental interference and ensuring the integrity and accuracy of monitoring data. A normal voiceprint baseline library is constructed through historical voiceprint data and a deep learning model, current voiceprint signals are compared in real time, the operation state is judged in combination with multi-dimensional parameters, a time sequence model is used for deducing the fault trend, and the effects of accurately recognizing the fault type, accurately dividing the equipment state and conducting early warning in advance are achieved.

Description

Technical Field

[0001] This invention relates to the field of equipment operation status monitoring technology, specifically a substation main transformer equipment operation status monitoring system based on voiceprint recognition. Background Technology

[0002] The operating status of substation main transformer equipment includes normal status, sub-health status, abnormal status, and fault status. Under normal status, the parameters of all core components and auxiliary systems of the main transformer are within the rated range, with no leakage, abnormal vibration, or abnormal noise, and it can continuously and stably carry the rated load. Under sub-health status, the core parameters of the main transformer are close to the rated threshold, or the auxiliary system shows slight abnormalities. Although it does not affect the normal power supply, there is a potential fault that requires continuous monitoring. Under abnormal status, some components of the main transformer exceed the standard or show obvious abnormalities. Although it does not lead to equipment shutdown, there is a risk of further deterioration of the fault, and it needs to be inspected and dealt with immediately. Under fault status, the core components of the main transformer are severely damaged or the critical system fails, making it unable to carry the load normally, and it may even cause a safety accident. Emergency shutdown and maintenance are required, resulting in unavoidable losses.

[0003] Generally, the methods for detecting the operating status of cooling fans and submersible oil pumps in the cooling system of substation main transformers include: manually inspecting the components by listening to their operating sounds to determine if there are abnormal noises, observing whether the components rotate smoothly and whether there are any leaks, touching the component casing to feel if the temperature is normal, verifying the component's heat dissipation effect and operating status by collecting operating parameters such as motor operating current and main transformer oil temperature, using a vibration analyzer to measure the component's vibration acceleration to troubleshoot mechanical faults, and using a megohmmeter to measure the insulation resistance of the motor windings to detect the insulation status. These detection methods have obvious drawbacks. The high noise generated by the components during manual inspection can mask slight abnormal sounds, and the detection results rely on human experience, leading to strong subjectivity and easy to miss early faults. The monitoring of operating parameters can only indirectly reflect the component status and cannot accurately locate the fault type. Vibration detection and insulation detection require specialized equipment and personnel, and insulation detection requires shutting down the components, which will affect the main transformer's heat dissipation efficiency. Overall, it is difficult to achieve early and accurate identification and continuous monitoring of faults.

[0004] Based on this, the present invention provides a substation main transformer equipment operation status monitoring system based on voiceprint recognition to solve the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this invention is to provide a substation main transformer equipment operation status monitoring system based on voiceprint recognition, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] This invention proposes a substation main transformer equipment operation status monitoring system based on voiceprint recognition, including a voiceprint acquisition module, a data preprocessing module, a voiceprint analysis module, a status assessment module, a collaborative monitoring module, an interactive early warning module, a blockchain evidence storage module, and an operation and maintenance optimization module;

[0008] The acoustic signature acquisition module is used for acquiring acoustic signature signals from various components of the main transformer and synchronously acquiring multi-dimensional operating parameters; the data preprocessing module is used for acoustic signature signal noise reduction and enhancement and spatiotemporal alignment of multi-source data; the acoustic signature analysis module is used for establishing normal acoustic signature baselines and identifying fault acoustic signature features; the status assessment module is used for real-time determination of equipment operating status and prediction of fault development trends; the collaborative monitoring module is used for real-time processing at the edge and collaborative linkage between acoustic signature monitoring terminals; the interactive early warning module is used for local sound and light early warning and remote information push; the blockchain notarization module is used for monitoring data notarization and operation and maintenance behavior incentives; and the operation and maintenance optimization module is used for model iterative optimization and dynamic scheduling of monitoring resources.

[0009] Preferably, the voiceprint acquisition module further includes an anti-interference voiceprint acquisition unit and a multi-dimensional parameter synchronization unit;

[0010] The anti-interference acoustic signature acquisition unit collects acoustic signature signals of each component during operation by deploying directional acoustic sensors near key components such as the main transformer core, windings, cooling fan, and submersible pump. The sensors have the function of resisting environmental noise shielding.

[0011] The multi-dimensional parameter synchronization unit integrates temperature sensors, vibration sensors, and current sensors to synchronously collect parameters such as the top oil temperature of the main transformer, winding temperature, component vibration acceleration, and operating current.

[0012] Preferably, the data preprocessing module further includes a noise reduction and enhancement unit and a spatiotemporal alignment unit;

[0013] The noise reduction and enhancement unit uses wavelet threshold noise reduction algorithm and spectral subtraction to filter out environmental noise and enhance the effective voiceprint signal of the collected voiceprint signal.

[0014] The spatiotemporal alignment unit uses a timestamp synchronization algorithm to accurately match and associate the pre-processed acoustic signature signal with the synchronously acquired temperature, vibration, and current parameters, forming a complete monitoring data chain.

[0015] Preferably, the voiceprint analysis module further includes a normal voiceprint baseline establishment unit and a fault voiceprint recognition unit;

[0016] The normal acoustic signature baseline establishment unit analyzes the historical acoustic signature data of each component of the main transformer under rated load and different oil temperature conditions, and combines it with a deep learning model to construct a normal operation acoustic signature baseline library for each component.

[0017] The fault acoustic signature recognition unit uses a convolutional neural network algorithm to compare the current preprocessed acoustic signature signal with the normal acoustic signature baseline in real time, and extracts the characteristic acoustic signatures corresponding to fan blade deformation, oil pump bearing wear, and winding loosening faults.

[0018] Preferably, the state assessment module further includes a real-time state determination unit and a trend prediction unit;

[0019] The real-time status determination unit combines the voiceprint recognition results with the synchronously collected temperature, vibration, and current parameters, and classifies the equipment operating status into normal state, sub-health state, abnormal state, and fault state according to a preset threshold system.

[0020] The trend prediction unit uses a time-series prediction model to analyze the changes in the intensity of fault acoustic signatures and the evolution of multi-dimensional parameters, thereby predicting the speed of fault development and the potential range of equipment damage.

[0021] Preferably, the collaborative monitoring module further includes an edge computing processing unit and an inter-device collaboration unit;

[0022] The edge computing processing unit deploys a lightweight voiceprint analysis model on the local monitoring terminal of the substation to achieve real-time processing of voiceprint signals and rapid extraction of fault features, thereby reducing data transmission latency.

[0023] The inter-device collaboration unit uses Internet of Things (IoT) communication technology to enable acoustic fingerprint monitoring terminals at different locations within the substation to automatically form a monitoring network, achieving full-area acoustic fingerprint coverage of the main transformer without blind spots.

[0024] Preferably, the interactive early warning module further includes a local early warning unit, a remote push unit, and a visualization display unit;

[0025] The local early warning unit activates tiered early warnings when it determines that the equipment is in an abnormal or faulty state through an audible and visual alarm deployed in the substation control room. Different states correspond to different audible and visual signals.

[0026] The remote push unit pushes equipment operating status, fault type, and fault location information to the mobile terminal of maintenance personnel via the mobile network; the visualization display unit displays the acoustic waveform, status judgment result, and fault development trend in a visual manner through the monitoring platform.

[0027] Preferably, the blockchain evidence storage module further includes a data evidence storage unit and an operation and maintenance incentive unit;

[0028] The data storage unit uses a blockchain distributed ledger to store the original voiceprint data, preprocessing results, status determination records, and fault handling records in an immutable manner, ensuring data traceability.

[0029] The operation and maintenance incentive unit sets incentive rules through smart contracts, and automatically issues points rewards to personnel who provide timely feedback on fault handling results and provide effective operation and maintenance suggestions. Points can be redeemed for operation and maintenance related services or materials through designated channels.

[0030] Preferably, the operation and maintenance optimization module further includes a feedback learning unit and a resource scheduling unit;

[0031] The feedback learning unit collects fault determination accuracy information and actual fault handling data from maintenance personnel, and uses them as training data for model optimization, continuously iteratively updating the normal voiceprint baseline library and the fault voiceprint recognition model.

[0032] The resource scheduling unit monitors the computing resource usage, storage resource usage, and network bandwidth consumption of each module in real time, and dynamically allocates resources according to the priority of the monitored tasks.

[0033] Based on the above system, the present invention also proposes a method for monitoring the operating status of equipment, the method comprising the following steps:

[0034] S1. Collect acoustic fingerprint signals and multi-dimensional operating parameters of the main transformer during operation, and deploy the collection equipment to ensure data coverage and integrity;

[0035] S2. Preprocess the collected data to achieve noise reduction and enhancement of the voiceprint signal and accurate spatiotemporal alignment of multi-source data;

[0036] S3. Construct a baseline library of acoustic signatures for the normal operation of each component of the main transformer based on historical data, covering different operating conditions;

[0037] S4. Real-time identification of fault acoustic signature features under the current operating state, and extraction of abnormal features through signal comparison;

[0038] S5. Combine multi-dimensional parameters to assess operational status and predict fault development trends;

[0039] S6. Real-time processing at the edge and coordinated linkage with the terminal enable full-area monitoring without blind spots;

[0040] S7. Activate the hierarchical interactive early warning mechanism to realize local alarms, remote push and visualization display;

[0041] S8. Use blockchain technology for data storage and implement operation and maintenance incentives through smart contracts;

[0042] S9. Collect operation and maintenance feedback data to optimize model performance, and dynamically schedule and monitor resources based on resource consumption.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention collects acoustic signature signals from key components using directional acoustic sensors and simultaneously acquires multi-dimensional operating parameters. It combines wavelet threshold denoising algorithms to filter out environmental noise and timestamp synchronization algorithms to eliminate environmental interference and ensure complete and accurate monitoring data. A normal acoustic signature baseline library is constructed using historical acoustic signature data and a deep learning model. Real-time comparison of current acoustic signature signals and multi-dimensional parameters are used to determine the operating status, and a time-series model is used to predict fault trends. This achieves accurate fault identification, precise equipment status classification, and early warning. Furthermore, a local lightweight model processes the monitoring data, and IoT technology is used to build a self-organizing network of terminals. In case of anomalies, the acquisition frequency of surrounding terminals is increased, thereby reducing data loss. Based on transmission delay, it achieves comprehensive monitoring without blind spots and focuses on covering abnormal areas. Through a hierarchical early warning mechanism and visualized data display, it enables maintenance personnel to grasp the equipment status in a timely and comprehensive manner and respond quickly to faults. By using blockchain to store monitoring and handling data, and smart contracts to issue fault handling tasks and calculate points rewards, it ensures that the data is tamper-proof and traceable, and incentivizes maintenance personnel to participate in handling. By supplementing the training set with maintenance feedback data to iterate the model, and dynamically allocating monitoring resources based on resource consumption, it continuously improves the accuracy of fault identification and ensures continuous and stable monitoring of abnormal areas and key components. It solves the problems of traditional inspection technology in practical applications. Attached Figure Description

[0045] Figure 1 This is a topology diagram of the substation main transformer equipment operation status monitoring system based on voiceprint recognition according to the present invention;

[0046] Figure 2 This is a flowchart of the substation main transformer equipment operation status monitoring method based on voiceprint recognition according to the present invention. Detailed Implementation

[0047] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1, please refer to Figure 1 This invention proposes a substation main transformer equipment operation status monitoring system based on voiceprint recognition, including a voiceprint acquisition module, a data preprocessing module, a voiceprint analysis module, a status assessment module, a collaborative monitoring module, an interactive early warning module, a blockchain evidence storage module, and an operation and maintenance optimization module;

[0049] It should be noted that the acoustic signature acquisition module is used for acoustic signature signal acquisition of various components of the main transformer and synchronous acquisition of multi-dimensional operating parameters; the data preprocessing module is used for acoustic signature signal noise reduction and enhancement and spatiotemporal alignment of multi-source data; the acoustic signature analysis module is used to establish normal acoustic signature baselines and identify fault acoustic signature characteristics; the status assessment module is used to determine the equipment operating status in real time and predict fault development trends; the collaborative monitoring module is used for real-time processing at the edge and collaborative linkage between acoustic signature monitoring terminals; the interactive early warning module is used for local sound and light early warning and remote information push; the blockchain evidence storage module is used for monitoring data evidence storage and operation and maintenance behavior incentives; and the operation and maintenance optimization module is used for model iterative optimization and dynamic scheduling of monitoring resources.

[0050] In this embodiment, it should also be noted that the voiceprint acquisition module further includes an anti-interference voiceprint acquisition unit and a multi-dimensional parameter synchronization unit;

[0051] The anti-interference acoustic signature acquisition unit collects acoustic signature signals of each component during operation by deploying directional acoustic sensors near key components such as the main transformer core, windings, cooling fan, and submersible pump. The sensors have the function of resisting environmental noise shielding.

[0052] The multi-dimensional parameter synchronization unit integrates temperature sensors, vibration sensors, and current sensors to synchronously collect parameters such as the top oil temperature of the main transformer, winding temperature, component vibration acceleration, and operating current.

[0053] In this embodiment, it should also be noted that the data preprocessing module further includes a noise reduction and enhancement unit and a spatiotemporal alignment unit;

[0054] The noise reduction and enhancement unit uses wavelet threshold noise reduction algorithm and spectral subtraction to filter out environmental noise and enhance the effective voiceprint signal of the collected voiceprint signal;

[0055] The spatiotemporal alignment unit uses a timestamp synchronization algorithm to accurately match and associate the pre-processed acoustic fingerprint signal with the synchronously acquired temperature, vibration, and current parameters, forming a complete monitoring data chain.

[0056] In this embodiment, it should also be noted that the voiceprint analysis module further includes a normal voiceprint baseline establishment unit and a fault voiceprint recognition unit.

[0057] The normal acoustic signature baseline establishment unit analyzes the historical acoustic signature data of each component of the main transformer under rated load and different oil temperature conditions, and combines it with a deep learning model to build a normal operation acoustic signature baseline library for each component.

[0058] The fault acoustic signature recognition unit uses a convolutional neural network algorithm to compare the current pre-processed acoustic signature signal with the normal acoustic signature baseline in real time, and extracts the characteristic acoustic signatures corresponding to fan blade deformation, oil pump bearing wear, and winding loosening faults.

[0059] In this embodiment, it should also be noted that the state assessment module further includes a real-time state determination unit and a trend prediction unit.

[0060] The real-time status determination unit combines the voiceprint recognition results with the synchronously collected temperature, vibration, and current parameters, and classifies the equipment operating status into normal state, sub-health state, abnormal state, and fault state according to a preset threshold system.

[0061] The trend prediction unit uses a time-series prediction model to analyze the changes in the intensity of fault acoustic signatures and the evolution of multi-dimensional parameters, predicting the speed of fault development and the potential range of equipment damage.

[0062] In this embodiment, it should also be noted that the collaborative monitoring module further includes an edge computing processing unit and an inter-device collaboration unit;

[0063] The edge computing processing unit deploys a lightweight acoustic signature analysis model on the local monitoring terminal of the substation to achieve real-time processing of acoustic signature signals and rapid extraction of fault features, thereby reducing data transmission latency.

[0064] The inter-equipment collaboration unit uses IoT communication technology to enable the acoustic fingerprint monitoring terminals in different locations within the substation to automatically form a monitoring network, achieving full-area acoustic fingerprint coverage of the main transformer without blind spots.

[0065] In this embodiment, it should also be noted that the interactive early warning module further includes a local early warning unit, a remote push unit, and a visualization display unit;

[0066] The local early warning unit activates tiered early warnings when it determines that the equipment is in an abnormal or faulty state by using audible and visual alarms deployed in the substation control room. Different states correspond to different audible and visual signals.

[0067] The remote push unit pushes equipment operating status, fault type, and fault location information to the mobile terminal of maintenance personnel via mobile network; the visualization display unit displays the acoustic waveform, status judgment results, and fault development trend in a visual manner through the monitoring platform.

[0068] In this embodiment, it should also be noted that the blockchain evidence storage module further includes a data evidence storage unit and an operation and maintenance incentive unit;

[0069] The data storage unit uses a blockchain distributed ledger to store the original voiceprint data, preprocessing results, status determination records, and fault handling records in an immutable manner, ensuring data traceability.

[0070] The operation and maintenance incentive unit sets incentive rules through smart contracts, automatically issuing points rewards to personnel who provide timely feedback on fault handling results and offer effective operation and maintenance suggestions. These points can be redeemed for operation and maintenance-related services or materials through designated channels.

[0071] In this embodiment, it should also be noted that the operation and maintenance optimization module further includes a feedback learning unit and a resource scheduling unit;

[0072] The feedback learning unit collects information on the accuracy of fault diagnosis and actual fault handling data from maintenance personnel, and uses it as training data for model optimization, continuously iterating and updating the normal voiceprint baseline library and the fault voiceprint recognition model.

[0073] The resource scheduling unit monitors the computing resource usage, storage resource usage, and network bandwidth consumption of each module in real time, and dynamically allocates resources according to the priority of the monitored tasks.

[0074] Example 2, please refer to Figure 2 In practical applications, this invention relates to a method for monitoring the operating status of substation main transformer equipment based on voiceprint recognition. Specifically, it includes the following steps:

[0075] S1. Collect acoustic fingerprint signals and multi-dimensional operating parameters of the main transformer during operation, and deploy the collection equipment to ensure data coverage and integrity;

[0076] Directional acoustic sensors are deployed around key components such as the main transformer core, windings, cooling fan, and submersible oil pump to collect the original acoustic signals of each component during operation. The sensors are activated in an anti-environmental noise shielding mode to reduce interference.

[0077] Simultaneously integrating temperature sensors, vibration sensors, and current sensors, it synchronously collects parameters such as top-level oil temperature, winding temperature, component vibration acceleration, and operating current of the main transformer. All acquisition devices are uniformly configured with timestamp modules to ensure that the acquisition actions are triggered synchronously, and to obtain time-stamped acoustic fingerprint signals and multi-dimensional parameter datasets.

[0078] S2. Preprocess the collected data to achieve noise reduction and enhancement of the voiceprint signal and accurate spatiotemporal alignment of multi-source data;

[0079] The original speaker signal is denoised using a wavelet threshold denoising algorithm, as shown in equation (1):

[0080]

[0081] In the formula, w ij w′ represents the wavelet coefficient of the i-th layer and j-th wavelet after wavelet transform of the original voiceprint signal. ij The denoised wavelet coefficients are λ, where λ is the threshold. σ is the noise standard deviation, and N is the signal length;

[0082] Subsequently, spectral subtraction was used to enhance the denoised signal, further suppressing residual environmental noise;

[0083] By using a timestamp synchronization algorithm, the timestamp information of each dataset is extracted. Based on the acquisition time of the voiceprint signal, parameters such as temperature, vibration, and current are matched one by one according to the timestamp to form a complete data chain containing voiceprint features and multi-dimensional parameters, ensuring the consistency of data association at the same time node.

[0084] S3. Construct a baseline library of acoustic signatures for the normal operation of each component of the main transformer based on historical data, covering different operating conditions;

[0085] Collect historical pre-processed acoustic fingerprint data and corresponding multi-dimensional parameters of the main transformer under rated load, different oil temperature, and different operating time conditions, and filter the historical data to remove outliers and invalid data;

[0086] The filtered voiceprint data was converted into Mel spectrogram feature maps and used as input to a deep learning model. The model was trained using a CNN-LSTM hybrid model, which consists of 3 convolutional layers, 2 LSTM layers and 1 fully connected layer. The convolutional kernel sizes of the convolutional layers are 3×3, 3×3 and 2×2, respectively, and the stride is 1. The number of hidden units in the LSTM layer is 128. The output dimension of the fully connected layer matches the number of component types.

[0087] By training and learning the operating rules of each component under different working conditions, normal voiceprint baselines corresponding to each component are generated, and a complete normal voiceprint baseline library is constructed according to component type and working condition.

[0088] S4. Real-time identification of fault acoustic signature features under the current operating state, and extraction of abnormal features through signal comparison;

[0089] The current voiceprint signal preprocessed by S2 is converted into a Mel spectrum feature map, which is then input into a pre-trained convolutional neural network model. The model extracts local features from the feature map through convolutional layers, reduces the dimensionality through pooling layers, and outputs a feature vector through a fully connected layer.

[0090] The feature vector is compared in real time with the baseline feature vectors of the corresponding components and operating conditions in the normal voiceprint baseline library constructed by S3, and the feature similarity is calculated. The similarity calculation is shown in equation (2):

[0091]

[0092] In the formula, Sim represents the feature similarity, and x k y is the k-th element of the current voiceprint feature vector. k is the kth element of the normal voiceprint baseline feature vector, where n is the dimension of the feature vector;

[0093] When Sim < 0.85, abnormal features are identified. The probability distribution of fault types is further calculated by the Softmax function of the model output layer, and the fault types corresponding to the fault features with the highest probability are extracted (including fan blade deformation, oil pump bearing wear, winding loosening, and abnormal core vibration).

[0094] S5. Combine multi-dimensional parameters to assess operational status and predict fault development trends;

[0095] Based on a preset threshold system, the voiceprint recognition results of S4 are fused and evaluated with the temperature, vibration, and current parameters synchronously collected by S2: when the voiceprint recognition is normal and all parameters are within the rated range (top oil temperature ≤ 85℃, winding temperature ≤ 95℃, vibration acceleration ≤ 4.5mm / s²), the following criteria apply: 2 When the operating current is 80%-110% of the rated value, it is considered to be in normal condition.

[0096] When the voiceprint recognition similarity is between 0.8 and 0.85 or a single parameter is close to the rated threshold, it is judged as a sub-healthy state;

[0097] When the voiceprint recognition is abnormal and at least one parameter exceeds the rated threshold, it is determined to be an abnormal state;

[0098] When the probability of voiceprint recognition failure is ≥0.9 and multiple parameters significantly exceed the rated threshold, it is determined to be a fault state;

[0099] The ARIMA time series prediction model was used to analyze the fault characteristic intensity and multi-dimensional parameter change data under abnormal and fault states, as shown in equation (3):

[0100] φ(L)(1-L) d X t =θ(L)ε t (3);

[0101] In the formula, φ(L) is the autoregressive operator, θ(L) is the moving average operator, d is the difference order, and X... t ε represents the characteristic intensity or parameter value at time t. t It is a white noise sequence;

[0102] By modeling the changing trends and parameter evolution patterns of fault characteristic intensity over the next 24 hours, the speed of fault development and the potential range of equipment damage can be predicted.

[0103] S6. Real-time processing at the edge and coordinated linkage with the terminal enable full-area monitoring without blind spots;

[0104] A lightweight voiceprint analysis model is deployed on the local monitoring terminal of the substation. Some of the data preprocessed by S2 is diverted to the edge terminal. The edge terminal model processes the data in real time, quickly extracts key features and makes a preliminary judgment on the operating status, reducing the data transmission pressure and latency of the central server.

[0105] Through IoT communication technology, the data acquisition terminals in different locations within the substation can automatically form a self-organizing network. The terminals can share monitoring data and preliminary judgment results in real time. When a terminal detects an anomaly, the surrounding terminals automatically increase the acquisition frequency to achieve key monitoring coverage of the abnormal area.

[0106] S7. Activate the hierarchical interactive early warning mechanism to realize local alarms, remote push and visualization display;

[0107] Based on the S5 status assessment results, a tiered warning system is activated: no warning is triggered under normal conditions;

[0108] The sub-health status is only displayed as a notification message on the local monitoring terminal;

[0109] An abnormal state triggers a local audible and visual alarm, emitting a low-frequency audible and visual signal (light flashing frequency 2 times / second, alarm sound frequency 500Hz). At the same time, the device's operating status, fault type, and fault location information are pushed to the mobile terminal of the maintenance personnel via the mobile network.

[0110] The fault condition triggers the local audible and visual alarm to operate at high frequency (light flashing frequency 5 times / second, alarm sound frequency 1000Hz), and simultaneously pushes emergency alarm information along with the fault development trend prediction results.

[0111] The monitoring platform displays the voiceprint waveform, feature similarity curve, status judgment result, and fault development trend in a visual manner using line charts and bar charts, allowing maintenance personnel to trace the data change process.

[0112] S8. Use blockchain technology for data storage and implement operation and maintenance incentives through smart contracts;

[0113] The raw voiceprint data, preprocessing results, status determination records, and fault handling records are organized into data blocks in chronological order. Each block contains the hash value of the previous block and is stored immutably through a blockchain distributed ledger to ensure that all monitoring data and handling processes are traceable.

[0114] The smart contract rules are preset so that when the equipment is determined to be in a faulty state, the contract will automatically issue a collaborative handling task to the terminals of all maintenance personnel within the substation operation and maintenance management scope;

[0115] Personnel who provide timely feedback on fault handling results and offer effective operation and maintenance suggestions will be rewarded with points calculated through contracts. The point calculation is shown in formula (4):

[0116] P = P0 × η × τ (4);

[0117] In the formula, P is the reward score, P0 is the base score (default 100 points), η is the handling efficiency coefficient (η = 1.5 when handling time ≤ 1 hour, η = 1.0 when handling time 1-3 hours, η = 0.5 when handling time > 3 hours), and τ is the handling effect coefficient (τ = 1.2 when the fault is completely resolved, τ = 0.8 when it is partially resolved, and τ = 0 when it is not resolved).

[0118] Points are recorded through a blockchain ledger and can be redeemed for maintenance-related services or materials through designated channels;

[0119] S9. Collect operation and maintenance feedback data to optimize model performance, and dynamically schedule and monitor resources based on resource consumption;

[0120] Collect information on the accuracy of fault diagnosis, the deviation between the actual fault type and the model recognition result, and the effective experience data in the handling process from the feedback of operation and maintenance personnel. Use these data as training samples to supplement the model training set, retrain the CNN-LSTM hybrid model and the convolutional neural network model, update the normal voiceprint baseline library and fault feature recognition parameters, and complete one model iteration every 100 sets of effective feedback data to improve the recognition accuracy.

[0121] Real-time monitoring of the computing resource utilization, storage resource utilization, and network bandwidth consumption of each acquisition terminal, edge terminal, and central server is used to allocate resources using a priority-based resource scheduling algorithm, as shown in equation (5):

[0122] R i =W1×C i +W2×S i +W3×B i (5);

[0123] In the formula, R i Let W1, W2, and W3 be the weighting coefficients (taking values ​​of 0.4, 0.3, and 0.3 respectively) for the resource allocation priority of the i-th monitoring node, and C be the weighting coefficients. i To calculate resource utilization, S i For storage resource utilization, B i Network bandwidth consumption rate;

[0124] Nodes with higher priority are allocated more resources to ensure the continuity and real-time monitoring of abnormal areas and critical components.

[0125] Through the above steps, this invention collects acoustic fingerprint signals of key components using directional acoustic sensors, simultaneously collects multi-dimensional operating parameters, and combines wavelet threshold noise reduction algorithm to filter out environmental noise and data associated with timestamp synchronization algorithm, thereby achieving the effect of eliminating environmental interference and ensuring the integrity and accuracy of monitoring data.

[0126] By constructing a normal voiceprint baseline library using historical voiceprint data and deep learning models, the system compares the current voiceprint signal in real time, determines the operating status by combining multi-dimensional parameters, and uses time series models to deduce fault trends, thereby achieving the functions of accurately identifying fault types, accurately classifying equipment status, and providing early warnings.

[0127] By processing monitoring data using a local lightweight model and building a self-organizing network of terminals using IoT technology, the collection frequency of surrounding terminals is increased when an anomaly occurs, thereby reducing data transmission latency, achieving full-area monitoring without blind spots, and focusing on covering the abnormal areas.

[0128] By using a tiered early warning mechanism and visualized data display, the system enables maintenance personnel to have a timely and comprehensive understanding of equipment status and to respond quickly to faults.

[0129] By using blockchain to store and monitor data, and by using smart contracts to issue fault handling tasks and calculate points rewards, the system aims to ensure that data is tamper-proof and traceable, and to incentivize maintenance personnel to participate in the handling process.

[0130] By supplementing the training set with operation and maintenance feedback data to iterate the model and dynamically allocating monitoring resources based on resource consumption, the accuracy of fault identification is continuously improved, and the continuous and stable monitoring of abnormal areas and key components is ensured. This solves the problems existing in the practical application of traditional inspection technologies.

[0131] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0132] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A substation main transformer equipment operation status monitoring system based on voiceprint recognition, characterized in that, It includes a voiceprint acquisition module, a data preprocessing module, a voiceprint analysis module, a status assessment module, a collaborative monitoring module, an interactive early warning module, a blockchain evidence storage module, and an operation and maintenance optimization module; The acoustic signature acquisition module is used for acoustic signature signal acquisition of various components of the main transformer and synchronous acquisition of multi-dimensional operating parameters; the data preprocessing module is used for acoustic signature signal noise reduction and enhancement and spatiotemporal alignment of multi-source data. The voiceprint analysis module is used to establish a normal voiceprint baseline and identify faulty voiceprint features. The status assessment module is used to determine the equipment operating status and predict the fault development trend in real time. The collaborative monitoring module is used for real-time processing at the edge and collaborative linkage between voiceprint monitoring terminals; the interactive early warning module is used for local sound and light early warning and remote information push; the blockchain evidence storage module is used for monitoring data evidence storage and operation and maintenance behavior incentives; and the operation and maintenance optimization module is used for model iterative optimization and dynamic scheduling of monitoring resources.

2. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 1, characterized in that: The voiceprint acquisition module also includes an anti-interference voiceprint acquisition unit and a multi-dimensional parameter synchronization unit; The anti-interference acoustic signature acquisition unit collects acoustic signature signals of each component during operation by deploying directional acoustic sensors near key components such as the main transformer core, windings, cooling fan, and submersible pump. The sensors have the function of resisting environmental noise shielding. The multi-dimensional parameter synchronization unit integrates temperature sensors, vibration sensors, and current sensors to synchronously collect parameters such as the top oil temperature of the main transformer, winding temperature, component vibration acceleration, and operating current.

3. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 2, characterized in that: The data preprocessing module also includes a noise reduction and enhancement unit and a spatiotemporal alignment unit; The noise reduction and enhancement unit uses wavelet threshold noise reduction algorithm and spectral subtraction to filter out environmental noise and enhance the effective voiceprint signal of the collected voiceprint signal. The spatiotemporal alignment unit uses a timestamp synchronization algorithm to accurately match and associate the pre-processed acoustic signature signal with the synchronously acquired temperature, vibration, and current parameters, forming a complete monitoring data chain.

4. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 3, characterized in that: The voiceprint analysis module also includes a normal voiceprint baseline establishment unit and a fault voiceprint recognition unit. The normal acoustic signature baseline establishment unit analyzes the historical acoustic signature data of each component of the main transformer under rated load and different oil temperature conditions, and combines it with a deep learning model to construct a normal operation acoustic signature baseline library for each component. The fault acoustic signature recognition unit uses a convolutional neural network algorithm to compare the current preprocessed acoustic signature signal with the normal acoustic signature baseline in real time, and extracts the characteristic acoustic signatures corresponding to fan blade deformation, oil pump bearing wear, and winding loosening faults.

5. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 4, characterized in that: The status assessment module also includes a real-time status determination unit and a trend prediction unit; The real-time status determination unit combines the voiceprint recognition results with the synchronously collected temperature, vibration, and current parameters, and classifies the equipment operating status into normal state, sub-health state, abnormal state, and fault state according to a preset threshold system. The trend prediction unit uses a time-series prediction model to analyze the changes in the intensity of fault acoustic signatures and the evolution of multi-dimensional parameters, thereby predicting the speed of fault development and the potential range of equipment damage.

6. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 5, characterized in that: The collaborative monitoring module also includes an edge computing processing unit and an inter-device collaboration unit; The edge computing processing unit deploys a lightweight voiceprint analysis model on the local monitoring terminal of the substation to achieve real-time processing of voiceprint signals and rapid extraction of fault features, thereby reducing data transmission latency. The inter-device collaboration unit uses Internet of Things (IoT) communication technology to enable acoustic fingerprint monitoring terminals at different locations within the substation to automatically form a monitoring network, achieving full-area acoustic fingerprint coverage of the main transformer without blind spots.

7. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 6, characterized in that: The interactive early warning module also includes a local early warning unit, a remote push unit, and a visualization display unit; The local early warning unit activates tiered early warnings when it determines that the equipment is in an abnormal or faulty state through an audible and visual alarm deployed in the substation control room. Different states correspond to different audible and visual signals. The remote push unit pushes equipment operating status, fault type, and fault location information to the mobile terminal of maintenance personnel via a mobile network. The visualization unit intuitively displays the acoustic waveform, status judgment results, and fault development trend in the form of charts through the monitoring platform.

8. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 7, characterized in that: The blockchain evidence storage module also includes a data evidence storage unit and an operation and maintenance incentive unit; The data storage unit uses a blockchain distributed ledger to store the original voiceprint data, preprocessing results, status determination records, and fault handling records in an immutable manner, ensuring data traceability. The operation and maintenance incentive unit sets incentive rules through smart contracts, and automatically issues points rewards to personnel who provide timely feedback on fault handling results and provide effective operation and maintenance suggestions. Points can be redeemed for operation and maintenance related services or materials through designated channels.

9. The substation main transformer equipment operation status monitoring system based on voiceprint recognition according to claim 8, characterized in that: The operation and maintenance optimization module also includes a feedback learning unit and a resource scheduling unit; The feedback learning unit collects fault determination accuracy information and actual fault handling data from maintenance personnel, and uses them as training data for model optimization, continuously iteratively updating the normal voiceprint baseline library and the fault voiceprint recognition model. The resource scheduling unit monitors the computing resource usage, storage resource usage, and network bandwidth consumption of each module in real time, and dynamically allocates resources according to the priority of the monitored tasks.

10. A method for monitoring the operating status of equipment in a substation main transformer based on voiceprint recognition, as described in any one of claims 1 to 9, characterized in that, The method includes the following steps: S1. Collect acoustic fingerprint signals and multi-dimensional operating parameters of the main transformer during operation, and deploy the collection equipment to ensure data coverage and integrity; S2. Preprocess the collected data to achieve noise reduction and enhancement of the voiceprint signal and accurate spatiotemporal alignment of multi-source data; S3. Construct a baseline library of acoustic signatures for the normal operation of each component of the main transformer based on historical data, covering different operating conditions; S4. Real-time identification of fault acoustic signature features under the current operating state, and extraction of abnormal features through signal comparison; S5. Combine multi-dimensional parameters to assess operational status and predict fault development trends; S6. Real-time processing at the edge and coordinated linkage with the terminal enable full-area monitoring without blind spots; S7. Activate the hierarchical interactive early warning mechanism to realize local alarms, remote push and visualization display; S8. Use blockchain technology for data storage and implement operation and maintenance incentives through smart contracts; S9. Collect operation and maintenance feedback data to optimize model performance, and dynamically schedule and monitor resources based on resource consumption.

Images