Power equipment multi-node quantum precision temperature measurement system and method based on deep learning

Abstract

The application provides a power equipment multi-node quantum precision temperature measurement system and method based on deep learning, relates to the technical field of precision temperature measurement, adopts a quantum ground state fingerprint, constructs a reference fingerprint library, and configures optimized quantum measurement parameters; entangled particles are distributed by a central control unit, and a pulse is applied to obtain key temperature measurement node fluorescence timing signals and total original phase conversion values, and pure quantum phase change data sets are extracted; time-space features are constructed according to the key temperature measurement node fluorescence timing signals, and instantaneous temperature values are inversed by fusing power auxiliary data, a three-dimensional temperature field distribution map is generated by using a structure area constraint method in combination with geographic coordinates; microscopic marking and macroscopic marking are performed, a fault diagnosis report and a key node list are generated; the optimized quantum measurement parameters are updated to obtain new cycle optimized quantum measurement parameters, so that cyclic measurement is performed; the dispersed temperature measurement data is converted into an understandable and predictable intelligent diagnosis system, and the equipment operation state monitoring capability is improved.

Description

Technical Field

[0001] This invention relates to the field of precision temperature measurement technology, specifically to a multi-node quantum precision temperature measurement system and method for power equipment based on deep learning. Background Technology

[0002] Existing multi-node temperature measurement technology for power equipment is usually based on multiple sensors independently collecting temperature data and processing the temperature of each node through simple statistical methods. Related research mainly follows two relatively independent directions: "improving the accuracy of single-point measurement" or "optimizing local data processing". One focuses on improving the data accuracy of a single temperature measurement node, while the other focuses on simple analysis of data from a single node or a small number of nodes. However, in practical applications, most existing technologies remain at the level of independent processing of single-node or local node data, lacking unified modeling of the spatial correlation and temporal evolution characteristics between multi-node data. This results in multi-node temperature measurement data exhibiting "discretization and isolation" characteristics, making it difficult to effectively characterize the overall thermal state of the equipment. Existing methods typically rely solely on the temperature values ​​themselves for analysis, lacking a unified feature expression and multi-source information fusion mechanism. They are unable to effectively compare and correlate data from different nodes, limiting their diagnostic capabilities and making it difficult to cope with multi-factor coupled faults under complex operating conditions. Therefore, existing technologies suffer from insufficient overall perception capabilities, low data utilization efficiency, and limited intelligence in fault identification in multi-node temperature measurement scenarios, making it difficult to achieve an effective transformation from "multi-point measurement" to "global state cognition". Summary of the Invention

[0003] To achieve the above objectives, the present invention provides the following technical solution: a multi-node quantum precision temperature measurement system for power equipment based on deep learning, the system comprising: Fingerprint calibration module: collects real-time magnetic field noise data and power-assisted data; deploys key temperature measurement nodes and NV color center sensors; extracts photodetector magnetic resonance spectral lines through quantum state initialization and microwave frequency sweep signals to construct quantum ground state fingerprints; constructs a benchmark fingerprint library using the dynamic benchmark freeform method; and optimizes quantum measurement parameters by fusing quantum ground state fingerprints and real-time magnetic field noise data. Common-mode noise reduction and extraction module: entangled particles are distributed through the central control unit, and microwave pulses and laser pulses are applied under the synchronous clock to obtain the fluorescence timing signals of key temperature measurement nodes and the total original phase conversion value. Based on the total original phase conversion value, common-mode noise is identified and eliminated, and a pure quantum phase change dataset is extracted. Spatiotemporal temperature field construction module: Spatiotemporal features are constructed based on fluorescence time-series signals from key temperature measurement nodes, and instantaneous temperature values ​​are retrieved by integrating power-assisted data. A three-dimensional temperature field distribution map is generated by combining geographic coordinates with a structural region constraint method. Diagnostic optimization module: Performs microscopic and macroscopic labeling, generates fault diagnosis reports and a list of key nodes; and updates the quantum measurement parameters for the new cycle to perform cyclic measurements.

[0004] Furthermore, key temperature measurement nodes and NV color center sensors are deployed, and photodetector magnetic resonance spectral lines are extracted through quantum state initialization and microwave frequency sweep signal to construct a quantum ground-state fingerprint, including: Deploy G key temperature measurement nodes and G NV color center sensors on power equipment, and obtain a list of geographic coordinates of the G key temperature measurement nodes and a list of G NV color centers; The central control unit sends quantum state initialization commands to G key temperature measurement nodes to obtain G initialized NV color center sensors; Based on G initialized NV color center sensors, a microwave sweep signal is applied to extract G key temperature measurement node photodetector magnetic resonance spectral lines; and G key temperature measurement node quantum ground state fingerprints are constructed.

[0005] Furthermore, the construction of the benchmark fingerprint database using the dynamic benchmark freeform method includes: Based on the geographical coordinates of G key temperature measurement nodes, obtain the distance and weight between the a-th key temperature measurement node and the b-th key temperature measurement node, and generate a distance matrix and a weight matrix. Based on the weight matrix, spatial relationship modeling is performed to obtain the predicted quantum ground state fingerprints of G key temperature measurement nodes; Based on the difference between the predicted quantum ground state fingerprint and the quantum ground state fingerprint of G key temperature measurement nodes, biased ground state fingerprint and outlier labeled fingerprint are constructed; A benchmark fingerprint library was constructed based on the geographical coordinates of G key temperature measurement nodes, quantum ground state fingerprints, predicted quantum ground state fingerprints, biased ground state fingerprints, and outlier labeled fingerprints.

[0006] Furthermore, the configuration optimization of quantum measurement parameters by fusing quantum ground-state fingerprint and real-time magnetic field noise data includes: Retrieve the status data of power equipment; Based on the quantum ground state fingerprint of G key temperature measurement nodes and the coincidence state data of power equipment, a set of candidate microwave frequencies is formed; Based on real-time magnetic field noise data from G key temperature measurement nodes, the noise level is obtained, and a measurement mode is triggered to obtain a dynamic decoupling pulse sequence. Optimize quantum measurement parameters by combining candidate microwave frequency sets and dynamically decoupled pulse sequences.

[0007] Furthermore, the extraction of the pure quantum phase change dataset includes: The central control unit sends corresponding entangled particles to G key temperature measurement nodes; Based on the global synchronization clock signal, within the Tr time window, microwave pulses and laser pulses are applied to the entangled particles corresponding to the G key temperature measurement nodes according to the optimized quantum measurement parameters, so as to obtain the fluorescence timing signals of the G key temperature measurement nodes within the Tr time window and convert them into the original phase conversion values. Obtain the total raw phase conversion value of G key temperature measurement nodes within the Tr time window; if the total raw phase conversion value is not 0, it indicates the presence of common-mode noise, and the pure quantum phase change dataset is obtained based on the common-mode noise; if the total raw phase conversion value is 0, it indicates the absence of common-mode noise, the pure quantum phase change is equal to the raw phase conversion value, and the pure quantum phase change dataset is directly generated.

[0008] Furthermore, the step of constructing spatiotemporal features based on fluorescence time-series signals from key temperature measurement nodes and fusing power-assisted data to retrieve instantaneous temperature values ​​includes: The fluorescence time series signals of G key temperature measurement nodes are spliced ​​with the fluorescence time series signals of k adjacent nodes to construct a spatiotemporal feature tensor; By introducing power-assisted data to the spatiotemporal feature tensor, the instantaneous temperature values ​​of G key temperature measurement nodes are retrieved.

[0009] Furthermore, the method of generating a three-dimensional temperature field distribution map using structural region constraints includes: Based on the geographical coordinates of the G key temperature measurement nodes and prior knowledge of the structure of the power equipment, the G key temperature measurement nodes are divided into F regions. Establish a combination of regional temperature field constraints based on the F regions; Using the instantaneous temperature values ​​and geographic coordinates of G key temperature measurement nodes as input, and combining regional temperature field constraints, the instantaneous temperature values ​​and geographic coordinates of HH temperature measurement points are obtained through numerical methods, and a three-dimensional temperature field distribution map is generated by combining the instantaneous temperature values ​​and geographic coordinates of G key temperature measurement nodes. Where HH represents the total number of temperature measurement points in the F regions, excluding the original G key temperature measurement nodes.

[0010] Furthermore, the process of performing microscopic and macroscopic labeling to generate a fault diagnosis report and a list of key nodes includes: For G key temperature measurement nodes, based on optimized quantum measurement parameters, microwave frequency sweep signal is executed to obtain the real-time optical detection magnetic resonance spectrum of the G key temperature measurement nodes, and fitting is performed to generate the real-time quantum ground state fingerprint of the G key temperature measurement nodes. The real-time quantum ground state fingerprints of the G key temperature measurement nodes are compared with the corresponding quantum ground state fingerprints in the benchmark fingerprint library to obtain the deformable quantum ground state fingerprints of the G key temperature measurement nodes. Microscopic labeling is performed by comparing the ground state fingerprints of the deformation sub-states of G nodes with the set anomaly type threshold; normal microscopic states are labeled as 0, and abnormal microscopic states are labeled as 1. The three-dimensional temperature field distribution map at time point Te is used as input to obtain Fd feature vectors; the Fd feature vectors are compared with the known fault mode library and macroscopically labeled; normal macroscopic features are labeled as 0 and abnormal macroscopic features are labeled as 1; The deformation sub-ground state fingerprint and Fd feature vectors corresponding to the key temperature measurement nodes with abnormal micro and macro markings are fused with multi-source information to obtain the normal state probability and JH fault type probabilities of the key temperature measurement node; and the warning level is classified according to the JH fault type probabilities. For key temperature monitoring nodes corresponding to the warning level, generate fault diagnosis reports and a list of nodes to focus on.

[0011] Furthermore, the update of the new cycle to optimize the quantum measurement parameters for performing cyclic measurements includes: Based on the list of key nodes of concern, retrieve the fault diagnosis reports of the corresponding key temperature measurement nodes to determine the quantum measurement parameters that need to be adjusted and optimized; Prioritize the key temperature measurement nodes according to the warning level to obtain priority labels; set up a new measurement resource allocation table based on the priority labels. For the quantum measurement parameters to be adjusted and optimized, the new measurement resource allocation table, and the optimized quantum measurement parameters of the nearest normal critical temperature measurement node of the early warning critical temperature measurement node, a new cycle of optimized quantum measurement parameters is generated; and a new round of measurement cycle is carried out.

[0012] A deep learning-based multi-node quantum precision temperature measurement method for power equipment, the method comprising: Step S1: Collect real-time magnetic field noise data and power-assisted data; deploy key temperature measurement nodes and NV color center sensors, extract photodetector magnetic resonance spectral lines, and construct quantum ground state fingerprints; construct a benchmark fingerprint library using the dynamic benchmark freeform method; and configure and optimize quantum measurement parameters by fusing quantum ground state fingerprints and real-time magnetic field noise data. Step S2: Distribute entangled particles using the central control unit, obtain fluorescence timing signals and total original phase conversion values ​​of key temperature measurement nodes, identify and eliminate common-mode noise based on the total original phase conversion values, and extract a pure quantum phase change dataset. Step S3: Construct spatiotemporal features based on fluorescence time-series signals from key temperature measurement nodes, and invert instantaneous temperature values ​​by integrating power-assisted data. Combine this with geographic coordinates and use the structural region constraint method to generate a three-dimensional temperature field distribution map. Step S4: Perform microscopic and macroscopic labeling to generate a fault diagnosis report and a list of key nodes; update the optimized quantum measurement parameters to obtain new periodic optimized quantum measurement parameters for cyclic measurement.

[0013] The technical effects and advantages of the deep learning-based multi-node quantum precision temperature measurement system and method for power equipment of this invention are as follows: By fusing quantum ground-state fingerprints with real-time magnetic field noise data, dynamically configuring and optimizing quantum measurement parameters, and combining entangled particle distribution with common-mode noise discrimination and elimination techniques under synchronous clocks, external interference such as environmental magnetic fields is effectively eliminated, and pure quantum phase changes are extracted, improving the sensitivity and accuracy of temperature measurement to the level of quantum precision measurement. At the same time, a three-dimensional temperature field distribution map is generated using spatiotemporal feature splicing and structural region constraint methods, breaking through the limitations of traditional single-point measurement. It can intuitively and precisely reflect the internal temperature gradient and thermal distribution state of the equipment, providing highly reliable data support for the early detection of potential faults such as local overheating.

[0014] It achieves full-process automation and intelligence from quantum sensing to intelligent diagnosis. On the one hand, it constructs a benchmark fingerprint database through dynamic benchmark liberalization and performs microscopic marking based on real-time quantum ground state fingerprint comparison. Combined with macroscopic marking by three-dimensional temperature field analysis, it realizes multi-dimensional and multi-level intelligent evaluation of equipment operating status, accurately identifies fault types and classifies warning levels, and automatically generates fault diagnosis reports and key node lists. On the other hand, it has closed-loop self-optimization capability. It can dynamically adjust the optimized quantum measurement parameters and measurement resources for the new cycle based on the diagnosis results and resource allocation priorities of the previous cycle, realizing the adaptive evolution of measurement strategy. This ensures high-precision monitoring of key areas while improving overall measurement. Attached Figure Description

[0015] Fig. 1 This is a block diagram of the deep learning-based multi-node quantum precision temperature measurement system for power equipment according to the present invention. Fig. 2 This is a flowchart of the diagnostic optimization module of the present invention; Fig. 3 This is a flowchart illustrating the steps of the deep learning-based multi-node quantum precision temperature measurement method for power equipment according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figs. 1-2 As shown, a deep learning-based multi-node quantum precision temperature measurement system for power equipment includes: Fingerprint calibration module: collects real-time magnetic field noise data and power-assisted data; deploys key temperature measurement nodes and NV color center sensors; extracts photodetector magnetic resonance spectral lines through quantum state initialization and microwave frequency sweep signals to construct quantum ground state fingerprints; constructs a benchmark fingerprint library using the dynamic benchmark freeform method; and optimizes quantum measurement parameters by fusing quantum ground state fingerprints and real-time magnetic field noise data. Common-mode noise reduction and extraction module: entangled particles are distributed through the central control unit, and microwave pulses and laser pulses are applied under the synchronous clock to obtain the fluorescence timing signals of key temperature measurement nodes and the total original phase conversion value. Based on the total original phase conversion value, common-mode noise is identified and eliminated, and a pure quantum phase change dataset is extracted. Spatiotemporal temperature field construction module: Spatiotemporal features are constructed based on fluorescence time-series signals from key temperature measurement nodes, and instantaneous temperature values ​​are retrieved by integrating power-assisted data. A three-dimensional temperature field distribution map is generated by combining geographic coordinates with a structural region constraint method. Diagnostic optimization module: Performs microscopic and macroscopic labeling, generates fault diagnosis reports and a list of key nodes; and updates the quantum measurement parameters for the new cycle to perform cyclic measurements.

[0018] Real-time magnetic field noise data of G key temperature measurement nodes (G is greater than or equal to 2, and the specific upper limit can be flexibly adjusted according to the actual structure of the power equipment) are obtained by conventional magnetic field noise measurement methods (such as adjacent node differential or ODMR spectral line broadening analysis, both of which are existing technologies). The power auxiliary data includes real-time voltage, current, ambient temperature, and ambient humidity of the equipment; Real-time voltage and current data are obtained by reading data from the online monitoring module of the power equipment itself and connecting to the central control unit through the data interface; ambient temperature and humidity data are obtained by deploying environmental sensors in key areas of the power equipment and uploading the collected data to the central control unit in real time via wired or wireless means. Deploy key temperature measurement nodes and NV color center sensors, extract photodetector magnetic resonance spectral lines through quantum state initialization and microwave frequency sweeping signals, and construct a quantum ground-state fingerprint, including: Deploy G key temperature measurement nodes on the power equipment, and deploy G NV color center sensors at the G key temperature measurement nodes to obtain a list of geographic coordinates of the G key temperature measurement nodes and a list of G NV color centers; The geographic coordinate list includes the geographic location and node number of the node; the NV color center list includes the sensor serial number, the wavelength of the integrated laser excitation optical path, the frequency of the microwave control antenna, and the parameters of the fluorescence collection module. Sensor serial numbers are bound to node numbers, and they are located at the same geographic coordinates; During deployment, positioning tools (such as GPS locators or total stations) are used to obtain the geographic coordinates of the deployment nodes; during the deployment of NV color center sensors, each NV color center sensor is tested through a dedicated NV color center characterization system (including a confocal microscope, microwave generator, and single-photon counter) to output an NV color center list; The central control unit sends quantum state initialization commands to G key temperature measurement nodes NV color center sensors to obtain G initialized NV color center sensors; The quantum state initialization command steps include: using a laser with a fixed center wavelength (e.g., 532nm) (laser power set to 10mW to 50mW), splitting it into G paths via an optical fiber splitter, each coupled to a G NV color center sensor; before each laser path enters the NV color center sensor, an acousto-optic modulator (AOM) is set up; the TTL pulse signal output by the central control unit controls the radio frequency drive switch of the AOM, thereby cutting the continuous laser into light pulses of precise duration; by adjusting the AOM, the laser power irradiating the NV color center sensor is stabilized within the range of 10mW-15mW; The light pulse, through the light pumping effect, polarizes the electron spin of the NV color center sensor from an arbitrary initial state to the ground state m_s=0, thus completing the quantum state initialization of the NV color center sensor; The central control unit is the central brain and command and dispatch center of the entire temperature measurement system. It is usually deployed in the main control room or data center cabinet. In terms of hardware, it consists of a high-performance arbitrary waveform generator (AWG), a microwave signal source, and a synchronous clock distributor. Based on G initialized NV color center sensors, a microwave sweep signal is applied, and a single-photon detector (such as an avalanche photodiode APD or a single-photon detector SPD) is used to extract G key temperature measurement node photodetector magnetic resonance spectral lines. The G key temperature measurement node photodetector magnetic resonance spectral lines are fitted (such as using the Lorentz fitting algorithm or the Gaussian fitting algorithm, both of which are existing technologies) to obtain the center frequency and spectral linewidth, and the center frequency and spectral linewidth are combined to form a quantum ground state fingerprint. Among them, the microwave sweep frequency signal is generated by the arbitrary waveform generator of the central control unit and applied to the G initialized NV color center sensors; By deploying NV color center sensors at key temperature measurement nodes and extracting photodetector magnetic resonance spectral lines, a high-precision quantum ground state fingerprint is constructed, enabling interference-resistant and fast-response instantaneous temperature measurement at a single node.

[0019] A benchmark fingerprint database is constructed using the dynamic benchmark liberalization method, including: Based on the list of geographic coordinates of G key temperature measurement nodes, obtain the distance and weight between the a-th key temperature measurement node and the b-th key temperature measurement node, and generate a distance matrix and a weight matrix. If there are 6 key temperature measurement nodes for power equipment, they are numbered Node1 to Node6. The coordinates of Node-1 are (0, 0, 0), and its quantum ground state fingerprint includes the center frequency and spectral linewidth, with the center frequency being 2870.5MHz. The coordinates of Node-5 are (2.0, 2.0, 0.5), and its center frequency is 2865MHz. The distance between Node-1 and each of the key temperature measurement nodes in Node-1 to Node-6 can be calculated using the Euclidean method, resulting in a distance matrix. The distance from Node-1 to Node-5 is 2.69, and the units are determined based on the coordinates of the key temperature measurement nodes for power equipment, including meters, decimeters, centimeters, millimeters, and other units representing distance. For the distance from Node-1 to Node-5, a weight function (such as the Gaussian kernel function, which is an existing technology) is used to calculate the weight of Node-1 and Node-5 as 0.035. Then, the weight of each node from Node-1 to Node-6 is calculated to generate a weight matrix. Based on the weight matrix, spatial relationship modeling is performed, and the predicted quantum ground state fingerprints of G key temperature measurement nodes are calculated. The predicted quantum ground state fingerprint includes the predicted center frequency and the predicted spectral linewidth. To calculate the predicted center frequency of Node-1, spatial relationship modeling is performed based on the distance and weights to Node-2, Node-3, Node-4, Node-5, and Node-6 (such as using a weighted average method or spatial interpolation, both of which are existing technologies), resulting in a predicted center frequency of 2870.49MHz. Similarly, the predicted center frequency of Node-5 is calculated to be 2870.5MHz. The predicted center frequency and predicted spectral linewidth of each node from Node-1 to Node-6 are calculated similarly to generate a predicted center frequency matrix and a predicted spectral linewidth matrix. The predicted quantum ground state fingerprint is then constructed from the predicted center frequency matrix and the predicted spectral linewidth matrix. Spatial modeling refers to characterizing and predicting the features of each node using the geographic / spatial relationships between nodes, allowing the prediction results to take spatial correlation into account rather than being calculated in isolation. For the predicted quantum ground state fingerprint and quantum ground state fingerprint of G key temperature measurement nodes, if the deviation calculation formula (existing technology) is used, the deviation ground state fingerprint of G key temperature measurement nodes is obtained; based on the comparison between the deviation ground state fingerprint and the outlier threshold range, the outlier labels of key temperature measurement nodes are performed, and the outlier label fingerprints of G key temperature measurement nodes are generated. Among them, the deviation ground state fingerprint includes center frequency deviation and spectral linewidth deviation; Set an outlier threshold range (e.g., using an empirical method). If the deviation ground state fingerprint is not within the outlier threshold range, the outlier is marked as 1, indicating an anomaly; if the deviation ground state fingerprint is within the outlier threshold range, the outlier is marked as 0, indicating normal. Calculate the outlier label for each node from Node-1 to Node-6 to generate an outlier label fingerprint. For example, the outlier threshold range includes the center frequency deviation range and the spectral linewidth deviation range. A center frequency deviation range greater than 2.0 MHz is defined as an outlier, and a spectral linewidth deviation range greater than 0.5 MHz is defined as an outlier. Node-1 has a center frequency deviation of 0.01 MHz and a spectral linewidth deviation of 0 MHz; Node-5 has a center frequency deviation of 5.50 MHz and a spectral linewidth deviation of 1.3 MHz. Therefore, Node-1 is marked as 0, indicating no outlier, and Node-5 is marked as 1, indicating an outlier. If only one of the center frequency deviation or spectral linewidth deviation falls within the defined range, it is also marked as 1, indicating an outlier. A benchmark fingerprint library is constructed based on the geographic coordinate list of G key temperature measurement nodes, quantum ground state fingerprint, predicted quantum ground state fingerprint, biased ground state fingerprint, and outlier labeled fingerprint; By utilizing node spatial relationships and deviation analysis to generate a benchmark fingerprint database, we can achieve unified modeling and anomaly comparison capabilities for data from multiple key temperature measurement nodes, providing a reference for overall status perception.

[0020] The quantum measurement parameters are optimized by integrating quantum ground-state fingerprints and real-time magnetic field noise data, including: Retrieve the electrical equipment compliance status data from the equipment rated operating condition database and historical temperature data; Based on the quantum ground state fingerprint of G key temperature measurement nodes and the coincidence state data of power equipment, a set of G candidate microwave frequencies is generated by utilizing the linear relationship between zero-field splitting of NV color center sensor and temperature (existing technology). Among them, the power equipment compliance status data includes: equipment rated operating condition parameters and historical temperature statistics; the equipment rated operating condition parameters include rated voltage, rated current, rated power, rated frequency, rated load rate and rated temperature rise, which are provided by the power equipment manufacturer, determined when the power equipment is put into operation, and stored in the equipment rated operating condition database; Historical temperature statistics include the historical instantaneous temperature mean, historical instantaneous temperature median, historical instantaneous temperature standard deviation, historical instantaneous temperature maximum and minimum, historical instantaneous temperature quartiles, historical instantaneous temperature rate of change, and historical temperature fluctuation spectrum characteristics, all stored in a historical temperature database.

[0021] For the real-time magnetic field noise data of G key temperature measurement nodes, conventional signal processing methods (such as integrated noise power or equivalent magnetic field noise root mean square value, both of which are existing technologies) are used to calculate the noise level of the G key temperature measurement nodes and trigger the measurement mode to obtain G dynamic decoupling pulse sequences. The measurement mode is as follows: if the noise level is less than sd, it indicates a low noise level; if the noise level is greater than or equal to sd, it indicates a high noise level; a low noise level triggers the aa sequence in the G dynamic decoupling pulse sequences, and a high noise level triggers the bb sequence in the G dynamic decoupling pulse sequences; sd represents the standard threshold for the noise level, which can be obtained from relevant materials. G optimized quantum measurement parameters are formed by combining G candidate microwave frequencies and G dynamically decoupled pulse sequences. By combining candidate microwave frequencies and real-time magnetic field noise data, optimized quantum measurement parameters are automatically generated, improving measurement accuracy and dynamic adaptability.

[0022] Extract the pure quantum phase transition dataset, including: The central control unit prepares G entangled particles in the maximally entangled state and sends them to G key temperature measurement nodes through optical fiber links, ensuring that each key temperature measurement node has one entangled particle. Among them, the G entangled particles are all correlated, and entangled particles can be constructed using GHZ states (which are existing technologies); Based on the global synchronization clock signal, within the Tr time window, microwave pulses and laser pulses are applied to the entangled particles corresponding to the G key temperature measurement nodes according to the G optimized quantum measurement parameters (the specific microwave pulses and laser pulses can be determined empirically); based on the fluorescence collection modules in the NV color center sensors of the G key temperature measurement nodes, the fluorescence timing signals of the G key temperature measurement nodes within the Tr time window are obtained; through the calibrated fluorescence intensity-layout number curve and the Ramsey interference principle, they are converted into the original phase conversion values; Obtain the total raw phase conversion value of G key temperature measurement nodes within the Tr time window; if the total raw phase conversion value is not 0, it indicates the presence of common-mode noise, and the pure quantum phase change dataset is obtained based on the common-mode noise; if the total raw phase conversion value is 0, it indicates the absence of common-mode noise, and the pure quantum phase change is equal to the raw phase conversion value, and the pure quantum phase change dataset is directly generated. A positive pure quantum phase transformation indicates an instantaneous temperature increase at the critical temperature measurement node; a negative pure quantum phase transformation indicates an instantaneous temperature decrease at the critical temperature measurement node; and a pure quantum phase transformation of 0 indicates that the instantaneous temperature of the critical temperature measurement node remains unchanged. The value of the Tr time window is usually between 50 microseconds and 1 second. The specific value needs to be determined based on the upper limit of the coherence time of the NV color center sensor and the thermal response characteristics of the power equipment. By distributing entangled particles and eliminating common-mode noise, pure quantum phase changes are extracted, enabling precise capture of instantaneous temperature changes and supporting real-time monitoring.

[0023] For example, if the original phase conversion values ​​of Node-1 to Node-6 in time window Tr are 0.03, 0.04, 0.02, 0.05, 0.01, and 0.03 respectively, and the total original phase conversion value of time window Tr (using the summation formula) is 0.18, according to the theoretical constraints of entangled particles (existing quantum mechanics theory, under ideal common-mode noise-free conditions, the algebraic sum of the original phase conversion values ​​of all key temperature measurement nodes is zero), it is determined that time window Tr includes the amount of common-mode noise. The common-mode noise amount is obtained according to the formula S=G*nv, where S is the total original phase conversion value, G is the number of key temperature measurement nodes, and nv is the common-mode noise amount. For example, if the total original phase conversion value is 0.18 and the number of nodes is 6, then the common-mode noise is 0.03. The pure quantum phase change is obtained by using the difference between the original phase conversion value and the common-mode noise of the G key temperature measurement nodes. For example, the pure quantum phase change of Node-1 to Node-6 is 0.00, 0.01, -0.01, 0.02, -0.02, and 0.00, respectively. Then the instantaneous temperature of Node-1 and Node-6 remains unchanged, the instantaneous temperature of Node-2 and Node-4 increases, and the instantaneous temperature of Node-3 and Node-5 decreases.

[0024] If the original phase conversion values ​​of Node-1 to Node-6 in the time window Tz are 0.02, -0.01, 0.03, -0.02, 0.01 and -0.03 respectively, and the total original phase conversion value of the time window Tz is 0, it indicates that there is no common-mode noise and the pure quantum phase change is equal to the original phase conversion value. Therefore, the pure quantum phase change values ​​of Node-1 to Node-6 are 0.02, -0.01, 0.03, -0.02, 0.01 and -0.03.

[0025] Spatiotemporal features are constructed based on fluorescence time-series signals from key temperature measurement nodes, and instantaneous temperature values ​​are retrieved by integrating power-assisted data, including: The fluorescence time-series signals of G key temperature measurement nodes are concatenated with the fluorescence time-series signals of k neighboring nodes to generate a spatiotemporal feature tensor; where k represents the number of neighboring nodes, k=1, 2, ..., G-1; the K-nearest neighbor algorithm (existing technology) can be used to determine the number of neighboring nodes; the specific concatenation can be performed using the torch.cat or tf.concat functions. Based on the pure quantum phase change, the preliminary instantaneous temperature values ​​of G nodes are obtained by utilizing the linear temperature relationship of the NV color center. The preliminary instantaneous temperature values ​​and spatiotemporal feature tensors of the corresponding key temperature measurement nodes are then input into a regression model (such as multilayer perceptron MLP, random forest, or lightweight neural network, all of which are existing technologies) to inversely derive the instantaneous temperature values ​​of the G key temperature measurement nodes. The critical temperature measurement nodes of power equipment are sparsely deployed (only the critical temperature measurement nodes have sensors), but it is necessary to know the instantaneous temperature distribution of the entire power equipment in order to fully assess the health status of the equipment. The spatiotemporal feature tensor is a multidimensional data structure; for example, three key temperature measurement nodes are deployed: Node-1, Node-2, and Node-3; within a certain time window, each key temperature measurement node collects fluorescence time-series signals at three time points (the values ​​have been normalized), {Node-1, t1(0.52), t2(0.53), t3(0.54)}, {Node-2, t1(0.61), t2(0.62), t3(0.54)}, {Node-2, t1(0.61), t2(0.62), t3(0.54)}. 0.63)},{Node-3,t1(0.48),t2(0.49),t3(0.50)},t1,t2,t3 are the three time points collected; among them, the fluorescence time series signals of Node-1 and Node-2 are spliced ​​together, and the spliced ​​spatiotemporal feature tensor is {[0.52,0.61],[0.53,0.62],[0.54,0.63]}, which is a 3*2 matrix (3 time points and 2 key temperature measurement nodes); The process by which the regression model retrieves the instantaneous temperature values ​​of the G key temperature measurement nodes is as follows: The initial instantaneous temperature values, spatiotemporal feature tensors, and power-assisted data are normalized and merged into a one-dimensional feature vector, which serves as the input feature of the regression model. At the same time, the instantaneous temperature values ​​of key temperature measurement nodes within the same time window are obtained through independent high-precision reference thermometers (such as fiber optic temperature sensors) or simulation calibration methods as output labels, thereby constructing a training sample set. The sample set is divided into a training set and a test set, and a regression model is trained. The architecture of the regression model can be a multilayer perceptron (MLP), a random forest, or a lightweight neural network. These models and their application processes are existing conventional technical solutions, and will not be elaborated here. It is only necessary to output the instantaneous temperature values ​​of key temperature measurement nodes based on the mapping relationship learned internally. By splicing together the fluorescence time-series signals from multiple key temperature measurement nodes to form a spatiotemporal feature tensor, and combining it with deep learning to invert the instantaneous temperature, a fine reconstruction from sparse nodes to global temperature is achieved.

[0026] The three-dimensional temperature field distribution map is generated using the structural region constraint method, including: Based on the geographical coordinates of the G key temperature measurement nodes and prior knowledge of the structure of the power equipment (such as CAD drawings), the G key temperature measurement nodes are divided into F regions (such as conductor region, insulation region, heat dissipation channel region, etc.). Establish a combination of regional temperature field constraints based on the F regions; Using the instantaneous temperature values ​​and geographic coordinates of G key temperature measurement nodes as input, and combining regional temperature field constraints, the instantaneous temperature values ​​and geographic coordinates of HH temperature measurement points are calculated using numerical methods, and a three-dimensional temperature field distribution map is generated by combining the instantaneous temperature values ​​and geographic coordinates of G key temperature measurement nodes. Where HH represents the total number of temperature measurement points in F regions, excluding the original G key temperature measurement nodes; the value of F is greater than or equal to 1. If there are two key temperature measurement nodes, they must be deployed in at least one region, so the lower limit of F is 1, and the upper limit is determined according to the specific region of the power equipment; the value of HH is greater than or equal to F, because HH is the total number of temperature measurement nodes in F regions and the connection between regions. If there is only one region, there must be at least one temperature measurement node. By constraining structural regions and performing numerical calculations, the temperature of key temperature measurement nodes is extended to the entire three-dimensional space of the equipment, enabling comprehensive, continuous, and visualized temperature field monitoring.

[0027] For example, monitoring a local area of ​​a transformer winding in power equipment, which includes a conductor area (copper winding) and an insulation area (insulating paper / oil gap); the conductor area has a high thermal conductivity, heat dissipates quickly, and the instantaneous temperature distribution is smooth and continuous; the insulation area has a low thermal conductivity, high thermal resistance, and allows for a large instantaneous temperature gradient on both sides. Six quantum sensor nodes were deployed in a local area of ​​the transformer winding, with known geographical coordinates and instantaneous temperature values; Based on prior knowledge of the equipment structure (such as CAD drawings), the conductor region is known to have the following ranges: x∈[0,4]cm, y∈[0,1]cm, z∈[0,1.5]cm (a cuboid region), and the insulation region is known to have the following ranges: x∈[0,4]cm, y∈[1,2.5]cm, z∈[0,1.5]cm (closely above the conductor region). The thermal conductivity coefficients of the conductor and insulation regions are known; x, y, and z represent the horizontal, vertical, and horizontal coordinates, respectively. The combination of constraints on the temperature field is as follows: Constraints within the conductor region: The instantaneous temperature change between any two monitoring nodes must be smooth, and drastic jumps are not allowed; the smoothing range of instantaneous temperature change can be set empirically. Constraints within the insulating region: Smoothness is also required, but due to the low thermal conductivity, the instantaneous temperature gradient can be larger than that in the conductor region; Interface constraint: At the interface between the conductor region and the insulation region, the instantaneous temperature value itself should be continuous (because the materials on both sides of the interface are in good contact), but the instantaneous temperature gradient can change abruptly. That is, when crossing from the conductor region to the insulation region, the instantaneous temperature value cannot jump, but the slope of the instantaneous temperature value change can suddenly become steep. The combination of temperature field constraints is set based on engineering standards and empirical data. The power industry has a large number of standards that specify the temperature limits and temperature rise characteristics of different equipment, which can be obtained from power-related GB standards, equipment manufacturer manuals and operation and maintenance procedures. Spatial interpolation was used to generate instantaneous temperature values ​​for mh nodes in the conductor region and hm nodes in the insulation region, respectively. Based on the interface, the instantaneous temperature values ​​of mh nodes and hm nodes in the conductor region are used to generate the instantaneous temperature values ​​of hh nodes at the interface according to the constraint optimization method (such as weighted residual minimization or iterative correction method, both of which are existing technologies). A three-dimensional temperature field distribution map (which can be gridded data) is generated based on the instantaneous temperature values ​​of mh nodes in the conductor region, hm nodes in the insulation region, and hh nodes at the interface. That is, each coordinate position in the three-dimensional coordinate space corresponds to an instantaneous temperature value. The total number of mh, hm, and hh is HH, and the values ​​of mh, hm, and hh are all greater than or equal to 1. Instantaneous temperature changes are the macroscopic manifestation of equipment failure. However, before the instantaneous temperature rises, the microstructure of the material may have already changed (such as changes in molecular structure due to insulation aging or abnormal local stress distribution due to poor contact). These microscopic changes directly affect the shape of the photodetector magnetic resonance spectral lines of quantum sensors, requiring very early warning of failures.

[0028] Perform micro- and macro-level labeling to generate a fault diagnosis report and a list of key nodes, including: For G key temperature measurement nodes, based on G optimized quantum measurement parameters, a microwave frequency sweep signal is executed (which can be set the same as in step 103), and the fluorescence time-series signals at L microwave frequencies (L ≥ 1) are recorded to obtain the real-time photodetector magnetic resonance spectra of the G key temperature measurement nodes. The real-time quantum ground state fingerprints of the G key temperature measurement nodes are then generated by fitting (such as Lorentz fitting or Gaussian fitting). The real-time quantum ground state fingerprints include the real-time center frequency and the real-time spectral linewidth. The real-time quantum ground state fingerprints of the G key temperature measurement nodes are compared with the corresponding quantum ground state fingerprints in the benchmark fingerprint library to obtain the deformable quantum ground state fingerprints of the G key temperature measurement nodes. Microscopic labeling is performed by comparing the ground state fingerprints of the deformation sub-states of G nodes with the set anomaly type threshold; normal microscopic states are labeled as 0, and abnormal microscopic states are labeled as 1. The three-dimensional temperature field distribution map in the time window Te is used as input, and a three-dimensional convolutional neural network or video classification network (which is the existing technology) is used to obtain Fd feature vectors. The Fd feature vectors are compared with a known fault mode library (the fault mode library is obtained by training through historical fault data) and macroscopic labels are made. Normal macroscopic labels are marked as 0, and abnormal macroscopic labels are marked as 1. The deformable sub-ground state fingerprints and Fd feature vectors corresponding to the key temperature measurement nodes with abnormal micro and macro markings are fused from multiple sources. A classifier (such as a random forest classifier) ​​is used to obtain the normal state probability and JH fault type probabilities (JH≥2) of the key temperature measurement node. The warning level is divided according to the JH fault type probabilities, including Level 1 warning, Level 2 warning and Level 3 warning. Multi-source information includes data with micro-anomaly markings, data with macro-anomaly markings, power auxiliary data, and historical fault data; For the key temperature measurement nodes corresponding to the warning level, generate a fault diagnosis report and a list of nodes to focus on; The threshold for abnormal types should be set by referring to relevant materials; The fault mode library includes fault types such as local overheating, insulation aging, poor contact, and electromagnetic interference. Each fault type has a corresponding feature vector template. For example, if the center frequency deviation in the ground state fingerprint of the deformable sub-state is within the range of a1 to a2 and the spectral linewidth is within the range of b1 to b2, then it is local overheating. "Center frequency deviation within the range of a1 to a2 and spectral linewidth within the range of b1 to b2" is the threshold range for local overheating in the anomaly type. The Fd feature vectors are high-dimensional representations (32≤Fd≤1024) learned by the network model during training, such as instantaneous temperature field statistical features and spatial gradient features. Instantaneous temperature field statistical features include the maximum instantaneous temperature value, minimum instantaneous temperature value, instantaneous temperature mean, instantaneous temperature variance, instantaneous temperature skewness, and instantaneous temperature kurtosis. Spatial gradient features include the average gradient and gradient rate of change of instantaneous temperature in the x, y, and z directions. The fault diagnosis report includes information on abnormal nodes (node ​​number, geographic coordinates), probability of fault type, warning level, and key evidence (deformed quantum ground state fingerprint, detection time window, duration of abnormality, and current optimized quantum measurement parameters). For example, for Node-1, the probability of normal state is 0.92, the probability of local overheating is 0.03, the probability of insulation aging is 0.02, the probability of poor contact is 0.02, and the probability of electromagnetic interference is 0.01. Therefore, the highest probability fault of Node-1 is the normal state. For Node-3, the probability of normal state is 0.05, the probability of local overheating is 0.75, the probability of insulation aging is 0.10, the probability of poor contact is 0.07, and the probability of electromagnetic interference is 0.03. Therefore, the highest probability fault of Node-3 is local overheating. The warning levels are divided according to the probability of each fault type JH; for example, a probability range of [0, 0.5] does not require a warning, a probability range of (0.5, 0.7] is a Level 1 warning (attention required, enhanced monitoring), a probability range of (0.7, 0.85] is a Level 2 warning (warning issued, prepare for maintenance), and a probability range of (0.85, 1] ​​is a Level 3 warning (danger, immediate shutdown). If Node-3 is a Level 2 alert, and only Node-3 is abnormal among Nodes 1 to Node-6, then the list of nodes to focus on is [Node-3]; if Node-4 and Node-5 are Level 3 alerts, then the list of nodes to focus on is [Node-4, Node-5, Node-3]. Based on quantum ground state fingerprint changes and three-dimensional temperature characteristics, combined with multi-source information, fault probability analysis is performed to achieve very early warning and refined fault location.

[0029] Update the quantum measurement parameters for a new cycle to perform cyclic measurements, including: Based on the list of key nodes of concern, retrieve the fault diagnosis reports of the corresponding key temperature measurement nodes, filter out high-quality data, and use optimization models (such as lightweight optimization models) to generate quantum measurement parameters to be adjusted and optimized. Prioritize the key temperature measurement nodes according to the warning level to obtain priority labels; set up a new measurement resource allocation table based on the priority labels. The high-quality data screening process involves: first, collecting historically accumulated fault diagnosis reports. Each report must be pre-labeled with a binary label by domain experts or according to established rules (such as completeness, consistency, and matching degree with a typical case library): 1 represents "high-quality data," and 0 represents "low-quality data." Key features are extracted from each report (such as center frequency deviation and spectral linewidth deviation in the deformable sub-ground state fingerprint, maximum probability of fault type, Fd feature vectors, normalized value of abnormal duration, and rationality score of power auxiliary data, etc.) to form feature vectors. These feature vectors are then input along with their corresponding quality labels. The data is fed into a random forest model for training, enabling the model to learn the nonlinear mapping relationship between features and quality levels. After training, the model enters the prediction application stage. When a new fault diagnosis report is input, its feature vector is extracted in the same way and fed into the trained random forest model. The model outputs the probability that the report belongs to the "high quality" category. If the probability exceeds a preset threshold (e.g., 0.85), the core fields and corresponding values ​​(e.g., deformation sub-ground state fingerprint, fault type probability, key feature vectors, etc.) are automatically extracted from the original report as a high-quality data subset for output. If the probability is lower than the preset threshold, it is judged as low-quality data and is not needed. Among them, weights can be set according to existing priority labels, and measurement resources of key temperature measurement nodes corresponding to priority labels can be allocated according to the weights. The new measurement resources include acquisition frequency, measurement cycle, and storage frequency; If Node-3 is a Level 2 warning, Node-4 and Node-5 are Level 3 warnings, and other nodes are normal, the list of nodes to be monitored is [Node-4, Node-5, Node-3]; and the priority labels are [(Node-4, Node-5) > Node-3 > (Node-1, Node-2, Node-6)]; then (Node-4, Node-5) has a greater weight than Node-3; Node-3 has a greater weight than (Node-1, Node-2, Node-6); if the total collection frequency is set to 100 in the new measurement resource allocation table, then the collection frequency allocated to (Node-4, Node-5) is greater than the collection frequency allocated to Node-3, and the collection frequency allocated to Node-3 is greater than the collection frequency allocated to (Node-1, Node-2, Node-6); For the quantum measurement parameters to be adjusted and optimized, the new measurement resource allocation table, and the optimized quantum measurement parameters of the nearest normal critical temperature measurement node (representing the closest normal critical temperature measurement node) of the critical temperature measurement node for the early warning (if there is no healthy node, the historical best optimized quantum measurement parameters are retrieved), the optimization model is used to generate new periodic optimized quantum measurement parameters; and a new round of measurement cycle is performed.

[0030] Among them, the optimization of quantum measurement parameters is aimed at G key temperature measurement nodes, the quantum measurement parameters to be adjusted and optimized are aimed at key temperature measurement nodes for anomaly warning, the new measurement resource allocation table is aimed at G key temperature measurement nodes, and the new cycle of quantum measurement parameter optimization is also aimed at G key temperature measurement nodes. Based on key monitoring nodes and early warning levels, the quantum measurement parameters and measurement resources are adaptively adjusted and optimized (the original measurement resources are allocated evenly; for example, if there are 5 key temperature measurement nodes and the total acquisition frequency is 100, then the acquisition frequency of each of the five key temperature measurement nodes is 20), to achieve closed-loop optimization and ensure the long-term efficient and intelligent operation of the temperature measurement system.

[0031] like Fig. 3 As shown, a deep learning-based multi-node quantum precision temperature measurement method for power equipment is presented, comprising the following steps: Step S1: Collect real-time magnetic field noise data and power-assisted data; deploy key temperature measurement nodes and NV color center sensors, extract photodetector magnetic resonance spectral lines, and construct quantum ground state fingerprints; construct a benchmark fingerprint library using the dynamic benchmark freeform method; and configure and optimize quantum measurement parameters by fusing quantum ground state fingerprints and real-time magnetic field noise data. Step S2: Distribute entangled particles using the central control unit, obtain fluorescence timing signals and total original phase conversion values ​​of key temperature measurement nodes, identify and eliminate common-mode noise based on the total original phase conversion values, and extract a pure quantum phase change dataset. Step S3: Construct spatiotemporal features based on fluorescence time-series signals from key temperature measurement nodes, and invert instantaneous temperature values ​​by integrating power-assisted data. Combine this with geographic coordinates and use the structural region constraint method to generate a three-dimensional temperature field distribution map. Step S4: Perform microscopic and macroscopic labeling to generate a fault diagnosis report and a list of key nodes; update the optimized quantum measurement parameters to obtain new periodic optimized quantum measurement parameters for cyclic measurement.

[0032] In this embodiment, by fusing quantum ground state fingerprints and real-time magnetic field noise data, quantum measurement parameters are dynamically configured and optimized. Combined with entangled particle distribution and common-mode noise discrimination and elimination techniques under synchronous clock, external interference such as environmental magnetic fields is effectively eliminated, and pure quantum phase changes are extracted, improving the sensitivity and accuracy of temperature measurement to the level of quantum precision measurement. At the same time, a three-dimensional temperature field distribution map is generated using spatiotemporal feature splicing and structural region constraint methods, breaking through the limitations of traditional single-point measurement. It can intuitively and precisely reflect the internal temperature gradient and thermal distribution state of the equipment, providing highly reliable data support for the early detection of potential faults such as local overheating. It achieves full-process automation and intelligence from quantum sensing to intelligent diagnosis. On the one hand, it constructs a benchmark fingerprint database through dynamic benchmark liberalization and performs microscopic marking based on real-time quantum ground state fingerprint comparison. Combined with macroscopic marking by three-dimensional temperature field analysis, it realizes multi-dimensional and multi-level intelligent evaluation of equipment operating status, accurately identifies fault types and classifies warning levels, and automatically generates fault diagnosis reports and key node lists. On the other hand, it has closed-loop self-optimization capability. It can dynamically adjust the optimized quantum measurement parameters and measurement resources for the new cycle based on the diagnosis results and resource allocation priorities of the previous cycle, realizing the adaptive evolution of measurement strategy. This ensures high-precision monitoring of key areas while improving the overall efficiency of measurement resource utilization and the long-term operational reliability of the system.

[0033] This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform the deep learning-based multi-node quantum precision temperature measurement system and method for power devices as described above.

[0034] The methods and systems according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store the deep learning-based multi-node quantum precision temperature measurement system and method for power equipment provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components in the electronic device shown in this application may be omitted according to actual needs.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-node quantum precision temperature measurement system for power equipment based on deep learning, characterized in that, The system includes: Fingerprint calibration module: collects real-time magnetic field noise data and power-assisted data; deploys key temperature measurement nodes and NV color center sensors; extracts photodetector magnetic resonance spectral lines through quantum state initialization and microwave frequency sweep signals to construct quantum ground state fingerprints; constructs a benchmark fingerprint library using the dynamic benchmark freeform method; and optimizes quantum measurement parameters by fusing quantum ground state fingerprints and real-time magnetic field noise data. Common-mode noise reduction and extraction module: entangled particles are distributed through the central control unit, and microwave pulses and laser pulses are applied under the synchronous clock to obtain the fluorescence timing signals of key temperature measurement nodes and the total original phase conversion value. Based on the total original phase conversion value, common-mode noise is identified and eliminated, and a pure quantum phase change dataset is extracted. Spatiotemporal temperature field construction module: Spatiotemporal features are constructed based on fluorescence time-series signals from key temperature measurement nodes, and instantaneous temperature values ​​are retrieved by integrating power-assisted data. A three-dimensional temperature field distribution map is generated by combining geographic coordinates with a structural region constraint method. Diagnostic optimization module: Performs microscopic and macroscopic labeling, generates fault diagnosis reports and a list of key nodes; and updates the quantum measurement parameters for the new cycle to perform cyclic measurements.

2. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 1, characterized in that, The deployment of key temperature measurement nodes and NV color center sensors, through quantum state initialization and microwave frequency sweep signal extraction of photodetector magnetic resonance spectral lines, constructs a quantum ground-state fingerprint, including: Deploy G key temperature measurement nodes and G NV color center sensors on power equipment, and obtain a list of geographic coordinates of the G key temperature measurement nodes and a list of G NV color centers; The central control unit sends quantum state initialization commands to G key temperature measurement nodes to obtain G initialized NV color center sensors; Based on G initialized NV color center sensors, a microwave sweep signal is applied to extract G key temperature measurement node photodetector magnetic resonance spectral lines; and G key temperature measurement node quantum ground state fingerprints are constructed.

3. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 2, characterized in that, The construction of the benchmark fingerprint database using the dynamic benchmark freeform method includes: Based on the geographical coordinates of G key temperature measurement nodes, obtain the distance and weight between the a-th key temperature measurement node and the b-th key temperature measurement node, and generate a distance matrix and a weight matrix. Based on the weight matrix, spatial relationship modeling is performed to obtain the predicted quantum ground state fingerprints of G key temperature measurement nodes; Based on the difference between the predicted quantum ground state fingerprint and the quantum ground state fingerprint of G key temperature measurement nodes, biased ground state fingerprint and outlier labeled fingerprint are constructed; A benchmark fingerprint library was constructed based on the geographical coordinates of G key temperature measurement nodes, quantum ground state fingerprints, predicted quantum ground state fingerprints, biased ground state fingerprints, and outlier labeled fingerprints.

4. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 3, characterized in that, The configuration of the fused quantum ground-state fingerprint and real-time magnetic field noise data optimizes quantum measurement parameters, including: Retrieve the status data of power equipment; Based on the quantum ground state fingerprint of G key temperature measurement nodes and the coincidence state data of power equipment, a set of candidate microwave frequencies is formed; Based on real-time magnetic field noise data from G key temperature measurement nodes, the noise level is obtained, and a measurement mode is triggered to obtain a dynamic decoupling pulse sequence. Optimize quantum measurement parameters by combining candidate microwave frequency sets and dynamically decoupled pulse sequences.

5. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 4, characterized in that, The extracted pure quantum phase change dataset includes: The central control unit sends corresponding entangled particles to G key temperature measurement nodes; Based on the global synchronization clock signal, within the Tr time window, microwave pulses and laser pulses are applied to the entangled particles corresponding to the G key temperature measurement nodes according to the optimized quantum measurement parameters, so as to obtain the fluorescence timing signals of the G key temperature measurement nodes within the Tr time window and convert them into the original phase conversion values. Obtain the total raw phase conversion values ​​of G key temperature measurement nodes within the Tr time window; If the total original phase conversion value is not 0, it indicates that there is common-mode noise. The pure quantum phase change dataset is obtained based on the common-mode noise. If the total original phase conversion value is 0, it indicates that there is no common-mode noise. The pure quantum phase change is equal to the original phase conversion value, and the pure quantum phase change dataset is directly generated.

6. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 5, characterized in that, The process of constructing spatiotemporal features based on fluorescence time-series signals from key temperature measurement nodes and fusing power-assisted data to retrieve instantaneous temperature values ​​includes: The fluorescence time series signals of G key temperature measurement nodes are spliced ​​with the fluorescence time series signals of k adjacent nodes to construct a spatiotemporal feature tensor. By introducing power-assisted data to the spatiotemporal feature tensor, the instantaneous temperature values ​​of G key temperature measurement nodes are retrieved.

7. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 6, characterized in that, The method of generating a three-dimensional temperature field distribution map using structural region constraints includes: Based on the geographical coordinates of the G key temperature measurement nodes and prior knowledge of the structure of the power equipment, the G key temperature measurement nodes are divided into F regions. Establish a combination of regional temperature field constraints based on the F regions; Using the instantaneous temperature values ​​and geographic coordinates of G key temperature measurement nodes as input, and combining regional temperature field constraints, the instantaneous temperature values ​​and geographic coordinates of HH temperature measurement points are obtained through numerical methods, and a three-dimensional temperature field distribution map is generated by combining the instantaneous temperature values ​​and geographic coordinates of G key temperature measurement nodes. Where HH represents the total number of temperature measurement points in the F regions.

8. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 7, characterized in that, The process involves performing microscopic and macroscopic labeling to generate a fault diagnosis report and a list of key nodes, including: For G key temperature measurement nodes, based on optimized quantum measurement parameters, microwave frequency sweep signal is executed to obtain the real-time optical detection magnetic resonance spectrum of the G key temperature measurement nodes, and fitting is performed to generate the real-time quantum ground state fingerprint of the G key temperature measurement nodes. The real-time quantum ground state fingerprints of the G key temperature measurement nodes are compared with the corresponding quantum ground state fingerprints in the benchmark fingerprint library to obtain the deformable quantum ground state fingerprints of the G key temperature measurement nodes. Microscopic labeling is performed by comparing the ground state fingerprints of the deformation sub-states of G nodes with a set anomaly type threshold; Using the three-dimensional temperature field distribution map at time point Te as input, Fd feature vectors are obtained; the Fd feature vectors are compared with the known fault mode library for macroscopic labeling. The deformation sub-ground state fingerprint and Fd feature vectors corresponding to the key temperature measurement nodes with abnormal micro and macro markings are fused with multi-source information to obtain the normal state probability and JH fault type probabilities of the key temperature measurement node; and the warning level is classified according to the JH fault type probabilities. For key temperature monitoring nodes corresponding to the warning level, generate fault diagnosis reports and a list of nodes to focus on.

9. The deep learning-based multi-node quantum precision temperature measurement system for power equipment according to claim 8, characterized in that, The update of the new cycle to optimize the quantum measurement parameters for performing cyclic measurements includes: Based on the list of key nodes of concern, the fault diagnosis reports of the corresponding key temperature measurement nodes are retrieved to obtain the quantum measurement parameters that need to be adjusted and optimized. Prioritize the key temperature measurement nodes according to the warning level to obtain priority labels; set up a new measurement resource allocation table based on the priority labels. For the quantum measurement parameters to be adjusted and optimized, the new measurement resource allocation table, and the optimized quantum measurement parameters of the nearest normal critical temperature measurement node of the early warning critical temperature measurement node, a new cycle of optimized quantum measurement parameters is generated; and a new round of measurement cycle is carried out.

10. A multi-node quantum precision temperature measurement method for power equipment based on deep learning, characterized in that, The method includes: Step S1: Collect real-time magnetic field noise data and power-assisted data; deploy key temperature measurement nodes and NV color center sensors, extract photodetector magnetic resonance spectral lines, and construct quantum ground state fingerprints; construct a benchmark fingerprint library using the dynamic benchmark freeform method; and configure and optimize quantum measurement parameters by fusing quantum ground state fingerprints and real-time magnetic field noise data. Step S2: Distribute entangled particles using the central control unit, obtain fluorescence timing signals and total original phase conversion values ​​of key temperature measurement nodes, identify and eliminate common-mode noise based on the total original phase conversion values, and extract a pure quantum phase change dataset. Step S3: Construct spatiotemporal features based on fluorescence time-series signals from key temperature measurement nodes, and invert instantaneous temperature values ​​by integrating power-assisted data. Combine this with geographic coordinates and use the structural region constraint method to generate a three-dimensional temperature field distribution map. Step S4: Perform microscopic and macroscopic labeling to generate a fault diagnosis report and a list of key nodes; update the optimized quantum measurement parameters to obtain new periodic optimized quantum measurement parameters for cyclic measurement.

Images