Deep learning switching-on safety risk assessment system fusing multi-source information and method of deep learning switching-on safety risk assessment system

Abstract

The invention discloses a deep learning switching-on safety risk assessment system and method fused with multi-source information, and relates to the technical field of switching-on safety risk assessment, the deep learning switching-on safety risk assessment system fused with the multi-source information and the method thereof realize accurate assessment of the switching-on safety risk of a circuit breaker through fusion of quantum and classical technologies, and improve the safety of the circuit breaker. Various beneficial effects are achieved; in the aspect of data acquisition, a quantum sensing array acquires microscopic quantum data in equipment, a classical module synchronizes SCADA system data, a multi-dimensional acquisition system is formed, and comprehensive information is provided; during data processing, a quantum-classical hybrid processor combines quantum convolution and a space-time attention mechanism to improve feature extraction efficiency and precision; in risk prevention and control, a risk decision engine scientifically and accurately evaluates the risk, an adaptive locking controller is quick in response, and the fault probability is reduced; meanwhile, the system guarantees stable and reliable operation through quantum noise compensation, high-fidelity quantum state transmission, data synchronization and other measures.

Description

Technical Field

[0001] The present invention relates to the technical field of closing safety risk assessment, and in particular to a deep learning closing safety risk assessment system and method integrating multi-source information. Background Art

[0002] The existing deep learning closing safety risk assessment system and method that integrates multi-source information still have the following drawbacks in actual use:

[0003] Traditional LSTM / CNN models face the problem of feature dimensionality explosion when processing multi-source heterogeneous data (such as partial discharge ultrasonic signals + infrared thermal imaging + quantum sensor data); a typical substation needs to complete risk assessment of 10^6 dimensional data within 3ms, and classical computers are difficult to meet real-time requirements; and existing systems fail to effectively integrate the microscopic parameters collected by quantum chips (such as magnetic field quantum state measurement data based on NV color centers), and only convert them into classical signal processing, losing the early fault characteristics contained in the quantum entangled state; conventional deep learning is limited by the thermal equilibrium assumption of the Boltzmann machine when predicting the probability of arc reignition, and the prediction accuracy of transient breakdown events caused by quantum tunneling effects (probability <10^-5) is low. Summary of the Invention

[0004] The purpose of the present invention is to provide a deep learning closing safety risk assessment system and method that integrates multi-source information to solve the above-mentioned problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a deep learning closing safety risk assessment system integrating multi-source information, comprising:

[0006] A quantum sensing array configured to collect microscopic magnetic field quantum state data of the circuit breaker operating mechanism through nitrogen vacancy color center quantum sensors and to obtain quantum tunneling current signals of the contact gap through a superconducting quantum interference device;

[0007] A classic data acquisition module configured to synchronously acquire three-phase voltage waveforms, closing coil currents, and ambient temperature and humidity parameters from a SCADA system;

[0008] A quantum-classical hybrid processor, comprising:

[0009] A quantum feature extraction unit configured to encode the microscopic magnetic field quantum state data into a quantum bit sequence , and realize quantum convolution operation through parameterized quantum circuits;

[0010] a classical feature fusion unit configured to perform cross-modal feature weighting on the quantum tunneling current signal and the three-phase voltage waveform using a spatiotemporal attention mechanism;

[0011] A risk decision engine is configured to generate a closing and locking control instruction based on the risk probability distribution output by the variational quantum eigensolver.

[0012] Furthermore, the quantum feature extraction unit includes:

[0013] A quantum encoding subunit is configured to map the microscopic magnetic field quantum state data to the amplitude component of the quantum state through amplitude encoding, and its mapping function is:

[0014] ;

[0015] A quantum convolution subunit is configured to execute a quantum circuit comprising 15 parameterized quantum gates, wherein each layer of quantum gates comprises an Ry rotation gate and a controlled Z gate.

[0016] Furthermore, the classical feature fusion unit includes:

[0017] A quantum data canonicalization module is configured to perform a projection measurement on the quantum bit sequence to obtain a classical probability distribution vector ;

[0018] A cross-modal attention module is configured to calculate the cross-covariance matrix of the quantum tunneling current signal and the three-phase voltage waveform, and generate feature weights through a Softmax function.

[0019] Furthermore, the risk decision engine includes:

[0020] A quantum probability generator configured to calculate a probability of a fault occurring by quantum Monte Carlo sampling , where the measurement operator ;

[0021] The adaptive locking controller is configured to When the lock signal is sent to the circuit breaker through the GOOSE message.

[0022] Furthermore, it also includes:

[0023] A quantum noise compensation module configured to model quantum decoherence effects based on the Lindblad master equation: ,in is the decoherence channel operator.

[0024] Furthermore, quantum state transmission is achieved between the quantum sensing array and the quantum-classical hybrid processor through a low-temperature microwave resonant cavity, with a transmission fidelity of ≥99.5%.

[0025] Furthermore, the data synchronization error between the classic data acquisition module and the SCADA system is less than 10 μs.

[0026] Furthermore, the response delay of the risk decision engine is less than 50 μs.

[0027] A method for evaluating the risk of circuit breaker closing safety based on the system comprises the following steps:

[0028] S1. Quantum state data acquisition:

[0029] The quantum sensor array is used to collect the microscopic magnetic field quantum state data of the circuit breaker operating mechanism, and the microscopic magnetic field quantum state data includes spin relaxation time and dephasing time ;

[0030] The three-phase voltage waveform is obtained by the classic data acquisition module and closing coil current ;

[0031] S2. Quantum-classical hybrid processing:

[0032] The microscopic magnetic field quantum state data is input into the quantum feature extraction unit, and the frequency domain features are extracted by quantum Fourier transform. ,in is the frequency measurement operator;

[0033] The quantum tunneling current signal and the three-phase voltage waveform are input into the classical feature fusion unit to calculate the cross-modal attention weight matrix ,in are the query vector and key vector of quantum signal and classical signal respectively;

[0034] S3. Risk decision generation:

[0035] The closing risk value is calculated by the quantum probability generator ,in is the quantum probability of the kth failure mode;

[0036] when When the self-adaptive locking controller is triggered to output a locking instruction.

[0037] Furthermore, the quantum circuit depth of the quantum Fourier transform in step S2 is , where n is the number of quantum bits.

[0038] Compared with the existing technology, the deep learning closing safety risk assessment system and method provided by the present invention that integrates multi-source information has the following beneficial effects:

[0039] This deep learning closing safety risk assessment system and method that integrates multi-source information achieves accurate assessment of circuit breaker closing safety risks through the integration of quantum and classical technologies, with multiple beneficial effects. In data acquisition, the quantum sensor array obtains microscopic quantum data inside the equipment, and the classical module synchronizes the SCADA system data to form a multi-dimensional acquisition system and provide comprehensive information. In data processing, the quantum-classical hybrid processor combines quantum convolution with the spatiotemporal attention mechanism to improve the efficiency and accuracy of feature extraction. In risk prevention and control, the risk decision engine scientifically and accurately assesses risks, and the adaptive locking controller responds quickly to reduce the probability of failure. At the same time, the system ensures stable and reliable operation through measures such as quantum noise compensation, high-fidelity quantum state transmission and data synchronization, providing support for the safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0041] Figure 1 Schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0043] See also Figure 1 , a deep learning closing safety risk assessment system integrating multi-source information, including:

[0044] A quantum sensing array configured to collect microscopic magnetic field quantum state data of the circuit breaker operating mechanism through nitrogen vacancy color center quantum sensors and to obtain quantum tunneling current signals of the contact gap through a superconducting quantum interference device;

[0045] Classic data acquisition module, the classic data acquisition module is configured to synchronously obtain the three-phase voltage waveform, closing coil current and ambient temperature and humidity parameters in the SCADA system;

[0046] Quantum-classical hybrid processor, quantum-classical hybrid processor includes:

[0047] A quantum feature extraction unit configured to encode microscopic magnetic field quantum state data into a quantum bit sequence , and realize quantum convolution operation through parameterized quantum circuits;

[0048] The classical feature fusion unit is configured to perform cross-modal feature weighting on the quantum tunneling current signal and the three-phase voltage waveform using a spatiotemporal attention mechanism;

[0049] The risk decision engine is configured to generate a closing and locking control instruction based on the risk probability distribution output by the variational quantum eigensolver.

[0050] Hardware Configuration

[0051] A diamond nitrogen vacancy (NV color center) quantum sensor is used, and the operating temperature is controlled in a low-temperature environment of 4K. The magnetic field signal is focused by a confocal microscope system, and the sampling frequency is set to 10MHz to capture the 10nT level magnetic field changes inside the operating mechanism.

[0052] The superconducting quantum interference device (SQUID) uses a niobium-titanium alloy Josephson junction, combined with a liquid helium cooling system (temperature ≤ 4.2K), to perform real-time sampling of the quantum tunneling current in the contact gap (resolution up to 1pA) at a sampling rate of 1GS / s.

[0053] The signal transmission quantum sensing array and hybrid processor are connected through a low-temperature microwave resonant cavity (operating frequency 6GHz). The inner wall of the cavity is coated with a superconducting niobium film and uses quantum state compression coding technology to ensure transmission fidelity ≥ 99.5%.

[0054] SCADA system connection

[0055] It interacts with the substation SCADA system through the OPCUA protocol, adopts a hardware clock synchronization module (based on the IEEE 1588 protocol), configures a master-slave clock architecture, and uses PTP message calibration to achieve data synchronization of three-phase voltage (sampling rate 10kHz) and closing coil current (sampling rate 1kHz) with an error of ≤10μs.

[0056] The ambient temperature and humidity sensor uses an RS485 interface type with a sampling interval of 100ms and is connected to the acquisition module via the Modbus protocol.

[0057] The quantum feature extraction unit includes:

[0058] The quantum encoding subunit is configured to map the microscopic magnetic field quantum state data to the amplitude component of the quantum state through amplitude encoding. The mapping function is:

[0059] ;

[0060] The quantum convolution subunit is configured to execute a quantum circuit containing 15 parameterized quantum gates, where each layer of quantum gates contains Ry rotation gates and controlled Z gates.

[0061] After being compiled through a quantum cloud computing platform (such as IBM Q Experience), it is executed on a 9-bit superconducting quantum processor with a circuit depth of O(n²).

[0062] The classic feature fusion unit includes:

[0063] Quantum data classicization module, the quantum data classicization module configuration performs projection measurement on the quantum bit sequence to obtain the classical probability distribution vector ;

[0064] Cross-modal attention module,The cross-modal attention module is configured to calculate the cross-covariance matrix of the,quantum tunneling current signal and the three-phase voltage waveform.

[0065] The cross-modal attention calculation defines the quantum tunneling current signal X∈R^(T×D1) and the voltage waveform Y∈R^(T×D2), and the cross-covariance matrix is ​​calculated as: ;

[0066] The weight matrix W∈R^(D1×D2) is generated through Softmax to achieve feature weighted fusion.

[0067] The risk decision engine includes:

[0068] A quantum probability generator configured to calculate the probability of a fault occurring by quantum Monte Carlo sampling , where the measurement operator ;

[0069] Adaptive blocking controller, the adaptive blocking controller is configured as When the lock signal is sent to the circuit breaker through the GOOSE message.

[0070] Also includes:

[0071] The quantum noise compensation module is configured to model quantum decoherence effects based on the Lindblad master equation: ,in is the decoherence channel operator.

[0072] Quantum state transmission is achieved between the quantum sensing array and the quantum-classical hybrid processor through a low-temperature microwave resonant cavity, with a transmission fidelity of ≥99.5%.

[0073] The data synchronization error between the classic data acquisition module and the SCADA system is less than 10μs.

[0074] The response delay of the risk decision engine is less than 50μs.

[0075] A system-based closing safety risk assessment method includes the following steps:

[0076] S1. Quantum state data acquisition:

[0077] The quantum sensor array is used to collect the microscopic magnetic field quantum state data of the circuit breaker operating mechanism, including the spin relaxation time. and dephasing time ;

[0078] Acquire three-phase voltage waveforms through classic data acquisition modules and closing coil current ;

[0079] S2. Quantum-classical hybrid processing:

[0080] The microscopic magnetic field quantum state data is input into the quantum feature extraction unit, and the frequency domain features are extracted through quantum Fourier transform. ,in is the frequency measurement operator;

[0081] The quantum tunneling current signal and the three-phase voltage waveform are input into the classical feature fusion unit to calculate the cross-modal attention weight matrix ,in are the query vector and key vector of quantum signal and classical signal respectively;

[0082] S3. Risk decision generation:

[0083] Calculate the closing risk value through quantum probability generator ,in is the quantum probability of the kth failure mode;

[0084] when When , the adaptive locking controller is triggered to output the locking command.

[0085] The quantum circuit depth of the quantum Fourier transform in step S2 is , where n is the number of quantum bits.

[0086] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A deep learning closing safety risk assessment system integrating multi-source information, characterized by: include: A quantum sensing array configured to collect microscopic magnetic field quantum state data of the circuit breaker operating mechanism through nitrogen vacancy color center quantum sensors and to obtain quantum tunneling current signals of the contact gap through a superconducting quantum interference device; A classic data acquisition module configured to synchronously acquire three-phase voltage waveforms, closing coil currents, and ambient temperature and humidity parameters from a SCADA system; A quantum-classical hybrid processor, comprising: A quantum feature extraction unit configured to encode the microscopic magnetic field quantum state data into a quantum bit sequence , and realize quantum convolution operation through parameterized quantum circuits; a classical feature fusion unit configured to perform cross-modal feature weighting on the quantum tunneling current signal and the three-phase voltage waveform using a spatiotemporal attention mechanism; A risk decision engine is configured to generate a closing and locking control instruction based on the risk probability distribution output by the variational quantum eigensolver.

2. The deep learning closing safety risk assessment system integrating multi-source information according to claim 1 is characterized in that: The quantum feature extraction unit comprises: A quantum encoding subunit is configured to map the microscopic magnetic field quantum state data to the amplitude component of the quantum state through amplitude encoding, and its mapping function is: ； A quantum convolution subunit is configured to execute a quantum circuit comprising 15 parameterized quantum gates, wherein each layer of quantum gates comprises an Ry rotation gate and a controlled Z gate.

3. The deep learning closing safety risk assessment system integrating multi-source information according to claim 1 is characterized in that: The classic feature fusion unit includes: A quantum data canonicalization module is configured to perform a projection measurement on the quantum bit sequence to obtain a classical probability distribution vector ; A cross-modal attention module is configured to calculate the cross-covariance matrix of the quantum tunneling current signal and the three-phase voltage waveform, and generate feature weights through a Softmax function.

4. The deep learning closing safety risk assessment system integrating multi-source information according to claim 1 is characterized in that: The risk decision engine includes: A quantum probability generator configured to calculate a probability of a fault occurring by quantum Monte Carlo sampling , where the measurement operator ; The adaptive locking controller is configured to When the lock signal is sent to the circuit breaker through the GOOSE message.

5. The deep learning closing safety risk assessment system integrating multi-source information according to claim 1 is characterized in that: Also includes: A quantum noise compensation module configured to model quantum decoherence effects based on the Lindblad master equation: ,in is the decoherence channel operator.

6. The deep learning closing safety risk assessment system integrating multi-source information according to claim 5 is characterized in that: Quantum state transmission is achieved between the quantum sensing array and the quantum-classical hybrid processor through a low-temperature microwave resonant cavity, with a transmission fidelity of ≥99.5%.

7. The deep learning closing safety risk assessment system integrating multi-source information according to claim 6 is characterized in that: The data synchronization error between the classic data acquisition module and the SCADA system is less than 10 μs.

8. The deep learning closing safety risk assessment system integrating multi-source information according to claim 7 is characterized in that: The response delay of the risk decision engine is less than 50 μs.

9. A method for evaluating the closing safety risk based on the system according to any one of claims 1 to 8, characterized in that: The following steps are involved: S1. Quantum state data acquisition: The quantum sensor array is used to collect the microscopic magnetic field quantum state data of the circuit breaker operating mechanism, and the microscopic magnetic field quantum state data includes spin relaxation time and dephasing time ; The three-phase voltage waveform is obtained by the classic data acquisition module and closing coil current ; S2. Quantum-classical hybrid processing: The microscopic magnetic field quantum state data is input into the quantum feature extraction unit, and the frequency domain features are extracted by quantum Fourier transform. ,in is the frequency measurement operator; The quantum tunneling current signal and the three-phase voltage waveform are input into the classical feature fusion unit to calculate the cross-modal attention weight matrix ,in are the query vector and key vector of quantum signal and classical signal respectively; S3. Risk decision generation: The closing risk value is calculated by the quantum probability generator ,in is the quantum probability of the kth failure mode; when When the self-adaptive locking controller is triggered to output a locking instruction.

10. The method according to claim 9, characterized in that The quantum circuit depth of the quantum Fourier transform in step S2 is , where n is the number of quantum bits.

Images