A method and system for multi-source data fusion analysis and adaptive control in energy terminals

Abstract

This invention relates to the field of multi-source data fusion technology, and in particular to a method and system for multi-source data fusion analysis and adaptive control of energy terminals. The method includes converting multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation; performing tensor fusion analysis on the manifold coordinate data based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure; and dynamically adjusting the multi-kernel tensor kernel function parameters of the fusion analysis module and the state equation parameters of the control decision module by collecting equipment operating status data and comparing the deviation between the predicted and measured values ​​of the fusion features, forming a closed-loop correction mechanism that can continuously self-optimize based on the actual operating conditions of the equipment.

Description

Technical Field

[0001] This invention relates to the field of multi-source data fusion technology, and in particular to a method and system for multi-source data fusion analysis and adaptive control of energy terminals. Background Technology

[0002] In today's era of rapid digital and intelligent development, energy management is crucial for achieving efficient energy use, reducing costs, and promoting sustainable development. As key nodes in energy consumption and use, energy terminals contain a wealth of information in their operational status and data; therefore, effective management and regulation of these terminals is an important area of ​​research in the energy field.

[0003] Currently, traditional multi-source data fusion analysis and adaptive control systems for energy terminals often employ single data sources or simple data fusion methods for data processing and regulation. For example, they might only collect power data from electricity terminals and control equipment startup and shutdown using simple threshold judgments. Alternatively, when dealing with multi-source data, they might use linear methods such as Principal Component Analysis (PCA) for dimensionality reduction and simply weight and fuse different types of energy terminal data (such as electricity and heat data). However, multi-source heterogeneous data from energy terminals has complex geometric structures and nonlinear relationships between data points. This traditional approach cannot effectively handle the complex geometric structures of multi-source heterogeneous data. Simple linear fusion loses the nonlinear relationships and important features between data points, resulting in insufficient utilization of data information and difficulty in accurately reflecting the true operating status of energy terminals, thus affecting the efficient management and regulation of energy terminals. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method and system for multi-source data fusion analysis and adaptive control of energy terminals. This solves the problem that traditional multi-source data fusion analysis and adaptive control systems for energy terminals often fail to fully utilize data information, making it difficult to accurately reflect the true operating status of energy terminals and thus affecting the efficient management and control of energy terminals.

[0005] In a first aspect, the present invention provides a method for multi-source data fusion analysis and adaptive control of energy terminals, which adopts the following technical solution:

[0006] A method for multi-source data fusion analysis and adaptive control of energy terminals, comprising:

[0007] Acquire multi-source heterogeneous data from energy terminals;

[0008] Multi-source heterogeneous data is converted into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation;

[0009] Tensor fusion analysis of manifold coordinate data is performed based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure.

[0010] The dynamic operation of energy terminal equipment is parametrically modeled by utilizing low-dimensional fusion features, and the control strategy is obtained by solving the minimum condition of the stochastic Hamiltonian function.

[0011] Generate equipment control commands based on the control strategy;

[0012] Adaptive optimization is performed based on feedback from the device's operating status under control commands.

[0013] Furthermore, the acquisition of multi-source heterogeneous data from the energy terminal includes acquiring multi-source heterogeneous data from the energy terminal through a sensor array and the IEEE 1588 time synchronization protocol. The multi-source heterogeneous data includes complex voltage phasor data of the power terminal, temperature and humidity joint distribution data of the thermal terminal, and vibration signal data of the equipment. Based on complex manifold coordinate transformation and probabilistic manifold metric calculation, the multi-source heterogeneous data is converted into manifold coordinate data.

[0014] Furthermore, the process of converting multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation includes performing a holomorphic mapping transformation on complex voltage phasor data, wherein the transformation formula for the holomorphic mapping transformation is:

[0015] ,

[0016] in, For complex voltage phasors, For the real part, The imaginary part is used to convert the complex voltage phasor data into two independent real-valued components, the real part and the imaginary part, to facilitate subsequent processing, thereby realizing the formal transformation of the complex voltage phasor data.

[0017] Furthermore, the process of converting multi-source heterogeneous data into manifold coordinate-based data through complex manifold coordinate transformation and probabilistic manifold metric calculation also includes constructing a Fisher information matrix as a Riemannian metric for the joint temperature and humidity distribution data. This Fisher information matrix is ​​then used as a metric tensor on the Riemannian manifold to map the joint temperature and humidity distribution data onto the Riemannian manifold, thus preserving the geometric structure of the data. The elements of the Fisher information matrix are represented as follows:

[0018] ,

[0019] in, For distribution parameters, Let be the joint probability density function. .

[0020] Furthermore, the tensor fusion analysis of manifold coordinated data based on multi-kernel tensor kernel functions includes constructing a joint data tensor from the manifold coordinated data and calculating the kernel matrix based on the multi-kernel tensor kernel function to obtain a multi-kernel tensor kernel matrix. The analysis unit is used to perform centering and eigenvalue decomposition on the multi-kernel tensor kernel matrix to obtain eigenvectors and eigenvalues. The generation unit is used to perform embedding mapping calculation on the eigenvalues ​​and eigenvectors based on the eigenvectors corresponding to the maximum number of preset eigenvalues ​​to generate low-dimensional fusion features that preserve geometric structure. The multi-kernel tensor kernel function is defined as the product of the kernel functions of each submanifold, expressed as:

[0021] ,

[0022] in, For the first A joint data tensor, For heterogeneous data types, For the first Submanifold kernel function, For the first The first Class data.

[0023] Furthermore, the tensor fusion analysis of manifold coordinated data based on multi-kernel tensor kernel functions also includes centering and eigenvalue decomposition of the multi-kernel tensor kernel matrix to obtain eigenvectors and eigenvalues. The centering process eliminates the influence of the data's mean, allowing subsequent analysis to proceed with the data under zero mean. The eigenvalue problem solved by eigenvalue decomposition is defined as... ,in, For a centralized kernel matrix, For a centered matrix, For feature vectors, For eigenvalues, Given the number of data samples, the multi-kernel tensor kernel matrix is ​​decomposed into eigenvectors and eigenvalues ​​through eigenvalue decomposition. The eigenvalues ​​reflect the variance of the data in different directions. The larger the variance, the richer the data information in that direction.

[0024] Furthermore, the parameterized modeling of the equipment operation dynamics of the energy terminal using low-dimensional fusion features includes, based on the mapping relationship between the low-dimensional fusion features and the physical characteristics of the energy terminal equipment, establishing a stochastic differential equation with feature parameterization to parameterize the equipment operation dynamics of the energy terminal. The low-dimensional fusion features retain key geometric information from multi-source heterogeneous data, and the equipment physical characteristics include the electrical parameters and mechanical characteristics of the equipment. By analyzing the mapping relationship, a state equation is established. ,in, For state vectors, For cumulative energy consumption, For the health status of the equipment, To accumulate carbon emissions and reflect the environmental impact of equipment operation; For real-time power input, It is a deterministic dynamic function; The random perturbation matrix; For Wiener process vectors.

[0025] Furthermore, the process of obtaining the control strategy by solving the minimum condition of the stochastic Hamiltonian function includes, based on the state equation, defining a multi-objective performance functional that includes energy consumption, equipment lifespan, and carbon emissions, obtaining a performance functional expression containing the expected integral, constructing a stochastic Hamiltonian function based on the performance functional expression, and solving the minimum condition to obtain the control strategy. The performance functional expression is:

[0026] ,

[0027] in, This is a parameter for electricity price weighting; As a weighting parameter for equipment maintenance costs; For carbon emission weighting parameters; For terminal health constraint weight parameters; For real-time electricity prices, For the regulation cycle.

[0028] Furthermore, the step of generating equipment control commands based on the control strategy includes: performing rolling optimization of the control strategy in the prediction time domain based on a model predictive control algorithm to obtain a reference power curve; designing a sliding mode surface and an exponential reaching law to robustly compensate for unmodeled disturbances by addressing the deviation between the actual operating state of the energy terminal equipment and the expected state corresponding to the reference power curve; obtaining a robust compensation term; and defining the sliding mode surface as... ,in, For state tracking error, For a positive definite matrix, the exponential reaching law is: ,in, and The sliding mode control parameters are used; finally, the reference power curve and the robust compensation term are compared using adaptive weighting coefficients. Linear superposition generates equipment control commands.

[0029] Furthermore, the adaptive optimization based on device operating status feedback under control commands includes: collecting real-time operating status data of the terminal device through sensors to obtain a real-time status dataset; inputting the real-time status dataset into a multi-core tensor kernel function to calculate the measured value of the fused feature; predicting the fused feature at the same time step based on the state equation; calculating the deviation between the two; and dynamically adjusting the parameters of the multi-core tensor kernel function and the parameters of the state equation based on the deviation to form a closed-loop correction mechanism. The formula for calculating the deviation is as follows: ,in, To fuse feature predictions, The measured values ​​are used to fuse features.

[0030] Secondly, an energy terminal multi-source data fusion analysis and adaptive control system includes:

[0031] The data acquisition module is configured to acquire multi-source heterogeneous data from energy terminals;

[0032] The popular coordinate module is configured to convert multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation;

[0033] The feature fusion module is configured to perform tensor fusion analysis on manifold coordinate data based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure.

[0034] The control strategy module is configured to use low-dimensional fusion features to parametrically model the dynamic operation of the energy terminal equipment and obtain the control strategy by solving the minimum condition of the stochastic Hamiltonian function.

[0035] The control command module is configured to generate device control commands based on the control strategy.

[0036] The feedback module is configured to perform adaptive optimization based on the device operating status feedback under control commands.

[0037] Thirdly, the present invention provides a computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device of the aforementioned method for multi-source data fusion analysis and adaptive control of an energy terminal.

[0038] Fourthly, the present invention provides a terminal device, including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor to provide the energy terminal multi-source data fusion analysis and adaptive control method.

[0039] In summary, the present invention has the following beneficial technical effects:

[0040] 1. This invention acquires multi-source heterogeneous data from energy terminals, obtains manifold coordinate-based data through complex manifold coordinate transformation and probabilistic manifold metric calculation, generates low-dimensional fusion features that preserve geometric structure based on multi-kernel tensor kernel functions, and constructs state equations and multi-objective performance functionals containing random perturbation terms using these features. A control strategy is obtained by solving the stochastic Hamiltonian function, and the strategy is implemented simultaneously. Equipment operating status is collected, and the parameters of the state equation are adaptively optimized to form a closed-loop correction mechanism. This achieves deep fusion and accurate analysis of multi-source data, fully utilizes data information, and reflects the true operating status of the equipment. Combined with a control strategy considering multiple objectives and uncertainties, and adaptive optimization, this invention solves the problem that traditional multi-source data fusion analysis and adaptive control systems for energy terminals often fail to fully utilize data information, making it difficult to accurately reflect the true operating status of the energy terminal, thus affecting the efficient management and control of the energy terminal.

[0041] 2. This invention converts multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation. It adopts a pure mapping transformation for complex voltage phasor data, constructs a Fisher information matrix as a Riemann metric for temperature and humidity joint distribution data, and then uses multi-kernel tensor kernel functions for tensor fusion analysis. This more completely captures the complex nonlinear relationships between data, making the low-dimensional fusion features more completely reflect the operating status of energy terminals and improving the utilization value of the data.

[0042] 3. This invention performs centering and eigenvalue decomposition on the multi-core tensor kernel matrix, and performs embedding mapping calculation on the eigenvalues ​​and eigenvectors based on the eigenvectors corresponding to the maximum number of preset eigenvalues, generating low-dimensional fusion features that retain the geometric structure. This reduces the data dimension and computational complexity while retaining key information, and improves the efficiency of data processing and analysis, enabling the system to run faster and more efficiently when processing large-scale multi-source heterogeneous data.

[0043] 4. This invention constructs a state equation that includes random disturbance terms and defines a multi-objective performance functional that comprehensively considers factors such as energy consumption, equipment lifespan, and carbon emissions. This enables the formulation of control strategies to take into account multiple needs such as energy consumption, equipment maintenance, and environmental protection, thereby maximizing the comprehensive benefits of energy terminal equipment. The control strategy is obtained by solving the minimum condition using a stochastic Hamiltonian function, which takes into account random disturbances during equipment operation, improves the robustness of the control strategy, and ensures the reliable operation of energy terminal equipment.

[0044] 5. This invention combines model predictive control and sliding mode control, uses rolling optimization to obtain a reference power curve, and designs a sliding surface and exponential reaching law for robust compensation for state deviations. The two are weighted and superimposed to generate control commands. This invention can utilize the forward-looking nature of model predictive control for optimization, and can also use sliding mode control to deal with unmodeled disturbances, effectively improving the accuracy and reliability of energy terminal equipment control and ensuring stable and efficient equipment operation.

[0045] 6. This invention collects equipment operating status data, compares the deviation between the predicted and measured values ​​of the fusion features, and dynamically adjusts the multi-core tensor kernel function parameters of the fusion analysis module and the state equation parameters of the control decision module to form a closed-loop correction mechanism. This mechanism can continuously optimize itself based on the actual operating conditions of the equipment, continuously improve the prediction and control capabilities of the operating status of energy terminal equipment, enhance long-term adaptability and control reliability, and further improve the performance and stability of the entire energy terminal control system. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of a multi-source data fusion analysis and adaptive control method for energy terminals according to Embodiment 1 of the present invention. Detailed Implementation

[0047] The present invention will be further described in detail below with reference to the accompanying drawings.

[0048] Example 1

[0049] Reference Figure 1 This embodiment of a method for multi-source data fusion analysis and adaptive control of energy terminals includes:

[0050] Acquire multi-source heterogeneous data from energy terminals;

[0051] Multi-source heterogeneous data is converted into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation;

[0052] Tensor fusion analysis of manifold coordinate data is performed based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure.

[0053] The dynamic operation of energy terminal equipment is parametrically modeled by utilizing low-dimensional fusion features, and the control strategy is obtained by solving the minimum condition of the stochastic Hamiltonian function.

[0054] Generate equipment control commands based on the control strategy;

[0055] Adaptive optimization is performed based on feedback from the device's operating status under control commands.

[0056] Specifically:

[0057] S1. Acquire multi-source heterogeneous data from energy terminals.

[0058] The acquisition and processing module includes a data acquisition unit and a data processing unit. The data acquisition unit is used to acquire multi-source heterogeneous data from energy terminals through a sensor array and the IEEE 1588 time synchronization protocol. The multi-source heterogeneous data includes complex voltage phasor data from power terminals, temperature and humidity joint distribution data from thermal terminals, and vibration signal data from equipment. The data processing unit is used to convert the multi-source heterogeneous data into manifold coordinate data based on complex manifold coordinate transformation and probabilistic manifold metric calculation.

[0059] S2. Transform multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation;

[0060] In the process of converting data into manifold coordinate data, a holomorphic mapping transformation is performed on the complex voltage phasor data, and a Fisher information matrix is ​​constructed as a Riemannian metric for the joint distribution data of temperature and humidity.

[0061] The transformation formula for holomorphic mapping transformation is:

[0062] ,

[0063] in, For complex voltage phasors, For the real part, It is the imaginary part;

[0064] The elements of the Fisher information matrix are:

[0065] ,

[0066] in, For distribution parameters, Let be the joint probability density function. .

[0067] Specifically, multi-source heterogeneous data is converted into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation. This module includes a data acquisition unit and a data processing unit. The data acquisition unit acquires multi-source heterogeneous data from energy terminals through a sensor array and the IEEE 1588 time synchronization protocol. This data includes complex voltage phasor data from power terminals, temperature and humidity joint distribution data from thermal terminals, and vibration signal data from equipment. The complex voltage phasor data from power terminals is acquired through voltage transformers and current transformers, the temperature and humidity joint distribution data from thermal terminals is acquired through a temperature and humidity sensor network, and the vibration signal data from equipment is acquired through accelerometers. The raw data acquired by the sensor array is synchronized in time using the IEEE 1588 time synchronization protocol to ensure consistency of multi-source data in the time dimension.

[0068] in,

[0069] (1) The data processing unit converts multi-source heterogeneous data into manifold coordinate data based on complex manifold coordinate transformation and probabilistic manifold metric calculation. For complex voltage phasor data, a holomorphic mapping transformation is used. Holomorphic mapping is an important concept in complex analysis; it preserves the complex structure and can effectively process complex data in power systems. The transformation formula for holomorphic mapping is:

[0070] ,

[0071] in, For complex voltage phasors, For the real part, This is the imaginary part. This formula converts complex voltage phasor data into two independent real-valued components, the real and imaginary parts, facilitating subsequent processing. In complex voltage phasor data processing, let the input complex voltage phasor... According to the holomorphic mapping transformation formula ,in The real part is 3; The imaginary part is 4. The output is... This transforms the original complex voltage phasor data into two independent real-valued components, real and imaginary, facilitating subsequent analysis and processing of power terminal data. This achieves a formal transformation of the complex voltage phasor data, making it more suitable for subsequent data processing workflows.

[0072] (2) For the joint distribution data of temperature and humidity, a Fisher information matrix is ​​constructed as a Riemannian metric. The Fisher information matrix is ​​an important tool in statistics for measuring the accuracy of parameter estimation. In manifold learning, it can be used as a metric tensor on the Riemannian manifold. The elements of the Fisher information matrix are:

[0073] ,

[0074] in, For distribution parameters, Let be the joint probability density function. .

[0075] By constructing a Fisher information matrix, the joint temperature and humidity distribution data are mapped onto a Riemannian manifold, thus preserving the geometric structure of the data. In the processing of the joint temperature and humidity distribution data, the distribution parameters are set... ,in For temperature, For humidity, the joint probability density function Given. Assume that the joint probability density function of temperature and humidity within a certain range is: (Example function only). For Fisher information matrix elements:

[0076] ,

[0077] when At that time, first seek ,

[0078] right Taking the logarithm, we get ,

[0079] but Similarly .

[0080] Substitute into the formula to calculate:

[0081] ,

[0082] After integration (assuming the integration range is...) ) can be obtained The value of .

[0083] By calculating the values ​​of each element of the Fisher information matrix, the Fisher information matrix is ​​constructed, which maps the joint temperature and humidity distribution data onto the Riemannian manifold, preserving the geometric structure of the joint temperature and humidity distribution data and providing a more suitable data format for subsequent analysis and processing of thermal terminal data.

[0084] (3) For the vibration signal data of the equipment, preprocessing is first performed, including noise reduction and filtering, and then feature parameters such as vibration frequency and amplitude are extracted. These feature parameters constitute the feature space of the vibration signal, and are mapped to a low-dimensional manifold space through the manifold learning algorithm to realize the dimensionality reduction and feature extraction of the data.

[0085] Through the aforementioned complex manifold coordinate transformation and probabilistic manifold metric calculation, multi-source heterogeneous data are converted into unified manifold coordinate data. This transformation method effectively preserves the inherent geometric structure of the data, providing a solid data foundation for subsequent fusion analysis. In practical applications at energy terminals, this acquisition and processing module can improve the accuracy and efficiency of data processing, providing strong support for the optimized control of energy terminals.

[0086] S3. Tensor fusion analysis is performed on manifold coordinate data based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure. Specifically, based on multi-kernel tensor kernel functions, the manifold coordinate data is constructed into a joint data tensor and the kernel matrix is ​​calculated to obtain the multi-kernel tensor kernel matrix. The analysis unit is used to perform centering and eigenvalue decomposition on the multi-kernel tensor kernel matrix to obtain eigenvectors and eigenvalues. The generation unit is used to perform embedding mapping calculation on the eigenvalues ​​and eigenvectors based on the eigenvectors corresponding to the maximum number of preset eigenvalues ​​to generate low-dimensional fusion features that preserve geometric structure.

[0087] The kernel function of a multi-kernel tensor is defined as the product of the kernel functions of each submanifold:

[0088] ,

[0089] in, For the first A joint data tensor, For heterogeneous data types, For the first Submanifold kernel function, For the first The first Class data;

[0090] The eigenvalue problem of solving eigenvalue decomposition is: ,in, For a centralized kernel matrix, For a centered matrix, For feature vectors, For eigenvalues, This represents the number of data samples.

[0091] Specifically,

[0092] (1) The fusion analysis module is used to perform tensor fusion analysis on manifold coordinate data based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure. This module includes fusion unit, analysis unit and generation unit.

[0093] The fusion unit, based on a multi-kernel tensor kernel function, constructs a joint data tensor from the manifold coordinated data and calculates the kernel matrix to obtain the multi-kernel tensor kernel matrix. The multi-kernel tensor kernel function is defined as the product of the kernel functions of each submanifold, i.e.:

[0094] ,

[0095] in, For the first A joint data tensor, For heterogeneous data types, For the first Submanifold kernel function, For the first The first Different types of submanifold kernel functions are selected based on the characteristics of the data, such as Gaussian kernel functions and polynomial kernel functions. Taking three heterogeneous data types—complex voltage phasor data from power terminals, joint temperature and humidity distribution data from thermal terminals, and vibration signal data from equipment—as examples, different submanifold kernel functions are used to construct a joint data tensor from these three types of manifold coordinate-based data. The kernel matrix is ​​then calculated using a multi-kernel tensor kernel function to capture the complex nonlinear relationships between different types of data.

[0096] Among them, the kernel function formula for multi-kernel tensors is:

[0097] ,

[0098] Let heterogeneous data types number These are the complex voltage phasor data of the power terminal (corresponding to...) ), temperature and humidity joint distribution data of heat terminals (corresponding to Vibration signal data of the equipment (corresponding) ). Assume For two joint data tensors, where For complex voltage phasor data, This is joint distribution data of temperature and humidity. Let the vibration signal data be the kernel function of the first type of submanifold. Gaussian kernel function Kernel function of submanifold of type 2 For polynomial kernel functions Kernel function of the third type of submanifold linear kernel function Substituting the above data into the multi-kernel tensor kernel function formula, we can obtain... After calculation, an element value is obtained from the kernel matrix. By calculating on different joint data tensors, a multi-kernel tensor kernel matrix is ​​constructed. This process enables the fusion calculation of different types of manifold coordinate data, capturing the complex nonlinear relationships between different types of data, and providing a foundation for subsequent analysis.

[0099] (2) The analysis unit performs centering and eigenvalue decomposition on the multi-kernel tensor kernel matrix to obtain eigenvectors and eigenvalues. Centering is performed to eliminate the influence of the mean in the data, allowing subsequent analysis to proceed with the data under zero mean. The eigenvalue problem solved by eigenvalue decomposition is... ,in, For a centralized kernel matrix, For a centered matrix, For feature vectors, For eigenvalues, The number of data samples is denoted as . Eigenvalue decomposition decomposes the multi-kernel tensor matrix into eigenvectors and eigenvalues. The eigenvalues ​​reflect the variance of the data in different directions; a larger variance indicates richer data information in that direction.

[0100] Specifically, for the formula of the eigenvalue problem Let the number of data samples be... Centralized kernel matrix For one The matrix, the centered matrix ,in for The identity matrix, for A column vector of all ones. Assume the multi-kernel tensor kernel matrix has already been obtained through the fusion unit. The matrix is ​​solved by eigenvalue decomposition, and the equations are solved using iterative algorithms (such as the power method). , to obtain the feature vector and eigenvalues For example, obtaining eigenvalues. And their corresponding eigenvectors Eigenvalues ​​reflect the variance of data in different directions. Through this eigenvalue decomposition, the multi-kernel tensor matrix is ​​decomposed into eigenvectors and eigenvalues, extracting key information of the data in different dimensions. This provides a basis for the subsequent generation of low-dimensional fusion features, realizing the extraction of data features from the multi-kernel tensor matrix.

[0101] (3) Based on the eigenvectors corresponding to the maximum number of preset eigenvalues, embedding mapping calculations are performed on the eigenvalues ​​and eigenvectors to generate low-dimensional fusion features that preserve geometric structure. Because the eigenvectors corresponding to the maximum eigenvalues ​​contain the main information of the data, embedding mapping calculations are performed on these eigenvalues ​​and eigenvectors to map the high-dimensional manifold coordinate data to a low-dimensional space while preserving the geometric structure of the data. This low-dimensional fusion feature reduces the data dimensionality and computational complexity while preserving the key geometric information of the original multi-source heterogeneous data, providing a more effective data representation for subsequent regulation and control decisions. Through the processing of the fusion analysis module, multi-source heterogeneous data from energy terminals can be presented in a more compact and representative form, improving data utilization efficiency and laying a good foundation for the optimized regulation and control of energy terminals.

[0102] S4. Utilizing low-dimensional fusion features, parameterize the dynamic operation of energy terminal equipment, construct a state equation containing random disturbance terms, and define a multi-objective performance functional including energy consumption, equipment lifespan, and carbon emissions. The control strategy is obtained by solving the minimum condition of the stochastic Hamiltonian function. The control decision module includes a stochastic state modeling unit, a performance functional definition unit, and a stochastic Hamiltonian solving unit. The stochastic state modeling unit is used to parameterize the dynamic operation of energy terminal equipment based on the mapping relationship between low-dimensional fusion features and the physical characteristics of energy terminal equipment. By establishing a stochastic differential equation with characteristic parameters, it constructs a state equation containing random disturbance terms. The performance functional definition unit is used to define a multi-objective performance functional including energy consumption, equipment lifespan, and carbon emissions based on the state equation, obtaining a performance functional expression containing the expected integral. The stochastic Hamiltonian solving unit is used to construct a stochastic Hamiltonian function based on the performance functional expression and solve the minimum condition to obtain the control strategy.

[0103] (1) The state equation is:

[0104] ,

[0105] in, For state vectors, For cumulative energy consumption, For the health status of the equipment, To accumulate carbon emissions, For real-time power input, For deterministic dynamic functions, The random perturbation matrix, The vector for the Wiener process;

[0106] The performance functional expression is:

[0107] ,

[0108] in, For electricity price weighting parameters, As a weighting parameter for equipment maintenance costs, For carbon emission weighting parameters, For terminal health constraint weight parameters, For real-time electricity prices, For the regulation cycle.

[0109] Specifically, the control and decision-making module utilizes low-dimensional fusion features to parametrically model the dynamic operation of energy terminal equipment, constructs state equations containing stochastic disturbance terms, and defines a multi-objective performance functional that includes energy consumption, equipment lifespan, and carbon emissions. The control strategy is obtained by solving the minimum condition of the stochastic Hamiltonian function. This module includes a stochastic state modeling unit, a performance functional definition unit, and a stochastic Hamiltonian solution unit.

[0110] The stochastic state modeling unit, based on the mapping relationship between low-dimensional fused features and the physical characteristics of energy terminal equipment, parametrically models the dynamic operation of energy terminal equipment by establishing characteristic parameterized stochastic differential equations. The low-dimensional fused features are obtained by the fusion analysis module, preserving key geometric information from multi-source heterogeneous data. Equipment physical characteristics include electrical parameters and mechanical properties. State equations are established by analyzing the mapping relationship between the two. .in, For state vectors, Cumulative energy consumption reflects the energy consumption during equipment operation; To assess the health status of equipment, including its aging and wear conditions; To accumulate carbon emissions and reflect the environmental impact of equipment operation. For real-time power input, It is a deterministic dynamic function that describes the operating behavior of the equipment under deterministic factors; The random perturbation matrix takes into account uncertainties such as environmental factors and accidental equipment failures. Let be the Wiener process vector, representing random noise. This state equation comprehensively describes the operational dynamics of the energy terminal equipment, providing a foundational model for subsequent decision-making.

[0111] For the state equation formula Let the input state vector be... ,in For cumulative energy consumption, For the health status of the equipment, For cumulative carbon emissions; control inputs Real-time power; deterministic dynamic function ,in This is the equipment degradation coefficient. Carbon emission coefficient; random perturbation matrix ,in These are the random disturbance coefficients for energy consumption, equipment health status, and carbon emissions, respectively. This is the Wiener process vector. Using this state equation, combined with the physical characteristics and operating environment of the equipment, the dynamic operation of the equipment is parametrically modeled, outputting a prediction of the equipment's state over a future period, including changes in energy consumption, evolution of equipment health status, and carbon emission growth. This yields a dynamic model of equipment operation, enabling accurate description and prediction of the operating state of energy terminal equipment, providing a foundation for subsequent control decisions.

[0112] (2) The performance functional definition unit is based on the state equation, defining a multi-objective performance functional that includes energy consumption, equipment lifespan, and carbon emissions. The performance functional expression is:

[0113] ,

[0114] in, The electricity price weighting parameter reflects the importance of electricity price in the objective function; The weighting parameter for equipment maintenance costs reflects the impact of equipment maintenance costs; It serves as a carbon emission weighting parameter, measuring the cost of carbon emissions; The terminal health constraint weight parameter emphasizes the importance of the device's health status; For real-time electricity prices, The regulation cycle is defined by this performance functional, which comprehensively considers economic costs (electricity prices, equipment maintenance costs), environmental impacts (carbon emissions), and the health status of the equipment itself, thus comprehensively measuring the overall benefits of energy terminal equipment operation.

[0115] For the performance functional expression:

[0116] ,

[0117] Let the input parameters be... For electricity price weighting parameters, As a weighting parameter for equipment maintenance costs, For carbon emission weighting parameters, For terminal health constraint weight parameters, For power, For real-time electricity prices, For the health status of the equipment, For carbon emissions, The regulation period is defined by this performance functional expression, which comprehensively considers factors such as energy consumption costs, equipment lifespan depreciation, and carbon emissions to output a comprehensive performance evaluation value. This value reflects the overall benefits of equipment operation throughout the entire regulation period, resulting in an index that can comprehensively measure the equipment's operating performance. This achieves multi-objective optimization evaluation of energy terminal equipment operation and provides a basis for finding the optimal control strategy.

[0118] (3) The stochastic Hamiltonian solving unit constructs a stochastic Hamiltonian function based on the performance functional expression and solves the minimum condition to obtain the control strategy. The stochastic Hamiltonian function is an important tool for solving stochastic optimal control problems. By solving its minimum condition, a control strategy that optimizes the performance functional under the consideration of stochastic factors can be obtained. This process involves complex mathematical derivation and solution. Using stochastic analysis theory and optimal control algorithms, the optimal values ​​of control variables such as real-time power input are determined, thereby realizing the optimized regulation of energy terminal equipment. Through the processing of the regulation decision module, a reasonable control strategy can be formulated under the consideration of multiple factors and stochastic disturbances, so as to realize the efficient, economical and environmentally friendly operation of energy terminal equipment.

[0119] S5. Control Execution Module: Used to generate and execute equipment control commands according to the control strategy. The control execution module includes a predictive control unit, a sliding mode control unit, a generation unit, and an execution unit. The predictive control unit is used to perform rolling optimization of the control strategy in the prediction time domain based on the model predictive control algorithm to obtain a reference power curve. The sliding mode control unit is used to design a sliding surface and an exponential reaching law to robustly compensate for unmodeled disturbances based on the deviation between the actual operating state of the energy terminal equipment and the expected state corresponding to the reference power curve, obtaining a robust compensation term. The sliding surface is defined as... ,in, For state tracking error, For a positive definite matrix, the exponential reaching law is: ,in, and The parameters are sliding mode control parameters. Unmodeled disturbances include equipment nonlinear dynamics and external environmental disturbances not included in the model predictive control algorithm. The generation unit is used to connect the reference power curve and the robust compensation term using adaptive weighting coefficients. Linear superposition generates equipment control commands. The execution unit converts the equipment control commands into analog signals through the digital-to-analog conversion module and transmits them to the equipment at the energy terminal, enabling the equipment at the energy terminal to execute the equipment control commands.

[0120] Specifically, the control execution module is used to generate and execute equipment control commands according to the control strategy. It consists of a predictive control unit, a sliding mode control unit, a generation unit, and an execution unit.

[0121] (1) The predictive control unit performs rolling optimization of the control strategy within the prediction time domain based on the model predictive control algorithm. The model predictive control algorithm predicts the operating status of the energy terminal equipment over a period of time based on the dynamic model of the equipment. In each control cycle, the unit optimizes the control strategy based on the actual operating status of the equipment and the predictive model. For example, considering factors such as the real-time power demand of the energy terminal equipment, the health status of the equipment, and changes in the external environment, the control strategy is continuously adjusted within the prediction time domain to obtain the optimal reference power curve. This rolling optimization method can respond promptly to various changes during equipment operation, improving the accuracy and adaptability of the control.

[0122] The sliding mode control unit addresses the deviation between the actual operating state of an energy terminal device and the desired state corresponding to a reference power curve. During actual operation, unmodeled disturbances, such as nonlinear dynamics of the device and external environmental interference, can cause the actual state to deviate from the desired state. The sliding mode control unit solves this problem by designing a sliding surface and an exponential reaching law. The sliding surface is defined as... ,in State tracking error, which is the difference between the actual state and the desired state. It is a positive definite matrix, which determines the rate and characteristics of the system state's convergence to the sliding surface. The exponential reaching law is... ,in and These are the sliding mode control parameters. By reasonably selecting these parameters, the system state can quickly and stably approach the sliding surface, and robust compensation can be performed on unmodeled disturbances to obtain robust compensation terms, thereby enhancing the system's resistance to uncertainties.

[0123] The generating unit uses adaptive weighting coefficients to combine the reference power curve with the robust compensation term. Linear superposition. Adaptive weighting coefficients. It will dynamically adjust based on factors such as the real-time operating status of the equipment and changes in the external environment. For example, when the equipment is operating relatively stably, the reference power curve will have a relatively large weight in the control command; while when the equipment is subjected to significant disturbances and the actual state deviates significantly from the expected state, the weight of the robust compensation term will increase accordingly. This will generate equipment control commands that ensure that the control commands can both follow the expected optimization strategy and effectively cope with sudden disturbances.

[0124] (2) The execution unit converts the equipment control commands into analog signals and transmits them to the equipment at the energy terminal through the digital-to-analog converter module. The digital-to-analog converter module converts the digital control commands into analog signals suitable for the equipment to receive, such as voltage or current signals, so that the equipment at the energy terminal can execute the equipment control commands, realize precise adjustment of the equipment operating status, and ensure the stable and efficient operation of the energy terminal equipment.

[0125] S6. State Feedback Module: This module collects the operating status of energy terminal equipment and feeds it back to the control and decision module, adaptively optimizing the parameters of the state equation. The state feedback module includes a state acquisition unit and a feedback optimization unit. The state acquisition unit collects real-time operating status data from the terminal equipment using sensors, including current, temperature, and vibration sensors. The operating status data includes power, temperature, and health index. The feedback optimization unit inputs the real-time state dataset into the multi-core tensor kernel function of the fusion analysis module, calculates the measured values ​​of the fusion features, predicts the predicted values ​​of the fusion features at the same time step based on the state equation of the control and decision module, calculates the deviation between the two, and dynamically adjusts the parameters of the multi-core tensor kernel function in the fusion analysis module and the parameters of the state equation in the control and decision module based on the deviation, forming a closed-loop correction mechanism. The formula for calculating the deviation is as follows: ,in, To fuse feature predictions, The measured values ​​are used to fuse features.

[0126] Specifically, the state feedback module is used to collect the operating status of energy terminal equipment and feed it back to the control decision module to adaptively optimize the parameters of the state equation. It consists of a state acquisition unit and a feedback optimization unit.

[0127] The status acquisition unit collects real-time operating status data from terminal equipment using current sensors, temperature sensors, vibration sensors, and other sensors, resulting in a real-time status dataset. These sensors are distributed across key components of the energy terminal equipment. Current sensors monitor changes in the equipment's current to obtain power information; temperature sensors sense the equipment's temperature in real time, reflecting its operating load and health status; and vibration sensors capture vibration signals, which can be used to determine if there are any mechanical faults or other anomalies. The collected operating status data includes power, temperature, and health indices, which constitute the real-time status dataset, providing a foundation for subsequent analysis.

[0128] The feedback optimization unit inputs the real-time state dataset into the multi-kernel tensor kernel function of the fusion analysis module. The multi-kernel tensor kernel function is a key tool in the fusion analysis module for processing multi-source heterogeneous data; it enables the fusion analysis of different types of state data. Through the calculation of the multi-kernel tensor kernel function, the measured values ​​of the fusion features are obtained. Simultaneously, based on the state equation of the control decision module, and using the currently known parameters and model, the predicted values ​​of the fusion features at the same time step are predicted. Here, the state equation is a mathematical model used by the control decision module to describe the dynamic operation of the equipment, containing key information such as the equipment's energy consumption, health status, and carbon emissions.

[0129] The formula for calculating the deviation between the predicted and measured values ​​of the fused features is as follows: ,in, To fuse feature predictions, The measured values ​​are used to fuse features. Based on this deviation, the parameters of the multi-core tensor kernel function in the fusion analysis module and the parameters of the state equation in the control decision module are dynamically adjusted. For example, if the deviation is large, it indicates that the current model parameters differ significantly from the actual situation. In this case, specific algorithms, such as gradient descent, are used to adjust the parameters of the multi-core tensor kernel function (such as the kernel function bandwidth) and the parameters of the state equation (such as the variance of the random perturbation matrix), enabling the model to better fit the actual data and forming a closed-loop correction mechanism. This closed-loop correction mechanism can continuously optimize the system model, improve the prediction and control accuracy of the operating status of energy terminal equipment, and make the entire energy terminal control system operate more stably and efficiently.

[0130] The data acquisition and processing module acquires multi-source heterogeneous data from energy terminals and transforms this data into manifold coordinate data using complex manifold coordinate transformation and probabilistic manifold metric calculations, laying the foundation for subsequent processing. The fusion analysis module performs tensor fusion analysis on the manifold coordinate data based on multi-kernel tensor kernel functions, generating low-dimensional fusion features that preserve geometric structure, effectively capturing nonlinear relationships between data and avoiding the data feature loss problem caused by traditional linear fusion. The control and decision-making module uses the low-dimensional fusion features to parametrically model the dynamic operation of the energy terminal equipment, constructing state equations including random disturbance terms. It also defines a multi-objective performance functional covering energy consumption, equipment lifespan, and carbon emissions. By solving the minimum condition of the stochastic Hamiltonian function, a control strategy is obtained, thus formulating a reasonable control strategy considering various influencing factors and uncertainties. The control execution module generates and executes equipment control commands according to the control strategy, ensuring the effective implementation of the strategy. The status feedback module collects the operating status of energy terminal equipment and feeds it back to the control and decision-making module. It adaptively optimizes the parameters of the state equation, forming a closed-loop correction mechanism. This allows the system to continuously adjust and optimize based on actual operating conditions, thereby achieving deep processing and fusion of multi-source heterogeneous data from energy terminals. It fully utilizes the effective information in the data and accurately reflects the true operating status of the energy terminals. Based on this, combined with control and decision-making considering multiple objectives and uncertainties, and the adaptive optimization mechanism, efficient adaptive control of energy terminals is achieved. This effectively solves the problem in traditional energy terminal multi-source data fusion analysis and adaptive control systems where insufficient data utilization and difficulty in accurately reflecting the true operating status of energy terminals hinder efficient management and control. It improves energy utilization efficiency, extends equipment lifespan, and provides strong support for intelligent management of energy terminals.

[0131] Example 2

[0132] This embodiment provides an energy terminal multi-source data fusion analysis and adaptive control system, including:

[0133] The data acquisition module is configured to acquire multi-source heterogeneous data from energy terminals;

[0134] The popular coordinate module is configured to convert multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation;

[0135] The feature fusion module is configured to perform tensor fusion analysis on manifold coordinate data based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure.

[0136] The control strategy module is configured to use low-dimensional fusion features to parametrically model the dynamic operation of the energy terminal equipment and obtain the control strategy by solving the minimum condition of the stochastic Hamiltonian function.

[0137] The control command module is configured to generate device control commands based on the control strategy.

[0138] The feedback module is configured to perform adaptive optimization based on the device operating status feedback under control commands.

[0139] A computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device, the aforementioned method for multi-source data fusion analysis and adaptive control of an energy terminal.

[0140] A terminal device includes a processor and a computer-readable storage medium, the processor being used to implement various instructions; the computer-readable storage medium being used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor as described in the method for multi-source data fusion analysis and adaptive control of an energy terminal.

[0141] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for multi-source data fusion analysis and adaptive control of energy terminals, characterized in that, include: Acquire multi-source heterogeneous data from energy terminals; Multi-source heterogeneous data is converted into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation; Tensor fusion analysis of manifold coordinate data is performed based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure. The dynamic operation of energy terminal equipment is parametrically modeled by utilizing low-dimensional fusion features, and the control strategy is obtained by solving the minimum condition of the stochastic Hamiltonian function. Generate equipment control commands based on the control strategy; Adaptive optimization is performed based on feedback of the device's operating status under control commands; The process of converting multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation includes performing a holomorphic mapping transformation on complex voltage phasor data. The transformation formula for the holomorphic mapping transformation is as follows: ， in, For complex voltage phasors, For the real part, The imaginary part is used; by converting the complex voltage phasor data into two independent real-valued components, the real part and the imaginary part, it is easier to process them in subsequent steps, thus realizing the formal transformation of the complex voltage phasor data. The process of converting multi-source heterogeneous data into manifold coordinate-based data through complex manifold coordinate transformation and probabilistic manifold metric calculation also includes constructing a Fisher information matrix as a Riemannian metric for the joint temperature and humidity distribution data. This Fisher information matrix is ​​then used as a metric tensor on the Riemannian manifold to map the joint temperature and humidity distribution data onto the Riemannian manifold, thus preserving the geometric structure of the data. The elements of the Fisher information matrix are represented as follows: ， in, For distribution parameters, Let be the joint probability density function. ; The tensor fusion analysis of manifold coordinate data based on multi-kernel tensor kernel functions includes constructing a joint data tensor from the manifold coordinate data and calculating the kernel matrix to obtain a multi-kernel tensor kernel matrix. The analysis unit performs centering and eigenvalue decomposition on the multi-kernel tensor kernel matrix to obtain eigenvectors and eigenvalues. The generation unit performs embedding mapping calculations on the eigenvalues ​​and eigenvectors based on the eigenvectors corresponding to the maximum number of preset eigenvalues ​​to generate low-dimensional fusion features that preserve geometric structure. The multi-kernel tensor kernel function is defined as the product of the kernel functions of each submanifold, expressed as: ， in, For the first A joint data tensor, For heterogeneous data types, For the first Submanifold kernel function, For the first The first Class data; The tensor fusion analysis of manifold coordinated data based on multi-kernel tensor kernel functions further includes centering and eigenvalue decomposition of the multi-kernel tensor kernel matrix to obtain eigenvectors and eigenvalues. Centering eliminates the influence of the mean in the data, allowing subsequent analysis to proceed with the data under zero mean. The eigenvalue problem solved by eigenvalue decomposition is defined as... ,in, For a centralized kernel matrix, For a centered matrix, For feature vectors, For eigenvalues, Given the number of data samples, the multi-kernel tensor matrix is ​​decomposed into eigenvectors and eigenvalues ​​through eigenvalue decomposition. The eigenvalues ​​reflect the variance of the data in different directions. The larger the variance, the richer the data information in that direction. The method of parametrically modeling the operational dynamics of energy terminal equipment using low-dimensional fusion features includes: based on the mapping relationship between low-dimensional fusion features and the physical characteristics of energy terminal equipment, establishing stochastic differential equations with feature parameterization to parametrically model the operational dynamics of energy terminal equipment. The low-dimensional fusion features retain key geometric information from multi-source heterogeneous data, and the physical characteristics of the equipment include its electrical and mechanical parameters. State equations are established by analyzing the mapping relationship. ,in, For state vectors, For cumulative energy consumption, For the health status of the equipment, To accumulate carbon emissions and reflect the environmental impact of equipment operation; For real-time power input, It is a deterministic dynamic function; The random perturbation matrix; For Wiener process vectors.

2. The method for multi-source data fusion analysis and adaptive control of energy terminals according to claim 1, characterized in that, The process of obtaining a control strategy by solving the minimum condition of a stochastic Hamiltonian function includes: defining a multi-objective performance functional based on the state equation, incorporating energy consumption, equipment lifespan, and carbon emissions; obtaining a performance functional expression containing the expected integral; constructing a stochastic Hamiltonian function based on the performance functional expression and solving the minimum condition to obtain the control strategy; and wherein the performance functional expression is: ， in, This is a parameter for electricity price weighting; As a weighting parameter for equipment maintenance costs; For carbon emission weighting parameters; For terminal health constraint weight parameters; For real-time electricity prices, For the regulation cycle.

3. The method for multi-source data fusion analysis and adaptive control of energy terminals according to claim 2, characterized in that, The process of generating equipment control commands based on the control strategy includes: using a model predictive control algorithm to perform rolling optimization of the control strategy in the prediction time domain to obtain a reference power curve; and designing a sliding mode surface and an exponential reaching law to robustly compensate for unmodeled disturbances by addressing the deviation between the actual operating state of the energy terminal equipment and the expected state corresponding to the reference power curve, resulting in a robust compensation term. The sliding mode surface is defined as... ,in, For state tracking error, For a positive definite matrix, the exponential reaching law is: ,in, and The sliding mode control parameters are used; finally, the reference power curve and the robust compensation term are compared using adaptive weighting coefficients. Linear superposition generates equipment control commands.

4. The method for multi-source data fusion analysis and adaptive control of energy terminals according to claim 3, characterized in that, The adaptive optimization based on device operating status feedback under control commands includes: collecting real-time operating status data of the terminal device through sensors to obtain a real-time status dataset; inputting the real-time status dataset into a multi-core tensor kernel function to calculate the measured value of the fused feature; predicting the fused feature value at the same time step based on the state equation; calculating the deviation between the two; and dynamically adjusting the parameters of the multi-core tensor kernel function and the parameters of the state equation based on the deviation to form a closed-loop correction mechanism. The formula for calculating the deviation is as follows: ,in, To fuse feature predictions, The measured values ​​are used to fuse features.

5. An energy terminal multi-source data fusion analysis and adaptive control system, employing the energy terminal multi-source data fusion analysis and adaptive control method as described in any one of claims 1-4, characterized in that, include: The data acquisition module is configured to acquire multi-source heterogeneous data from energy terminals; The popular coordinate module is configured to convert multi-source heterogeneous data into manifold coordinate data through complex manifold coordinate transformation and probabilistic manifold metric calculation; The feature fusion module is configured to perform tensor fusion analysis on manifold coordinate data based on multi-kernel tensor kernel functions to generate low-dimensional fusion features that preserve geometric structure. The control strategy module is configured to use low-dimensional fusion features to parametrically model the dynamic operation of the energy terminal equipment and obtain the control strategy by solving the minimum condition of the stochastic Hamiltonian function. The control command module is configured to generate device control commands based on the control strategy. The feedback module is configured to perform adaptive optimization based on the device operating status feedback under control commands.

Images