Multi-energy balance margin adaptive regulation and control method and device

Abstract

The invention discloses a multi-energy balance margin adaptive regulation and control method and device, and relates to the technical field of quantum computing crossover, and the method comprises the steps: carrying out the optimization of a four-dimensional input matrix through a quantum encoder, executing quantum annealing calculation to generate a ground state solution, integrating an inertia demand increment with the ground state solution through quantum bit entanglement, and carrying out the optimization of a four-dimensional input matrix; calculating a comprehensive inertia gap value, and generating a hierarchical scheduling instruction by adopting a dynamic threshold hierarchical strategy; and according to the hierarchical scheduling instruction, a topology reconstruction instruction is generated through a virtual inertia dynamic algorithm, dynamic monitoring is performed by applying multi-agent collaborative game optimization, and an intelligent regulation and control report is generated. After three-dimensional fault features and space-time dimensions of a multi-energy fluctuation ratio are aligned based on a dynamic time warping-covariance matching method, dynamic fusion is carried out on standard field phase fiber bundles by adopting aged-Simons topology weaving, and self-adaptive regulation and control of a multi-energy balance margin are realized.

Description

Technical Field

[0001] This invention relates to the field of quantum computing interdisciplinary technology, and in particular to a method and apparatus for adaptive control of multi-energy balance margin. Background Technology

[0002] Against the backdrop of the rapid development of multi-energy internet and multi-energy complementary systems, the monitoring of the operating status of power equipment and the coordinated control of inertia have become key technologies for ensuring system stability. Existing technologies typically employ acoustic signature analysis for mechanical fault diagnosis, with mainstream methods including acoustic signature spectral feature extraction combined with deep learning models for status identification. Meanwhile, inertia control generally relies on real-time power fluctuation data, using classical control theory or data-driven models to predict inertia demand and generate scheduling commands. The signal processing flow, control response delay, and secure data transmission of these methods have been clearly defined and regulated. These technologies have been widely applied in equipment health management systems for wind power, nuclear power, and other scenarios, forming a relatively complete technical path of "signal acquisition - feature analysis - control decision-making."

[0003] Existing methods have room for improvement in dynamic fusion of multimodal features and quantum collaborative optimization: First, the alignment of acoustic and energy features usually adopts covariance matching with a fixed time window, which has not fully considered the time non-stationary characteristics of rotating machinery under varying operating conditions, resulting in a hysteresis deviation between the inertia increment prediction and the actual equipment state transition; Second, the energy conservation constraint modeling of the Hamiltonian of the mechanical system in the inertia ground state solution of quantum annealing is not perfect. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multi-energy balance margin adaptive control method to solve the problem of improving the modeling of inertia prediction hysteresis bias and energy conservation constraints.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a multi-energy balance margin adaptive control method, which includes,

[0008] The raw voiceprint signal is acquired in real time and preprocessed. A convolutional pulse neural network is used to dynamically analyze the preprocessed raw voiceprint signal to generate voiceprint spectral features. Finally, a three-dimensional fault feature vector is generated by a CNN-SNN fusion algorithm.

[0009] Based on the three-dimensional fault feature vector, a dynamic quantization vector is generated by feature matching through a multimodal feature fusion algorithm, and a dynamic weight allocation strategy is adopted to predict the inertia demand increment.

[0010] The three-dimensional fault feature vector generated earlier is called, and the original energy parameters of the multi-energy system (such as wind power, photovoltaic output, hydro current, etc.) are collected. The volatility of the multi-energy system is calculated by time series analysis algorithm. Then, the three-dimensional fault feature vector and the volatility of the multi-energy system are spatiotemporally aligned by dynamic time warping-covariance matching method. At the same time, a four-dimensional input matrix is ​​constructed by phase space manifold weaving engine.

[0011] The four-dimensional input matrix is ​​optimized by a quantum encoder, and a quantum annealing calculation is performed to generate a ground state solution. The inertia demand increment is integrated with the ground state solution through qubit entanglement, the comprehensive inertia gap value is calculated, and a hierarchical scheduling instruction is generated by a dynamic threshold hierarchical strategy.

[0012] Based on the hierarchical scheduling instructions, a topology reconfiguration instruction is generated through a virtual inertia dynamic algorithm, and dynamic monitoring is performed using multi-agent collaborative game optimization to generate an intelligent control report.

[0013] As a preferred embodiment of the multi-energy balance margin adaptive control method of the present invention, the following steps are taken: real-time acquisition and preprocessing of the original acoustic signature signal; dynamic analysis of the preprocessed original acoustic signature signal using a convolutional pulse neural network to generate acoustic signature spectral features; and generation of a three-dimensional fault feature vector using a CNN-SNN fusion algorithm.

[0014] The original acoustic signature spectrum includes fundamental harmonic groups, high-frequency resonant peaks, modulation sidebands, broadband impulse energy, and acoustic signature chromaticity map.

[0015] The preprocessing includes dynamic noise suppression, key frequency band extraction, anti-aliasing sampling, wavelet packet denoising, and signal-to-noise ratio;

[0016] The acoustic signature spectral features are extracted from the original acoustic signature spectrum by multi-scale power spectrum integration and intelligent resonant band tracking using an acoustic signature physical feature decoupling device.

[0017] Fault fingerprints are obtained by dynamically analyzing acoustic signature spectral features using a convolutional pulse neural network, and a three-dimensional fault feature vector is generated.

[0018] As a preferred embodiment of the multi-energy balance margin adaptive control method of the present invention, the specific steps for generating a dynamic quantization vector by feature matching through a multi-modal feature fusion algorithm based on the three-dimensional fault feature vector are as follows:

[0019] The three-dimensional fault feature vector is aligned with the preset working condition fingerprint database through a multimodal feature fusion algorithm. The optimal working condition mode is dynamically matched by the Mahalanobis distance-temporal correlation dual criteria in the degradation representation domain to generate a quantized encoding vector.

[0020] The impact and fatigue components are assigned based on the material fatigue coefficient, and the fusion weights are obtained by weighted dot product. The fused feature vector is then integrated with the quantization encoding vector to generate a dynamic quantization vector.

[0021] As a preferred embodiment of the multi-energy balance margin adaptive control method of the present invention, the specific steps of using a dynamic weight allocation strategy to predict the inertia demand increment are as follows:

[0022] Based on the dynamic quantization vector, the inertia control coefficient matrix for each dimension of inertia demand is calculated using a dynamic weight allocation strategy.

[0023] An inertia demand prediction function is constructed by using a deep reinforcement learning model, and an embedded adversarial mechanism is used to dynamically and adversarially optimize the inertia control coefficient matrix to predict the inertia demand increment.

[0024] As a preferred embodiment of the multi-energy balance margin adaptive control method of the present invention, the following steps are taken: First, a previously generated three-dimensional fault feature vector is invoked, and simultaneously, the original energy parameters of the multi-energy system (such as wind power, photovoltaic output, hydropower, etc.) are collected. The multi-energy volatility is then calculated using a time-series analysis algorithm. Subsequently, the three-dimensional fault feature vector and the multi-energy volatility are spatiotemporally aligned using a dynamic time warping-covariance matching method. Simultaneously, a four-dimensional input matrix is ​​constructed using a phase space manifold weaving engine. The specific steps are as follows:

[0025] Three-dimensional fault features and multi-energy volatility are extracted by dynamic time warping window, and timestamp alignment is performed by dynamic time warping-covariance matching method to generate three-dimensional fault feature trajectory.

[0026] The three-dimensional fault feature trajectory is constructed into a Cantor differential manifold basis using a phase space manifold weaving engine. At the same time, the Cantor differential manifold basis is decomposed by a gauge field transformation to generate a gauge field phase fiber bundle.

[0027] Dynamically fuse gauge field phase fiber bundles through Chen-Simons topological weaving to generate a four-dimensional phase space structure, and then use qubit logic topological mapping to transform it into a quantum processor.

[0028] The quantum processor is analyzed by quantum annealing eigenmodes to generate a four-dimensional input matrix.

[0029] As a preferred embodiment of the multi-energy balance margin adaptive control method of the present invention, the following steps are taken: The four-dimensional input matrix is ​​optimized using a quantum encoder; a quantum annealing calculation is performed to generate a ground-state solution; the inertia demand increment is integrated with the ground-state solution through qubit entanglement; the comprehensive inertia gap value is calculated; and a dynamic threshold hierarchical strategy is used to generate hierarchical scheduling instructions.

[0030] The four-dimensional input matrix is ​​mapped to the qubit space by a quantum encoder for optimization, and quantum annealing calculation is performed by a variable quantum feature solver to generate the ground state solution.

[0031] Based on the ground state solution, an incremental-ground state coupling field is constructed through qubit entanglement, and the inertia demand increment is integrated with the ground state solution under Pauli Z-gate constraints to generate the inertia gap trajectory function;

[0032] Based on the inertia gap trajectory function, the comprehensive inertia gap value is calculated by iteratively solving the Schrödinger-Poisson equation, and hierarchical scheduling instructions are generated through a dynamic threshold hierarchical strategy.

[0033] As a preferred embodiment of the multi-energy balance margin adaptive control method of the present invention, the steps of generating topology reconfiguration instructions based on hierarchical scheduling instructions using a virtual inertia dynamic algorithm, and dynamically monitoring using multi-agent cooperative game optimization to generate an intelligent control report are as follows.

[0034] Based on the hierarchical scheduling instructions, the vulnerable nodes in the power grid topology are analyzed using a virtual inertia dynamic algorithm to generate topology reconfiguration instructions.

[0035] Based on the topology reconstruction instructions, a fused topology record is generated through a distributed reinforcement learning control unit, and dynamic monitoring is performed using multi-agent collaborative game optimization to generate an intelligent control report.

[0036] Secondly, the present invention provides a multi-energy balance margin adaptive control device, comprising,

[0037] The data acquisition module is used to acquire raw voiceprint signals in real time and perform preprocessing. It uses a convolutional pulse neural network to dynamically analyze the preprocessed raw voiceprint signals to generate voiceprint spectral features, and generates a three-dimensional fault feature vector through a CNN-SNN fusion algorithm.

[0038] The inertia demand calculation module is used to generate a dynamic quantization vector by performing feature matching through a multimodal feature fusion algorithm based on the three-dimensional fault feature vector, and to predict the inertia demand increment by adopting a dynamic weight allocation strategy.

[0039] A four-dimensional input matrix module is constructed to call the previously generated three-dimensional fault feature vector and collect the original energy parameters of the multi-energy system (such as wind power, photovoltaic output, hydropower, etc.). The volatility of the multi-energy system is calculated through time series analysis algorithm. Then, the three-dimensional fault feature vector and the volatility of the multi-energy system are spatiotemporally aligned by dynamic time warping-covariance matching method. At the same time, a phase space manifold weaving engine is used to construct the four-dimensional input matrix.

[0040] The hierarchical scheduling instruction generation module is used to optimize the four-dimensional input matrix through the quantum encoder, perform quantum annealing calculation to generate the ground state solution, integrate the inertia demand increment with the ground state solution through qubit entanglement, calculate the comprehensive inertia gap value, and generate hierarchical scheduling instructions using a dynamic threshold hierarchical strategy.

[0041] The module for generating control reports is used to generate topology reconfiguration instructions based on hierarchical scheduling instructions through a virtual inertia dynamic algorithm, and to perform dynamic monitoring using multi-agent collaborative game optimization to generate intelligent control reports.

[0042] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the multi-energy balance margin adaptive control method as described in the first aspect of the present invention.

[0043] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the multi-energy balance margin adaptive control method as described in the first aspect of the present invention.

[0044] The beneficial effects of this invention are as follows: An incremental-ground state coupled field is constructed through qubit entanglement, and the incremental inertia requirement is physically integrated with the ground state solution under Pauli Z-gate constraints, thus achieving a mandatory guarantee of mechanical energy conservation. Furthermore, based on the dynamic time warping-covariance matching method, after aligning the spatiotemporal dimensions of three-dimensional fault characteristics and multi-energy volatility, Chern-Simons topological weaving is used to dynamically fuse the gauge field phase fiber bundle, achieving adaptive control of the multi-energy balance margin. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 The flowchart shows the adaptive control method for multi-energy balance margin.

[0047] Figure 2 This is a schematic diagram of a multi-energy balance margin adaptive control device.

[0048] Figure 3 A flowchart for constructing a four-dimensional input matrix.

[0049] Figure 4 A flowchart for generating hierarchical scheduling instructions. Detailed Implementation

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0053] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a multi-energy balance margin adaptive control method, comprising the following steps:

[0054] S1. Real-time acquisition of raw voiceprint signals and preprocessing, dynamic analysis of preprocessed raw voiceprint signals using convolutional pulse neural networks to generate voiceprint spectral features, and generation of three-dimensional fault feature vectors through CNN-SNN fusion algorithm.

[0055] The original voiceprint signal includes fundamental harmonic groups, high-frequency resonant peaks, modulation sidebands, broadband impulse energy, and voiceprint chromaticity map.

[0056] Preprocessing includes dynamic noise suppression, key frequency band extraction, anti-aliasing sampling, wavelet packet denoising, and signal-to-noise ratio;

[0057] The acoustic signature physical feature decoupler is used to extract the acoustic signature spectral features from the original acoustic signature spectrum by multi-scale power spectrum integration and intelligent resonant band tracking.

[0058] It should be noted that the acoustic fingerprint physical feature decoupler performs multi-scale power spectrum integration on the acoustic fingerprint spectrum features to quantify the five-dimensional features of fundamental harmonic group distortion rate, high-frequency resonant peak offset, modulation sideband energy ratio, broadband impulse band integral intensity, and acoustic fingerprint chromaticity matrix. Then, based on these five-dimensional features, it performs intelligent resonance band tracking operation, calling the bearing raceway resonance model, gear meshing resonance model, and insulator corona resonance model in the device acoustic fingerprint library to dynamically track Q factor mutation points and extract acoustic fingerprint spectrum features in the 8-20kHz frequency band.

[0059] Fault fingerprints are obtained by dynamically analyzing acoustic signature spectral features using a convolutional spiking neural network, and a three-dimensional fault feature vector is generated using a CNN-SNN fusion algorithm.

[0060] Furthermore, the convolutional spiking neural network dynamic analysis operation performs fault fingerprint acquisition on the acoustic signature spectral features: First, a 5×5 dynamic convolutional kernel is used to slide on the acoustic signature chromaticity matrix to extract the fault fingerprint. Then, the pulse firing layer performs pulse temporal coding on the fundamental harmonic group distortion rate, high-frequency resonant peak offset, and modulation sideband energy ratio to capture millisecond-level transient impact events. Finally, the CNN-SNN fusion algorithm compresses the convolutional spatial features and pulse temporal features into a three-dimensional fault feature vector containing voltage fluctuation confidence, time-frequency distortion intensity, and acoustic signature coding density.

[0061] S2. Based on the three-dimensional fault feature vector, a dynamic quantization vector is generated by feature matching through a multimodal feature fusion algorithm, and a dynamic weight allocation strategy is adopted to predict the inertia demand increment.

[0062] The three-dimensional fault feature vector is aligned with the preset working condition fingerprint database through a multimodal feature fusion algorithm. The optimal working condition mode is dynamically matched by the Mahalanobis distance-temporal correlation dual criteria in the degradation representation domain to generate a quantized encoding vector.

[0063] The degradation characterization domain refers to the mathematical space of equipment failure evolution constructed by the dual criteria of Mahalanobis distance and temporal correlation.

[0064] The preset process of the operating condition fingerprint database is to extract feature vectors by collecting historical operating data of equipment (including voltage fluctuations, voiceprint coding density, and speed sequence), classify and encapsulate them into 300+ operating condition fingerprints such as bearing overheating mode, gear tooth breakage mode, and insulator flashover mode according to GB / T 6075.6 standard, and store them in the form of degradation characterization domain coordinate points.

[0065] The optimal operating condition mode refers to the historical operating condition fingerprint point within the degradation representation domain that has a Mahalanobis distance ≤ 0.35 and a temporal correlation coefficient ≥ 0.91 with the current three-dimensional fault feature vector. Its core feature is the best matching solution between the physical degradation trajectory and the real-time operating status of the equipment in the cross-modal space.

[0066] It should be noted that, firstly, the three-dimensional fault feature vector is loaded into the degradation representation domain; then, the preset working condition fingerprint database is called to perform Mahalanobis distance-temporal correlation dual-criterion matching. The Mahalanobis distance calculates the multi-dimensional spatial similarity between voltage fluctuation confidence and time-frequency distortion intensity (for example, the distance difference between the fault vector of a circuit breaker mechanism box and the ID107 pattern in the fingerprint database is 0.35), and the temporal correlation analyzes the autocorrelation characteristics of the voiceprint coding density within a 1-second window (for example, a correlation coefficient > 0.91 indicates a gear wear pattern); finally, the dual-criterion scores are fused to generate a quantized coding vector.

[0067] The impact and fatigue components are assigned based on the material fatigue coefficient, and the fusion weights are obtained by weighted dot product. The fused feature vector is then integrated with the quantization encoding vector to generate a dynamic quantization vector.

[0068] Furthermore, the material fatigue coefficient is first input into the degradation trajectory equation to allocate the impact component weight coefficient and the fatigue component weight coefficient; then, the impact component weight coefficient is multiplied by the first six bits of the quantization encoding vector, and the fatigue component weight coefficient is multiplied by the last six bits to obtain the fused feature vector; finally, the weighted results are integrated to generate a dynamic quantization vector.

[0069] The process of allocating impact components to the material fatigue coefficient is achieved through a dynamic weight mapping and physical coupling mechanism. First, a 5-20kHz bandpass filter is used to capture the transient waveform of the impact, and the peak energy and time-domain integral energy of the impact are quantified. Then, the impact coefficient is generated through dynamic weight mapping, and the impact components are classified and allocated based on the characteristics of the impact dominant frequency.

[0070] The fatigue component refers to the fatigue component weight coefficient calculated by the cumulative running time of the equipment and the load spectrum. It is used to perform a dot product operation on the lower six bits of the quantization encoding vector to generate the component in the dynamic quantization vector that represents the degree of material fatigue accumulation.

[0071] The impact component refers to the weighting coefficient (range 0.6~1.2) used to quantify the high six bits of the encoding vector by converting the material fatigue coefficient. Its physical essence is to dynamically characterize the instantaneous impact intensity of rotating machinery impact events on the equipment's inertia demand.

[0072] Based on the dynamic quantization vector, the inertia control coefficient matrix for each dimension of inertia demand is calculated using a dynamic weight allocation strategy.

[0073] Furthermore, the physical meaning of the impact and fatigue components in the dynamic quantization vector is first analyzed; then, a speed range weight allocation table is selected based on the real-time speed value of the equipment, and the quantization vectors of each dimension are integrated to generate dynamic weights. Then, the inertia requirement is obtained by fusing the cooperative gain term, and the aging attenuation coefficient is selected in combination with the cumulative running time of the equipment. Based on the aging attenuation coefficient, the fixed weights of the impact component, fatigue component, and voltage confidence dimension are integrated to generate an inertia control coefficient matrix.

[0074] The expression for generating the inertia control coefficient matrix is: ,

[0075] in, This represents the inertia control coefficient matrix. Indicates the voltage confidence dimension. Indicates a time point. Indicates the impact component The time-related decay coefficient, Represents fatigue component The time-related decay coefficient, express Impact component, express Fatigue weight over time This represents the fixed weight of the voltage confidence dimension.

[0076] Inertia requirement refers to the additional mechanical inertia required to maintain the rated speed, and its value is the dynamic difference between the current state of the equipment and the ground state inertia reference.

[0077] An inertia demand prediction function is constructed using a deep reinforcement learning model, and an embedded adversarial mechanism is used to dynamically and adversarially optimize the inertia control coefficient matrix to predict the incremental inertia demand.

[0078] The deep reinforcement learning model adopts a dual-track decision-making framework with time series awareness, and its core consists of three major architectures: a state parser, a policy generator, and a value evaluator.

[0079] The training of deep reinforcement learning models follows a dynamic closed loop of "exploration-verification-calibration". When the agent performs control actions in the environment, an adaptive noise strategy is introduced. In the early stage, high-intensity random exploration is carried out and then gradually converges. The state-action pairs and real-time rewards generated by each interaction are stored in the experience replay library. Batch samples are randomly selected for iterative optimization and incremental training.

[0080] It should be noted that, firstly, the inertia control coefficient matrix is ​​input into the strategy network to generate the initial inertia demand prediction value, while the value network obtains the value assessment based on the real-time rotation speed of the equipment and the volatility of multiple energy sources; then, an adversarial coefficient perturbation is constructed through the generator network with an embedded adversarial mechanism (e.g., the impact component weight is increased by ±0), and the discriminator network verifies the effectiveness of the perturbation under thermodynamic constraints (e.g., entropy increase rate ≤ 4.2 J / K·s); finally, the strategy network iteratively optimizes and outputs the adversarially enhanced inertia demand increment.

[0081] S3. Call the previously generated three-dimensional fault feature vector and collect the original energy parameters of the multi-energy system (such as wind power, photovoltaic output, hydropower, etc.). Calculate the multi-energy volatility through time series analysis algorithm. Then, align the three-dimensional fault feature vector and the multi-energy volatility in time and space through dynamic time warping-covariance matching method. At the same time, use the phase space manifold weaving engine to construct a four-dimensional input matrix.

[0082] First, determine the range of raw energy parameter collection;

[0083] Based on the type of multi-energy system, collect raw parameters that are strongly correlated with "inertia balance":

[0084] Electrical energy: Real-time output power of wind power / photovoltaic / thermal power (sampling frequency ≥ 50Hz), grid frequency deviation;

[0085] Non-electric energy (such as integrated energy systems): natural gas pipeline pressure, heating network water supply temperature, water network flow rate (sampling frequency ≥ 1Hz);

[0086] Next, a multi-energy volatility calculation algorithm is selected;

[0087] Based on the timescale of regulation requirements (multi-energy balance margin regulation typically requires millisecond to second-level responses), the following algorithm is used for calculation:

[0088] Short-term volatility (millisecond level, matching the time scale of fault characteristics): using the "250ms sliding window standard deviation method", the formula is as follows: ,

[0089] in, For a moment Volatility at any given moment For the first in the sliding window One original power value; N represents the average power within the window, and N is the number of data points within the window (e.g., when the sampling frequency is 50Hz, N=13).

[0090] Medium- to long-term volatility (second-level, matching the energy system volatility cycle): The "1-minute time series difference absolute value mean method" is used, with the following formula: ,

[0091] Where M is the number of sampling points within 1 minute (e.g., M=60 when the sampling frequency is 1Hz).

[0092] Three-dimensional fault features and multi-energy volatility are extracted by dynamic time warping window, and timestamp alignment is performed by dynamic time warping-covariance matching method to generate three-dimensional fault feature trajectory.

[0093] It should be noted that, firstly, a 250ms sliding window is used to extract voltage fluctuation confidence, time-frequency distortion intensity, and acoustic coding density to form a three-dimensional fault feature vector slice, and multi-energy fluctuation rates are collected simultaneously within the same time window; then, the dynamic time warping-covariance matching method is used to align the timestamps of the two types of data: in the covariance tensor analysis, the cross-covariance value between the three-dimensional fault feature vector and the water pressure / flow / power fluctuation is calculated, and the time offset is corrected according to the peak value of the cross-covariance; finally, a three-dimensional fault feature trajectory with strict time synchronization is generated.

[0094] The three-dimensional fault feature trajectory is constructed into a Cantor differential manifold basis using a phase space manifold weaving engine. At the same time, the Cantor differential manifold basis is decomposed by a gauge field transformation to generate a gauge field phase fiber bundle.

[0095] Furthermore, the voltage fluctuation confidence trajectory, time-frequency distortion intensity trajectory, and voiceprint coding density trajectory are first mapped to three-dimensional phase space coordinate axes, and then the local tangent space basis is calculated through the Jacobian matrix; then the cantor differential manifold basis is decomposed by Yang-Mills decomposition using the gauge field transformation (the connection form of the basis fiber bundle is calculated under the SU(3) group representation and projected onto the gauge potential field); finally, the gauge field phase fiber bundle carrying the gauge group phase information is output.

[0096] Dynamically fuse gauge field phase fiber bundles through Chen-Simons topological weaving to generate a four-dimensional phase space structure, and then use qubit logic topological mapping to transform it into a quantum processor.

[0097] Furthermore, the outer product of the fiber bundle connection form is first obtained through the Chen-Simons form, and the 8-dimensional generator of the SU(3) group is subjected to fiber bundle tensor product operation with the phase loop [0,2π]. Then, the four-dimensional phase space structure (voltage / distortion / density / time four axes) is converted into a topological network that can be recognized by the quantum processor by using the quantum bit logic topology mapping, and finally mapped to the quantum processor.

[0098] The quantum processor is analyzed by quantum annealing eigenmodes to generate a four-dimensional input matrix.

[0099] It should be noted that, firstly, a Hamiltonian model is constructed on the quantum processor, and then a quantum annealing scheduling curve is executed on the Hamiltonian model to obtain the probability amplitude vector of the qubit; then, the ground state wave function is obtained through quantum state tomography; then, the eigenmode ground state is extracted from the ground state wave function through a variable quantum feature solver, and mapped to four-dimensional phase space coordinates through qubit measurement; finally, based on the four-dimensional phase space coordinates, a four-dimensional input matrix is ​​generated using the multi-scale quantum resonance differential entropy method, consisting of voltage fluctuation confidence coordinates, time-frequency distortion intensity coordinates, voiceprint coding density coordinates, and time evolution coordinates.

[0100] The core of the Hamiltonian model is to mathematically formalize the energy relationships of spin interactions in magnetic materials. First, the geometry of the lattice and its connections to adjacent nodes are defined (e.g., each spin point in a two-dimensional grid is adjacent to the four points above, below, left, and right). Second, the direction of the magnetic moment at each node is represented by the spin variable σ (with values ​​of ±1). Then, two types of energy are quantified: the magnetic coupling effect of adjacent spins (energy is reduced when they are in the same direction, and the sum is weighted by the coefficient J) and the directional effect of the external magnetic field on the spin (weighted by the field strength h). Finally, by minimizing the energy associated with negative signs, all spin states are integrated to construct the Hamiltonian model.

[0101] The execution of the quantum annealing scheduling curve is a competition between the quantum coherence length and the decoherence rate of the control system.

[0102] S4. Optimize the four-dimensional input matrix through a quantum encoder, perform quantum annealing calculation to generate the ground state solution, integrate the inertia demand increment with the ground state solution through qubit entanglement, calculate the comprehensive inertia gap value, and generate hierarchical scheduling instructions using a dynamic threshold hierarchical strategy.

[0103] The four-dimensional input matrix is ​​mapped to the qubit space by a quantum encoder for optimization, and quantum annealing calculation is performed by a variable quantum feature solver to generate the ground state solution.

[0104] It should be noted that, firstly, the quantum encoder maps the four-dimensional input matrix (containing data in four dimensions: voltage fluctuation confidence, time-frequency distortion intensity, voiceprint coding density, and time evolution) to the qubit space for optimization. Then, it encodes the four-dimensional elements into qubit probability amplitudes through a Pauli operator chain. Subsequently, it performs quantum annealing calculations in the qubit space using a variable quantum feature solver, and relaxes the qubit to the ground state through a quantum phase transition process based on the rotating mechanical Hamiltonian. Finally, it measures the collapse probability of the qubit group under the z-basis vector to generate the ground state solution.

[0105] The quantum annealing calculation expression is: ,

[0106] in, Indicates the quantum annealing evolution time. This represents a time-varying transverse magnetic field strength function. Indicates the first Pauli qubits Operator, Representing a quantum bit and The coupling coefficient between them (within the range of 0.5~1). Indicates the first Pauli qubits Operator, Indicates the first Pauli qubits Operator, Indicates the first Local field strength of each qubit A pair of qubits that are physically connected. This represents the ground-state solution generated during the quantum annealing process. Indicates the index of qubits;

[0107] The time-varying transverse magnetic field strength function refers to the transverse field strength parameter that decays linearly with time during quantum annealing, and it controls the quantum state tunneling probability to achieve ground state convergence.

[0108] Based on the ground state solution, an incremental-ground state coupling field is constructed through qubit entanglement, and the inertia demand increment is integrated with the ground state solution under Pauli Z-gate constraints to generate the inertia gap trajectory function;

[0109] It should be noted that, firstly, the entanglement of qubits constructs an incremental-ground state coupling field, and the probability amplitude vector of the qubits is entangled with the inertia demand increment in a Bell state. Then, an Adama Gate transform is performed on the ground state solution (qubit 12), and entanglement is established through a CNOT gate. Subsequently, an energy conservation term is applied under the Pauli Z-gate constraint. Finally, the time-domain function is extracted through the inverse quantum Fourier transform to generate an inertia gap trajectory function containing the time evolution phase.

[0110] The probability magnitude vector of a qubit is a core mathematical tool in quantum mechanics for describing quantum states. It precisely characterizes the superposition of all possible states of a quantum system in complex vector form.

[0111] Based on the inertia gap trajectory function, the comprehensive inertia gap value is solved iteratively by the Schrödinger-Poisson equation, and hierarchical scheduling instructions are generated by a dynamic threshold hierarchical strategy.

[0112] Furthermore, the Schrödinger-Poisson equation is first iteratively solved for the inertia gap trajectory function. Then, the inertia gap trajectory function is discretized into a spatiotemporal grid, and the Schrödinger equation is solved under the Laplace operator. The potential energy term converges through the Poisson equation after five iterations, and the comprehensive inertia gap value is output. Then, a hierarchical scheduling instruction is generated by matching the preset 12 thresholds through a dynamic threshold hierarchical strategy.

[0113] The preset process for the 12 thresholds is based on the comprehensive inertia gap value. According to the safety guidelines for nuclear power equipment, 12 dynamic thresholds are divided, and each threshold corresponds to a specific scheduling action and is preset.

[0114] S5. Based on the hierarchical scheduling instructions, generate topology reconstruction instructions through a virtual inertia dynamic algorithm, and use multi-agent collaborative game optimization for dynamic monitoring to generate intelligent control reports.

[0115] Based on the hierarchical scheduling instructions, the vulnerable nodes in the power grid topology are analyzed using a virtual inertia dynamic algorithm to generate topology reconfiguration instructions.

[0116] Topology reconfiguration instructions refer to hexadecimal control codes generated by a virtual inertia dynamic algorithm. Their physical meaning is to execute a predefined circuit breaker operation sequence and adjust the line impedance at the coordinates of vulnerable nodes in the power grid.

[0117] It should be noted that, firstly, the virtual inertia dynamic algorithm performs grid topology vulnerability node analysis on the hierarchical scheduling command, then obtains the inertia sensitivity factor of each bus node through the regional grid, and then selects a scheme from the preset topology reconfiguration strategy library according to the scheduling command level, generating a topology reconfiguration command containing the target node coordinates and operation sequence.

[0118] The predefined process for circuit breaker operation sequences involves establishing a standard operation template library, dynamically integrating the basic operation procedures fixed by international standards with the equipment's cumulative operation count and ambient temperature state function, generating physical execution instructions with adaptive correction, and then predefining them.

[0119] Based on the topology reconstruction instructions, a fused topology record is generated through a distributed reinforcement learning control unit, and dynamic monitoring is performed using multi-agent collaborative game optimization to generate an intelligent control report.

[0120] Fusion topology records refer to the real-time grid state structured dataset generated by the distributed reinforcement learning control unit during the grid reconfiguration process, which includes key parameters such as circuit breaker location, line impedance correction value, energy storage unit status, and load rate.

[0121] Furthermore, the distributed reinforcement learning control unit first performs a parsing operation on the topology reconstruction command (example input code: 0xC304), then decodes the command to obtain the target node coordinates and operation sequence (circuit breaker action sequence), and reconstructs the topology mapping map through the regional power grid digital twin model; then it generates a fused topology record containing circuit breaker status, line impedance correction value, and real-time load rate; then it uses multi-agent collaborative game optimization to implement dynamic monitoring: the energy storage unit control agent, the circuit breaker scheduling agent, and the frequency stabilization agent negotiate the control strategy through Nash equilibrium game, execute 4 strategy game iterations per second, and output an intelligent control report containing grid oscillation suppression rate, response delay, and predicted state transition path.

[0122] This embodiment also provides a multi-energy balance margin adaptive control device, including:

[0123] The data acquisition module is used to acquire raw voiceprint signals in real time and perform preprocessing. It uses a convolutional pulse neural network to dynamically analyze the preprocessed raw voiceprint signals to generate voiceprint spectral features, and generates a three-dimensional fault feature vector through a CNN-SNN fusion algorithm.

[0124] The inertia demand calculation module is used to generate a dynamic quantization vector by performing feature matching through a multimodal feature fusion algorithm based on the three-dimensional fault feature vector, and to predict the inertia demand increment by adopting a dynamic weight allocation strategy.

[0125] A four-dimensional input matrix module is constructed to call the previously generated three-dimensional fault feature vector and collect the original energy parameters of the multi-energy system (such as wind power, photovoltaic output, hydropower, etc.). The volatility of the multi-energy system is calculated through time series analysis algorithm. Then, the three-dimensional fault feature vector and the volatility of the multi-energy system are spatiotemporally aligned by dynamic time warping-covariance matching method. At the same time, a phase space manifold weaving engine is used to construct the four-dimensional input matrix.

[0126] First, determine the range of raw energy parameter collection;

[0127] Based on the type of multi-energy system, collect raw parameters that are strongly correlated with "inertia balance":

[0128] Electrical energy: Real-time output power of wind power / photovoltaic / thermal power (sampling frequency ≥ 50Hz), grid frequency deviation;

[0129] Non-electric energy (such as integrated energy systems): natural gas pipeline pressure, heating network water supply temperature, water network flow rate (sampling frequency ≥ 1Hz);

[0130] Next, a multi-energy volatility calculation algorithm is selected;

[0131] Based on the timescale of regulation requirements (multi-energy balance margin regulation typically requires millisecond to second-level responses), the following algorithm is used for calculation:

[0132] Short-term volatility (millisecond level, matching the time scale of fault characteristics): using the "250ms sliding window standard deviation method", the formula is as follows: ,

[0133] in, For a moment Volatility at any given moment For the first in the sliding window One original power value; N represents the average power within the window, and N is the number of data points within the window (e.g., when the sampling frequency is 50Hz, N=13).

[0134] Medium- to long-term volatility (second-level, matching the energy system volatility cycle): The "1-minute time series difference absolute value mean method" is used, with the following formula: ,

[0135] Where M is the number of sampling points in 1 minute (e.g., when the sampling frequency is 1 Hz, M = 60).

[0136] Then, the data is standardized.

[0137] The calculated multi-energy volatility needs to be aligned with the numerical range of the three-dimensional fault feature vector (e.g., normalized to the [0,1] interval) to avoid affecting the accuracy of subsequent covariance matching due to differences in dimensions (e.g., power volatility is in kW, while the fault feature vector is in dimensionless confidence).

[0138] The hierarchical scheduling instruction generation module is used to optimize the four-dimensional input matrix through the quantum encoder, perform quantum annealing calculation to generate the ground state solution, integrate the inertia demand increment with the ground state solution through qubit entanglement, calculate the comprehensive inertia gap value, and generate hierarchical scheduling instructions using a dynamic threshold hierarchical strategy.

[0139] The module for generating control reports is used to generate topology reconfiguration instructions based on hierarchical scheduling instructions through a virtual inertia dynamic algorithm, and to perform dynamic monitoring using multi-agent collaborative game optimization to generate intelligent control reports.

[0140] This embodiment also provides a computer device applicable to the multi-energy balance margin adaptive control method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multi-energy balance margin adaptive control method proposed in the above embodiment.

[0141] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0142] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the adaptive control method for multi-energy balance margin as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0143] In summary, this invention achieves mandatory protection of mechanical energy conservation by constructing an incremental-ground state coupled field through qubit entanglement and physically integrating the incremental inertia requirement with the ground state solution under Pauli Z-gate constraints. Furthermore, after aligning the spatiotemporal dimensions of three-dimensional fault characteristics and multi-energy volatility based on the dynamic time warping-covariance matching method, it employs Chern-Simons topological weaving to dynamically fuse the gauge field phase fiber bundles, achieving adaptive control of the multi-energy balance margin.

[0144] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-energy balance margin adaptive control method, characterized in that: include, Real-time acquisition of raw voiceprint signals and preprocessing; dynamic analysis of preprocessed raw voiceprint signals using convolutional pulse neural networks to generate voiceprint spectral features; and generation of three-dimensional fault feature vectors using CNN-SNN fusion algorithm. A multimodal feature fusion algorithm is used to perform feature matching on the three-dimensional fault feature vector to generate a dynamic quantization vector. A dynamic weight allocation strategy is then used to predict the inertia demand increment of the dynamic quantization vector. The three-dimensional fault feature vector generated earlier is called, and the original energy parameters of the multi-energy system are collected. The multi-energy volatility is calculated by time series analysis algorithm. Then, the three-dimensional fault feature vector and the multi-energy volatility are spatiotemporally aligned by dynamic time warping-covariance matching method. At the same time, a four-dimensional input matrix is ​​constructed by phase space manifold weaving engine. The four-dimensional input matrix is ​​optimized by a quantum encoder, and the optimization result is calculated by a quantum annealing algorithm to generate a ground state solution. The inertia demand increment is integrated with the ground state solution by qubit entanglement to obtain a comprehensive inertia gap value. Based on the comprehensive inertia gap value, a dynamic threshold hierarchical strategy is adopted to generate hierarchical scheduling instructions. Based on the hierarchical scheduling instructions, a topology reconfiguration instruction is generated through a virtual inertia dynamic algorithm, and dynamic monitoring is performed using multi-agent collaborative game optimization to generate an intelligent control report.

2. The multi-energy balance margin adaptive control method as described in claim 1, characterized in that: The process involves real-time acquisition and preprocessing of raw voiceprint signals, followed by dynamic analysis of the preprocessed raw voiceprint signals using a convolutional pulse neural network to generate voiceprint spectral features. A three-dimensional fault feature vector is then generated using a CNN-SNN fusion algorithm. The specific steps are as follows: The original voiceprint signal includes fundamental harmonic groups, high-frequency resonant peaks, modulation sidebands, broadband impulse energy, and voiceprint chromaticity map. The preprocessing includes dynamic noise suppression, key frequency band extraction, anti-aliasing sampling, wavelet packet denoising, and signal-to-noise ratio; The acoustic signature spectral features are extracted from the original acoustic signature spectrum by multi-scale power spectrum integration and intelligent resonant band tracking using an acoustic signature physical feature decoupling device. Fault fingerprints are obtained by dynamically analyzing acoustic signature spectral features using a convolutional spiking neural network, and a three-dimensional fault feature vector is generated using a CNN-SNN fusion algorithm.

3. The multi-energy balance margin adaptive control method as described in claim 2, characterized in that: Based on the three-dimensional fault feature vector, a dynamic quantization vector is generated through feature matching using a multimodal feature fusion algorithm. The specific steps are as follows. The three-dimensional fault feature vector is aligned with the preset working condition fingerprint database through a multimodal feature fusion algorithm. The optimal working condition mode is dynamically matched by the Mahalanobis distance-temporal correlation dual criteria in the degradation representation domain to generate a quantized encoding vector. The impact and fatigue components are assigned based on the material fatigue coefficient, and the fusion weights are obtained by weighted dot product. The fused feature vector is then integrated with the quantization encoding vector to generate a dynamic quantization vector.

4. The multi-energy balance margin adaptive control method as described in claim 3, characterized in that: The specific steps for predicting the incremental inertia demand using a dynamic weight allocation strategy are as follows. Based on the dynamic quantization vector, the inertia control coefficient matrix for each dimension of inertia demand is calculated using a dynamic weight allocation strategy. An inertia demand prediction function is constructed using a deep reinforcement learning model, and an embedded adversarial mechanism is used to dynamically and adversarially optimize the inertia control coefficient matrix to predict the incremental inertia demand.

5. The multi-energy balance margin adaptive control method as described in claim 4, characterized in that: The process involves calling the previously generated three-dimensional fault feature vector, simultaneously collecting the original energy parameters of the multi-energy system, and calculating the multi-energy volatility using a time-series analysis algorithm. Subsequently, the three-dimensional fault feature vector and the multi-energy volatility are spatiotemporally aligned using a dynamic time warping-covariance matching method. A four-dimensional input matrix is ​​then constructed using a phase space manifold weaving engine. The specific steps are as follows: The three-dimensional fault feature vector and multi-energy volatility are extracted by dynamic time warping window, and the timestamp alignment is performed by dynamic time warping-covariance matching method to generate three-dimensional fault feature trajectory. The three-dimensional fault feature trajectory is constructed into a Cantor differential manifold basis using a phase space manifold weaving engine. At the same time, the Cantor differential manifold basis is decomposed by a gauge field transformation to generate a gauge field phase fiber bundle. Dynamically fuse gauge field phase fiber bundles through Chen-Simons topological weaving to generate a four-dimensional phase space structure, and then use qubit logic topological mapping to transform it into a quantum processor. The quantum processor is analyzed by quantum annealing eigenmodes to generate a four-dimensional input matrix.

6. The multi-energy balance margin adaptive control method as described in claim 5, characterized in that: The process involves optimizing the four-dimensional input matrix using a quantum encoder, performing quantum annealing to generate a ground-state solution, integrating the inertia demand increment with the ground-state solution through qubit entanglement, calculating the comprehensive inertia gap value, and generating hierarchical scheduling instructions using a dynamic threshold hierarchical strategy. The specific steps are as follows: The four-dimensional input matrix is ​​mapped to the qubit space by a quantum encoder for optimization, and quantum annealing calculation is performed by a variable quantum feature solver to generate the ground state solution. Based on the ground state solution, an incremental-ground state coupling field is constructed through qubit entanglement, and the inertia demand increment is integrated with the ground state solution under Pauli Z-gate constraints to generate the inertia gap trajectory function; Based on the inertia gap trajectory function, the comprehensive inertia gap value is calculated by iteratively solving the Schrödinger-Poisson equation, and hierarchical scheduling instructions are generated through a dynamic threshold hierarchical strategy.

7. The multi-energy balance margin adaptive control method as described in claim 6, characterized in that: The process involves generating topology reconfiguration instructions based on hierarchical scheduling commands using a virtual inertia dynamic algorithm, and then dynamically monitoring these instructions using multi-agent cooperative game theory to generate an intelligent control report. The specific steps are as follows: Based on the hierarchical scheduling instructions, the vulnerable nodes in the power grid topology are analyzed using a virtual inertia dynamic algorithm to generate topology reconfiguration instructions. Based on the topology reconstruction instructions, a fused topology record is generated through a distributed reinforcement learning control unit, and dynamic monitoring is performed using multi-agent collaborative game optimization to generate an intelligent control report.

8. A multi-energy balance margin adaptive control device, characterized in that: include, The data acquisition module is used to acquire raw voiceprint signals in real time and perform preprocessing. It uses a convolutional pulse neural network to dynamically analyze the preprocessed raw voiceprint signals to generate voiceprint spectral features, and generates a three-dimensional fault feature vector through a CNN-SNN fusion algorithm. The inertia demand calculation module is used to generate a dynamic quantization vector by performing feature matching through a multimodal feature fusion algorithm based on the three-dimensional fault feature vector, and to predict the inertia demand increment by adopting a dynamic weight allocation strategy. A four-dimensional input matrix module is constructed to call the previously generated three-dimensional fault feature vector and collect the original energy parameters of the multi-energy system. The volatility of the multi-energy system is calculated through a time series analysis algorithm. Then, the three-dimensional fault feature vector and the volatility of the multi-energy system are spatiotemporally aligned using the dynamic time warping-covariance matching method. At the same time, a phase space manifold weaving engine is used to construct the four-dimensional input matrix. The hierarchical scheduling instruction generation module is used to optimize the four-dimensional input matrix through the quantum encoder, perform quantum annealing calculation to generate the ground state solution, integrate the inertia demand increment with the ground state solution through qubit entanglement, calculate the comprehensive inertia gap value, and generate hierarchical scheduling instructions using a dynamic threshold hierarchical strategy. The module for generating control reports is used to generate topology reconfiguration instructions based on hierarchical scheduling instructions through a virtual inertia dynamic algorithm, and to perform dynamic monitoring using multi-agent collaborative game optimization to generate intelligent control reports.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the multi-energy balance margin adaptive control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the multi-energy balance margin adaptive control method according to any one of claims 1 to 7.

Images