287results about "Electric winding testing" patented technology

The invention relates to the technical field of fault testing, in particular to a dry-type air-core reactor turn-to-turn short circuit monitoring method based on a magnetic field difference method. According to the method, regularization constraint is introduced in an initial baseline stage to construct a reference electromagnetic coupling matrix, a heat accumulation index is recursively analyzed based on three-phase current energy, a thermally-induced magnetic field drift model is fitted in combination with magnetic field residual errors, and combined with combined constraint of the reference electromagnetic coupling matrix and the thermal drift model, a three-phase electromagnetic coupling model is established. The normal thermally induced magnetic field drift component is dynamically stripped in real-time monitoring; and under the constraint of a reference electromagnetic coupling matrix, fitting and optimizing a current vector through weighted least square state estimation, and integrating magnetic field and current two-dimensional deviation to obtain a structure consistency deviation index for fault monitoring. According to the invention, monitoring is carried out in consideration of a linear electromagnetic transfer rule and nonlinear thermal drift compensation, a complex magnetic field environment and a variable load operation condition of the dry-type air-core reactor are adapted, and continuous and reliable monitoring guarantee is provided for long-term stable operation of equipment.

The invention discloses a high-precision transformer fault detection method and system, relates to the technical field of transformer detection, is used for solving the problems of insensitive fault characterization and delayed diagnosis under a light load condition, and judges whether to enter a load judgment mechanism or not by acquiring input side access voltage information and counting low-load accumulated duration. When entering a load discrimination mechanism, detecting the start-stop behavior of a load branch and the temperature rise trend of a winding, calculating a light load index, identifying a light load state, carrying out region division on an iron core, collecting magnetic flux density data, and calculating iron loss balance characteristics aiming at iron loss deviation which is easy to occur under a light load condition; the insulation degradation coefficient is calculated by fusing the oil temperature peak value and the dielectric loss factor, and the abnormal index is generated with the iron loss balance characteristic, so that the early warning of the early hidden fault of the transformer can be realized in a light load scene in which the traditional temperature rise and current indexes are insensitive, and the operation reliability and maintenance efficiency of the transformer in a power distribution system are improved.

The invention relates to a fault detection method and device of a converter transformer, computer equipment and a medium. The method comprises the steps of obtaining a vibration time-domain waveform signal and a current time-domain waveform signal of a winding of the converter transformer in a preset time period, determining vibration frequency response data of the winding according to the vibration time-domain waveform signal and the current time-domain waveform signal, and determining the current time-domain waveform signal of the winding according to the vibration frequency response data and preset reference vibration frequency response data. And determining a fault detection result of the winding. By adopting the method, the timeliness of fault detection of the converter transformer can be improved.

The invention relates to the technical field of power equipment state detection, in particular to a transformer winding state detection method, system and equipment based on dynamic frequency band adaptation and a storage medium. Collecting full-band frequency response data of the healthy reference winding, and performing adaptive adjustment on a preset basic frequency band based on the peak-valley distribution density to obtain a reference target frequency band matched with the frequency spectrum characteristics of the winding; generating a reference polar coordinate feature map through a phase-amplitude mapping rule, and extracting pattern texture features to construct a reference feature library; collecting detected winding data and taking the reference target frequency band as a detection frequency band to generate a detected polar coordinate feature map; a multi-dimensional evaluation result is obtained through graph space similarity comparison and feature difference quantification in combination with amplitude anomaly recognition; and outputting a normal, slight fault or serious fault state judgment result of the tested winding based on the preset evaluation threshold interval and the working condition adaptation correction rule. The problems that in the prior art, frequency band division is fixed, and threshold values lack working condition adaptation are solved.

The invention discloses a relay fault detection circuit, method and system applied to an energy storage system, and relates to the technical field of electrical control, and the method comprises the following steps: resistance dynamic compensation, coil fault analysis, contact fault analysis, mechanical performance analysis and comprehensive fault risk assessment. The comprehensive performance of the relay is analyzed through a comprehensive evaluation algorithm in combination with a coil aging coefficient, a contact health index H and a mechanical degradation index, a comprehensive fault index is obtained, the overall performance condition of the relay is judged according to the comprehensive fault index, and corresponding measures are taken. According to the method, the temperature fluctuation interference is eliminated, the accuracy of contact resistance detection is improved, the mechanical fault of the relay can be found in time through coil aging analysis, contact health evaluation and mechanical performance evaluation algorithms in combination with compensated data, multi-dimensional fault detection is realized, misjudgment and missed judgment are effectively reduced, and the fault judgment accuracy of the relay is remarkably improved.

The invention relates to a fault monitoring and protection technology for a flexible direct current transmission system, in particular to a fault detection and polarity judgment method and system for the flexible direct current transmission system, and the method comprises the steps: firstly constructing a differential current and a zero-mode current through positive electrode current and negative electrode current, and carrying out the preprocessing through band-pass filtering; then, in the short-time detection window and the long-time confirmation window, key features of difference and zero mode are extracted by adopting an adaptive truncation and robust statistical method, and zero mode compensation is realized in combination with covariance correction; and finally, through multi-scale fusion of short window and long window results, a fusion mean value and variance are calculated, a confidence interval is constructed, and rapid fault detection of the flexible direct-current power transmission system and accurate determination of the polarity of the positive electrode and the negative electrode are realized. According to the method, rapidity and robustness are taken into consideration, interference of noise and abnormal points is effectively suppressed, rapid and accurate fault identification of the flexible direct current system can be realized in a complex operation environment, and technical support is provided for operation safety and protection control of a flexible direct current power grid.

The invention discloses a dry type reactor turn-to-turn short circuit fault positioning method based on a magnetic field direction angle. Obtaining radial and axial magnetic field component data of an outer side designated area under different working conditions; dividing a measurement area according to a reactor structure and magnetic field distribution characteristics; screening optimal measurement areas under different working conditions based on the divided measurement areas and initial moment magnetic field components in the obtained radial and axial magnetic field component data; calculating a magnetic field direction angle of the optimal measurement area under different screened working conditions; making a sensor layout scheme according to the calculated magnetic field direction angles of the optimal measurement area under different working conditions and the change characteristics of the magnetic field direction angles of the measurement points under different working conditions, and determining each measurement point; fault positioning is realized by analyzing the magnetic field direction angle difference between the short-circuit working condition and the normal working condition of each measuring point. According to the method, the problems of insensitive response to the turn-to-turn short circuit fault of the inner layer encapsulation and low positioning precision of a traditional method are solved, and a high-reliability technical scheme is provided for fault detection of the dry-type air-core reactor.

The invention relates to the technical field of motor insulation state monitoring, and discloses a motor winding turn-to-turn discharge monitoring method and device based on Rogowski coils. The high-pass integrator is matched to carry out integral operation on coil output signals in a centralized mode to obtain high-frequency discharge current data, then fault characteristic quantity is constructed based on the data, the fault position of the motor cable is determined, an independent monitoring module does not need to be configured for each phase of cable, and the single high-pass integrator can complete centralized processing of multi-coil signals. Not only is the use amount of core monitoring parts greatly reduced to reduce the monitoring cost, but also the hardware layout is remarkably simplified to save the installation space, and meanwhile, the data processing nodes are reduced to reduce processor resource occupation, and the problems of high monitoring cost, large space and resource consumption and low monitoring efficiency in the prior art are effectively solved.

The invention provides an online power supply switch mutual inductor fault detection method and system, a medium and a product, and relates to the technical field of online monitoring and fault diagnosis of power equipment, and the method comprises the steps: generating a synchronous triggering instruction at the zero crossing point moment of the power frequency current of a mutual inductor; in response to the instruction, injecting an excitation pulse into a secondary circuit of the mutual inductor and synchronously acquiring a feedback voltage to obtain a mixed transient signal; extracting a pure response waveform only containing a high-frequency component from the mixed transient signal to carry out multi-resolution time-frequency transformation, and calculating and outputting a multi-dimensional feature vector representing a winding state; inputting the multi-dimensional feature vector into a pre-trained neural network model for classification operation to obtain a fault classification probability of the current transformer state; and comparing the fault classification probability with a preset confidence threshold to generate a fault diagnosis result of the mutual inductor. By implementing the method, the recognition sensitivity of the system on early weak faults and high-frequency distribution parameter changes of the mutual inductor can be improved.

The invention provides a fault monitoring method and device for a motor winding of an electric shovel. The method comprises the following steps: collecting a current signal of the motor winding at a preset frequency; calculating the instantaneous change rate of each current sampling point, and dividing the sampling points of the current monitoring period into a sudden change section and / or a stable section according to a self-adaptive threshold value; determining early warning states of all the sudden change sections and all the stable sections, and marking the sudden change sections and the stable sections with abnormal early warning states as abnormal sections; detecting whether a fault section meeting a first preset condition exists or not; and determining the fault type of the fault section according to the fault matching rule and the feature matching score of the fault section. According to the method, fine division of the abrupt change section and the stable section is realized, association of the abrupt change section is completed, segmentation of the same fault is avoided, dynamic segmentation and combination rules are protected, features are extracted in a classified manner, abnormality is dynamically judged, and different from single feature early warning, feature extraction and early warning logic of a sub-scene is protected; and finally, realizing accurate fault distinguishing in combination with feature matching scores.

The invention belongs to the technical field of signal detection, and discloses an arrangement optimization method of detection coils for transformer turn-to-turn short circuit fault detection, which comprises the step of uniformly arranging five detection coils in a high-low voltage winding gap corresponding to each coil section along the axial direction so as to ensure efficient detection and accurate positioning of a transformer turn-to-turn short circuit fault. The method is not a pure data black box, but intelligently fuses a physical mechanism and data driving, and ensures that an optimization scheme not only conforms to an electromagnetic physical rule, but also has data self-adaptive capability. The PNN supports end-to-end learning, features can be directly extracted from original signals, the defects of traditional feature engineering are overcome, and the determination efficiency and objectivity of an optimization scheme are greatly improved. Layout optimization and accurate positioning are realized synchronously: in a primary optimization process, the algorithm can accurately position a fault wire turn while outputting an optimal sensor layout scheme.

The invention relates to the technical field of fault detection, in particular to a generator stator winding turn-to-turn short circuit detection method and system, and the method comprises the steps: synchronously collecting stator branch current, an air-gap magnetic field and an engine base vibration signal; preprocessing the signal and standardizing the working condition; extracting and adaptively fusing three types of feature vectors of branch current circulation, magnetic field asymmetry and vibration feature spectrum correlation to generate a comprehensive fault feature vector; inputting the short-circuit fault into a trained turn-to-turn short-circuit intelligent classifier, and outputting a fault judgment result and a continuous short-circuit fault confidence coefficient index; and finally, combining a confidence index trend and a feature contribution degree backtracking analysis result to generate a comprehensive state evaluation report. According to the invention, high-sensitivity and high-reliability online detection and early warning of early weak turn-to-turn short circuit are realized.

The invention discloses a dry-type air-core reactor turn-to-turn short circuit fault detection method and system based on magnetic field direction angle change, and the method comprises the following steps: firstly, obtaining a plurality of operation conditions of a dry-type air-core reactor, including a normal operation condition and a turn-to-turn short circuit fault condition containing different short circuit positions; time sequence magnetic field vector data of a measurement point position is preset on the outer side; secondly, calculating a magnetic field direction angle of each measurement point according to the acquired data, and drawing a time sequence change curve of the magnetic field direction angle under a corresponding working condition; and finally, comparing the time sequence curves of the magnetic field direction angles of the measuring points under different working conditions, and judging whether a turn-to-turn short circuit fault occurs or not based on the deviation degree characteristics of the magnetic field direction angles of the turn-to-turn short circuit working condition and the normal operation working condition, thereby completing fault diagnosis. The weak magnetic field characteristic of turn-to-turn short circuit can be sensitively captured, the limitation of a traditional monitoring method is effectively overcome, and a powerful guarantee is provided for stable operation of the dry-type air-core reactor.

A pumped storage unit main transformer winding inter-turn short circuit fault diagnosis positioning method and system, comprising: pulse injection is carried out on the main transformer of the pumped storage unit in the shutdown state, and the voltage response waveform is obtained; the characteristic waveform is calculated, the different connection modes of the pumped storage unit main transformer winding are combined, whether the pumped storage unit main transformer winding exists fault and the fault phase are judged; the VMD decomposition optimization model and the NTEO wave head extraction optimization model with optimal parameters are constructed, and the instantaneous energy spectrum is obtained based on the model and any characteristic phase containing characteristic phase; the fault positioning formula is constructed based on the instantaneous energy spectrum, the fault positioning formula corresponding to the relationship between the characteristic phase and the fault phase in the voltage response waveform is selected according to the relationship between the characteristic phase and the fault phase, and the positioning of the fault coil is obtained. The method has high reliability, high fault diagnosis accuracy and small positioning result error, and can meet the needs of pumped storage unit main transformer winding inter-turn short circuit fault diagnosis and positioning in engineering.

The invention belongs to the technical field of power equipment state monitoring and fault diagnosis, and particularly relates to a dry-type air-core reactor turn-to-turn short circuit early warning method, which comprises the following steps of: establishing a circuit model and a voltage equation set of a dry-type air-core reactor in a normal operation state and a turn-to-turn short circuit fault state; acquiring a voltage signal and a current signal of the dry-type air-core reactor in a running process in real time; calculating one or more key electrical parameters of the dry-type air-core reactor based on the voltage signal and the current signal; comparing and analyzing the calculated key electrical parameters with corresponding preset thresholds or reference parameters in a normal state; and when the variable quantity or the absolute value of at least one key electrical parameter exceeds a corresponding preset range, triggering turn-to-turn short circuit early warning. According to the invention, the early and accurate early warning of the turn-to-turn short circuit fault of the dry-type air-core reactor is realized, and the technical bottleneck of slow action of traditional overcurrent protection is fundamentally broken through.

This invention relates to the field of fault testing technology, specifically to a method for monitoring inter-turn short circuits in dry-type air-core reactors based on the magnetic field difference method. This method constructs a reference electromagnetic coupling matrix by introducing regularization constraints in the initial baseline stage. Based on the recursive analysis of the three-phase current energy thermal accumulation index, it combines the magnetic field residual to fit a thermally induced magnetic field drift model. By combining the joint constraints of the reference electromagnetic coupling matrix and the thermal drift model, the normal thermally induced magnetic field drift component is dynamically removed during real-time monitoring. Under the constraints of the reference electromagnetic coupling matrix, the current vector is optimized through weighted least squares state estimation. The structural consistency deviation index is obtained by comprehensively considering the dual-dimensional deviations of the magnetic field and current for fault monitoring. This invention balances linear electromagnetic transfer laws with nonlinear thermal drift compensation for monitoring, adapting to the complex magnetic field environment and variable load operating conditions of dry-type air-core reactors, providing continuous and reliable monitoring assurance for the long-term stable operation of the equipment.

The invention discloses a dry-type transformer fault intelligent diagnosis method and system based on an improved BGRU, and is suitable for configuring an online monitoring scene of a cooling fan, continuously collecting a load current sequence and a winding temperature sequence, and obtaining the start-stop state of the cooling fan. According to the method, an effective thermal time constant is determined based on a winding temperature sequence and in combination with a cooling fan start-stop state, a thermal response blind area is determined based on the effective thermal time constant and a preset proportionality coefficient, and the preset proportionality coefficient is selected according to a load current transient sudden change direction. The thermal response blind area is used for constraining the processing of confusion of the overload state and the winding short circuit fault by the improved BGRU model, so that the two types of faults are not diagnosed too early in an unrecognizable stage, thereby reducing false alarms and keeping available diagnosis output.

The invention discloses a multi-core combination diagnosis method for turn-to-turn short circuit faults of a five-phase permanent magnet motor. Firstly, a preprocessing layer with differential pooling is used for downsampling, and original signal fault features are reserved; secondly, a feature extractor composed of a multi-scale group convolution module and a grouping channel attention mechanism is provided for extracting and enhancing features; the module generates a feature map containing multi-scale time domain information through a multi-scale group convolution module, and selectively emphasizes features containing key fault severity information by using a packet channel attention mechanism, thereby improving the interpretability of a network model. The operation condition of the five-phase permanent magnet synchronous motor is complex, so that current signals collected under turn-to-turn short circuit faults of different degrees are highly similar, however, the method can perform tasks of retention, multi-level extraction and enhancement on fault features so as to distinguish current features under different fault degrees, and the accuracy of the current features is improved. Therefore, early turn-to-turn short circuit fault diagnosis under different short circuit conditions is realized, and the method has the advantage of high diagnosis precision.

A winding deformation online detection method for drawing a Lissajous figure based on voltage and current of a single-side port of a converter transformer comprises the following steps: 1) setting a reference parameter threshold value based on a winding in a healthy state; 2) synchronously acquiring a voltage signal and a current signal of a single-side port of the converter transformer, and preprocessing the voltage signal and the current signal; 3) drawing a Lissajous figure in a Cartesian coordinate system based on the preprocessed signal; 4) extracting key geometric feature parameters from the Lissajous figure; (5) judging whether the key geometric characteristic parameters exceed a reference parameter threshold value or not, if yes, judging that the winding state is abnormal, and if not, judging that the winding state is healthy and returning to the step (2); and 6) inputting the key geometric characteristic parameters into a typical fault characteristic library for matching for the winding in an abnormal state, identifying the deformed winding, and uploading the deformed winding to a monitoring system for real-time online detection.According to the method, excessive maintenance is avoided through predictive maintenance, and the full life cycle cost of the converter transformer is reduced.

The application discloses a kind of synchronous generator insulation fault detection device and method thereof, synchronous generator includes winding, winding is stator winding or rotor winding;Insulation fault detection device includes: one or several of DC voltage drop detection unit, voltage impulse detection unit or open transformer detection unit;Therefore, by the above detection device and detection method, when synchronous generator fails, the fault position can be positioned faster, to help workers repair in time, reduce power outage time, reduce loss.

The invention belongs to the technical field of connectors, and provides a motor turn-to-turn short circuit fault diagnosis method, device and system, and the diagnosis method comprises the following steps: (1) collecting a current signal of a stator and the rotating speed of a rotor; (2) processing the current signal to filter out a fundamental frequency, and performing fast Fourier transform to obtain an actually measured frequency spectrum; (3) calculating a slip ratio according to the rotating speed, and determining a turn-to-turn short circuit fault characteristic frequency and a modulation frequency generated by load fluctuation; and (4) combining the actually measured frequency spectrum, the short-circuit fault characteristic frequency and the modulation frequency to diagnose the short-circuit fault state of the motor. According to the method, the load modulation frequency and the turn-to-turn short circuit fault characteristic frequency in a healthy state can be accurately separated, and the accuracy and reliability of asynchronous motor turn-to-turn short circuit fault diagnosis under the variable load condition are remarkably improved.

The application discloses an improved BGRU-based dry-type transformer fault intelligent diagnosis method and system, which is suitable for an online monitoring scene of a cooling fan, continuously collects a load current sequence and a winding temperature sequence, and obtains a cooling fan start-stop state. The method determines an effective thermal time constant based on the winding temperature sequence and in combination with the cooling fan start-stop state, and determines a thermal response blind area based on the effective thermal time constant and a preset proportion coefficient, wherein the preset proportion coefficient is selected according to a load current transient mutation direction. The thermal response blind area is used to constrain the improved BGRU model to handle the confusion of the overload state and the winding short-circuit fault, so that the two types of faults do not occur in the early diagnosis stage in the unidentifiable stage, thereby reducing false positives and keeping the diagnosis output available.

The utility model discloses an anti-interference shielding case for deformation detection of a transformer winding, which relates to the technical field of transformer detection equipment and comprises a shielding case main body and a movable door slidably arranged in the shielding case main body. The shielding cover main body, the movable door and the top plate are arranged to form a shielding structure for wrapping the transformer winding, electromagnetic interference of the external environment on the transformer winding can be isolated in the detection process of the transformer winding, the heat conduction copper pipes are arranged in the middle of the top plate, and the heat conduction copper pipes are provided with the horizontal parts and the inclined parts, so that the heat conduction efficiency is improved. The horizontal part can achieve heat exchange with heat in air in the shielding structure through the heat conduction fins on the outer side of the horizontal part, and the inclined part can achieve heat exchange between heat in the heat conduction copper pipe and outside air through the heat dissipation fins on the outer side of the inclined part. And the transformer winding is prevented from being damaged by heat accumulation while the anti-electromagnetic interference on the transformer winding is realized.

The invention belongs to the field of power supply and distribution, and particularly relates to a method for monitoring and identifying a broken line grounding fault in a high-voltage winding of a distribution transformer in a Dyn wiring mode in real time. The method comprises the following steps: S1, monitoring high-voltage side zero-sequence voltage and low-voltage side voltage data of the distribution transformer in real time; s2, obtaining high-voltage side zero-sequence voltage data, and setting the high-voltage side zero-sequence voltage data as a judgment basis 1; s3, synchronously collecting three-phase voltage of a low-voltage side, and designing a judgment basis 2; and S4, in combination with the judgment basis 1 and the judgment basis 2, diagnosing the internal condition of the high-voltage winding in the corresponding transformer, and correspondingly prompting the fault condition. High-voltage side data are collected through a voltage transformer, and low-voltage side data are collected in real time through an existing port meter or a low-voltage fusion terminal. Zero-sequence voltage is calculated through high-voltage-side three-phase voltage data collected by a voltage transformer, or the zero-sequence voltage is directly measured through a secondary circuit opening triangle of the voltage transformer, and generation of the high-voltage-side zero-sequence voltage serves as a starting condition.

The application belongs to the field of power system fault diagnosis, and provides a three-phase power transformer inter-turn fault detection method based on equivalent magnetic circuit to solve the technical problem that it is difficult to identify inter-turn fault through each phase working voltage. The equivalent magnetic circuit under normal operation state is established according to the structural characteristics of the three-phase power transformer, and is simplified. Any phase in the equivalent magnetic circuit under normal operation state after simplification is selected as a fault phase, and the equivalent magnetic circuit under inter-turn fault state is established. The equivalent magnetic circuit under inter-turn fault state is analyzed, and the following conclusion is derived: only when inter-turn fault occurs, N f I f ≠ 0, wherein N f represents the number of turns shortened on the primary or secondary winding due to inter-turn fault, i.e. fault turn number; I f represents the induced current on the fault turn number, i.e. fault current; whether inter-turn fault occurs is judged according to N f I f . The application has economy, and can stably detect low-level inter-turn fault during system transient and external fault because the phase voltage is decoupled.

The application discloses a kind of wind generator stator fault detection method, device and fault early warning system, the method includes: obtaining the operation data set of wind generator stator to be diagnosed, after frequency tracking algorithm and pre-design calculation method, the fault characteristic quantity and phase difference of wind generator stator to be diagnosed are obtained;Whether the stator winding asymmetry fault of wind generator stator to be diagnosed occurs based on fault characteristic quantity determination;When wind generator stator to be diagnosed occurs the stator winding asymmetry fault, the fault phase of wind generator stator to be diagnosed is determined based on phase difference.The application obtains the fault characteristic quantity and phase difference of wind generator stator to be diagnosed by using frequency tracking algorithm and pre-design calculation method, and whether the stator winding asymmetry fault occurs based on fault characteristic quantity determination, corresponding fault phase is determined based on phase difference, can continuously and quickly identify the fault defect of stator winding, improve the stator fault detection efficiency and detection accuracy.

The invention provides a sound and vibration combined turn-to-turn short circuit monitoring method and system for a dry-type air-core reactor, electronic equipment and a storage medium, and the method comprises the steps: synchronously collecting multi-source sound signals and vibration signals of a bone conduction pickup, a microphone at a specific position and a vibration acceleration sensor through a multi-mode data collection module; after uncorrelated Gaussian white noise is added to the collected signals for multiple times, EMD decomposition is carried out, M times of same-order IMF component arithmetic averaging is carried out to obtain multiple groups of IMF components, effective IMF components are screened in combination with a joint criterion of a physical consistency coefficient and an adaptive cut-off frequency, fault sensitive features are extracted through phase perception adaptive synchronous compression transformation and iterative phase verification and reconstruction, and fault detection is carried out on the fault sensitive features. And then taking normal state feature distribution as a reference, introducing a physical weight coefficient related to a fault mechanism to complete feature standardization, finally training a special expert model and an intelligent gating unit through a subdivided sample, allocating weight aggregation according to the verification accuracy of the expert model, and outputting a diagnosis result.

This invention relates to the field of permanent magnet motor fault diagnosis, specifically to a method for identifying inter-turn short-circuit faults in permanent magnet synchronous motors. The method includes: 1) constructing an experimental platform for inter-turn short-circuit faults in permanent magnet synchronous motors and building a high-dimensional feature representation space; 2) collecting and preprocessing multi-modal operating signals during motor operation, dynamically tracking the energy distribution changes of the fault signals, and fusing different modal features to form a high-dimensional joint feature tensor; 3) inputting the joint feature tensor into a constructed deep model to extract spatial features, temporal features, and key feature weights; 4) outputting the motor operating state category at the classification layer, performing lightweight optimization on the trained model, and combining transfer learning and domain adaptation mechanisms to achieve adaptive diagnosis between different motor platforms. This method achieves dynamic fusion of multi-modal features, highlighting key features and weakening redundant information, thereby significantly improving diagnostic accuracy and model interpretability.

The invention provides a machine learning device capable of diagnosing the insulation state of a winding without stopping a shield motor pump. A machine learning device (4) generates a learning model (6) for an insulation state diagnosis device (5) for diagnosing the insulation state of a winding (31) of a stator (23) used in a shield motor pump (2). A machine learning device (4) uses unprocessed data or data obtained by pre-processing at least a portion of the unprocessed data as input data, and inputs learning data including at least the input data into a learning model (6). Therefore, the learning model learns the correlation between the input data and the diagnosis information of the insulation state of the winding. The unprocessed data is data of a voltage value and a current value supplied to the motor unit (20), data of a temperature of a winding (31) of the stator (23) or a temperature, a temperature difference, or a pressure related thereto, and data of an inflow temperature and a flow rate of the working fluid in the pump unit (10).

The invention provides a current transformer polarity verification device and method, and belongs to the technical field of power system relay protection, and the device comprises a primary side verification module connected with the primary end of a to-be-verified current transformer and a secondary side detection module connected with the secondary end of the to-be-verified current transformer, and the primary side verification module and the secondary side detection module are independently arranged. The primary side verification module comprises a plastic shell, a wireless receiving module, a relay, a rechargeable lithium battery, an electric quantity display, a self-locking button switch, a rocker arm switch and a primary line binding post; the secondary side detection module comprises a secondary wire binding post, a middle pointer ammeter, an adjustable resistor and a remote control switch; wherein the secondary wire binding post, the middle pointer ammeter and the adjustable resistor are sequentially connected in series to form a detection loop connected with the secondary end of the current transformer to be calibrated. The system has the advantages of single-person operation, wireless control, integration and low cost, and can meet the efficient, low-cost and convenient on-site verification requirements of a newly built transformer substation.