Distributed energy credible sensing method and system based on multi-source data fusion

Abstract

The invention discloses a distributed energy credible sensing method and system based on multi-source data fusion, and relates to the technical field of energy monitoring and evaluation, and the method comprises the steps: carrying out the fusion of distributed energy data through a hierarchical fusion structure; based on the fused distributed energy data, constructing a credibility measurement system from three dimensions of data, a model and a system by using a hierarchical dynamic credibility evaluation framework; based on a credibility measurement system, adopting a closed-loop feedback multi-target dynamic correction framework, and combining fusion data features and energy domain knowledge to carry out adaptive correction; the method can adapt to the dynamic change of the distributed energy system in real time, effectively cope with a complex and changeable operation environment, reduce the operation risk of the system, and improve the energy utilization efficiency and the overall stability of the system; the method provides powerful technical support for safe and efficient operation of a distributed energy system, and has remarkable theoretical value and wide application prospect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy monitoring and evaluation, and particularly relates to a distributed energy credible perception method and system based on multi-source data fusion. BACKGROUND

[0002] With the increasing emphasis on environmental protection and sustainable development worldwide, the energy structure is undergoing profound transformation. The traditional centralized energy supply mode has problems such as resource shortage and environmental pollution, and distributed energy is becoming increasingly important in the power system due to its advantages of being clean, efficient and flexible. It covers various forms such as solar photovoltaic and wind power, and can be directly installed on the user side or load center, reducing power transmission loss and improving energy utilization.

[0003] Power system stable operation needs to maintain real-time balance between power generation and power consumption, and large-scale access of distributed energy provides new regulation means for power auxiliary services such as frequency regulation, voltage regulation and backup, which can improve system reliability, stability and economy. However, the existing technology has many deficiencies in distributed energy monitoring and evaluation. Distributed energy system data is multi-source and heterogeneous, different devices and environmental factors will produce different data, and the existing data acquisition system is difficult to uniformly collect and integrate multi-source data, and cannot fully reflect the operating state. The credibility evaluation mechanism is also not perfect, lacking a scientific evaluation model, relying too much on simple empirical formulas or static indicators, and not fully considering the dynamic characteristics of energy and system interaction, and unable to adapt to changes in the energy system. In addition, the data processing and fusion capability is insufficient, and the collected data has quality problems such as outliers and missing values, and traditional data fusion algorithms are difficult to cope with the diversity, time-varying and uncertainty of distributed energy data, and cannot effectively mine data information. Although there are related researches at present, they mostly only focus on single aspect improvement, and cannot fully solve the key problems of multi-source data collection and integration, scientific credibility evaluation and data fusion. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] Therefore, the technical problem solved by the present application is: how to realize efficient collection and integration of multi-source and heterogeneous data to fully reflect the operating state of distributed energy; how to establish a scientific and reasonable credibility evaluation model to accurately quantify the reliability of distributed energy participating in power auxiliary services.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application embodiment provides a distributed energy credible perception method based on multi-source data fusion, comprising:

[0008] The distributed energy data is fused through a hierarchical fusion structure;

[0009] Based on the fused distributed energy data, a hierarchical dynamic credibility evaluation framework is used to construct a credibility measurement system from three dimensions of data, models and systems;

[0010] Based on the credibility measurement system, a closed-loop feedback multi-objective dynamic correction framework is adopted to combine the fusion data features and energy domain knowledge for adaptive correction;

[0011] For specific scenarios of distributed energy, a distributed energy correction mechanism is adopted, including a photovoltaic cloud impact compensator;

[0012] Specifically, for the photovoltaic output fluctuation caused by cloud layer change, a compensation algorithm is designed as follows:

[0013]

[0014] Wherein, P orig (t) and P corr (t) are the power values before and after correction at time t, G clear (t) is the clear sky irradiance model value, G meas (t) is the measured irradiance value, and C shad (t) is the cloud shadow compensation coefficient based on multi-source data fusion.

[0015] As a preferred scheme of the distributed energy credible perception method based on multi-source data fusion, wherein:

[0016] The distributed energy data is fused through the hierarchical fusion structure, including:

[0017] A three-layer fusion structure including data layer, feature layer and decision layer is constructed;

[0018] The data layer fusion includes:

[0019] For data of the same type and dimension, an improved state estimation equation is designed, expressed as:

[0020] x(k)=Φ(k|k-1)x(k-1)+G(k-1)w(k-1)+F(k-1)u(k-1)

[0021] Wherein, x(k) is the energy system state vector at time k, x(k-1) is the energy system state vector at time k-1; Φ(k|k-1) is the state transition matrix, G(k-1) is the noise gain matrix, w(k-1) is the process noise, F(k-1) is the control matrix, and u(k-1) is the control input vector.

[0022] The feature layer fusion includes:

[0023] A multi-modal feature extraction and fusion method based on neural network is designed for different types of but interrelated data, which is represented as:

[0024] f i =CNN i (X i )

[0025] F=Attention(f1,f2,...,f n )

[0026] Where X i is the i-th type of data, CNN i is the corresponding feature extraction network, f i is the extracted feature vector, Attention is a self-attention fusion function, and F is the fused feature.

[0027] As an optimal solution of the distributed energy credible perception method based on multi-source data fusion, wherein:

[0028] The distributed energy data is fused through the hierarchical fusion structure, and the method further comprises:

[0029] A time correlation modeling method is proposed for the multi-scale time characteristics of distributed energy data, and the multi-scale time correlation modeling equation is represented as:

[0030]

[0031] Where x(k) represents the energy system state vector at time k, m is the different time scale, τ i is the time interval corresponding to the time scale m, Φ i is the corresponding state transition matrix, α i is the time scale weight coefficient, which is determined by online learning, and w(k) is the process noise vector.

[0032] As an optimal solution of the distributed energy credible perception method based on multi-source data fusion, wherein:

[0033] The distributed energy data is fused through the hierarchical fusion structure, and the method further comprises:

[0034] The physical model and prior knowledge of distributed energy are introduced into the fusion process, and a knowledge-constrained correction mechanism is designed, which is represented as:

[0035]

[0036] Where, represents the corrected system state estimate, is the algorithm estimate, For the estimation value based on the physical model, λ is a dynamically adjusted weight coefficient.

[0037] As a preferred solution of the distributed energy trusted perception method based on multi-source data fusion, wherein:

[0038] The hierarchical dynamic trustworthiness evaluation framework based on the fused distributed energy data includes:

[0039] Based on the uncertainty characteristics of the fused data, information entropy is introduced to quantify the data trustworthiness;

[0040] Based on the time-frequency domain characteristics of the fused data, multi-scale fluctuation feature trustworthiness indicators are introduced;

[0041] Based on the deep features of the fused distributed energy data, an anomaly detection model based on deep autoencoder is designed.

[0042] As a preferred solution of the distributed energy trusted perception method based on multi-source data fusion, wherein:

[0043] The closed-loop feedback multi-objective dynamic correction framework is adopted based on the trustworthiness evaluation system, combined with the fusion data features and energy domain knowledge for adaptive correction, including:

[0044] Based on the trustworthiness evaluation system, a hierarchical correction strategy is designed to take differentiated correction measures for different trustworthiness levels, including:

[0045] High-level basic model structure correction:

[0046] When the trustworthiness evaluation shows that the system performance is in a long-term low level, the model structure correction is triggered:

[0047]

[0048] Wherein, M * is the final selected optimal model, is the candidate model set, D train represents the training data set, is the training loss function, C is the model complexity measure, Q total is the comprehensive trustworthiness evaluation result, and α and β are the trade-off parameters;

[0049] Middle-level model parameter correction: when the specific dimension trustworthiness indicators fluctuate, the corresponding parameter correction is triggered;

[0050] Low-level real-time data correction: for the sudden anomalies in the fused data, a real-time correction mechanism is designed.

[0051] As a preferred solution of the distributed energy trusted perception method based on multi-source data fusion, wherein:

[0052] The closed-loop feedback multi-objective dynamic correction framework is adopted based on the credibility measurement system, and self-adaptive correction is performed in combination with the fusion data features and the energy domain knowledge, including:

[0053] The fusion weight is dynamically adjusted based on the performance after the fusion of different data sources;

[0054] According to the fusion evaluation feedback, the parameters of the multi-scale time correlation model are dynamically adjusted.

[0055] In the second aspect, the embodiments of the present application provide a distributed energy trusted perception system based on multi-source data fusion, including:

[0056] The fusion module is configured to fuse the distributed energy data through the hierarchical fusion structure;

[0057] The credibility evaluation module is configured to construct a credibility measurement system from three dimensions of data, model and system based on the fused distributed energy data by using a hierarchical dynamic credibility evaluation framework;

[0058] The self-adaptive correction module is configured to adopt a closed-loop feedback multi-objective dynamic correction framework based on the credibility measurement system, and perform self-adaptive correction in combination with the fusion data features and the energy domain knowledge;

[0059] For the specific scenario of the distributed energy, a distributed energy correction mechanism is adopted, including a photovoltaic cloud influence compensator;

[0060] Specifically, for the photovoltaic output fluctuation caused by the change of the cloud layer, a compensation algorithm is designed and represented as:

[0061]

[0062] Wherein, P orig (t) and P corr (t) are the power values before and after correction at time t, G clear (t) is the clear sky irradiance model value, G meas (t) is the measured irradiance value, and C shad (t) is a cloud shadow compensation coefficient based on multi-source data fusion.

[0063] In the third aspect, the embodiments of the present application provide a computing device, including:

[0064] A memory and a processor;

[0065] The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method for distributed energy trusted perception based on multi-source data fusion according to any one of the embodiments of the present application.

[0066] In a fourth aspect, the embodiments of the present application provide a computer readable storage medium storing computer executable instructions, and the computer executable instructions are executed by a processor to implement the method for distributed energy trusted perception based on multi-source data fusion.

[0067] The multi-dimensional data acquisition system designed by the present application can comprehensively collect various types of data of distributed energy, and the preprocessing algorithm ensures high-quality data, laying a foundation for subsequent analysis; the hierarchical adaptive multi-source data fusion framework solves the problem of insufficient energy data for traditional algorithms. The data layer fusion considers the influence of external scheduling instructions; the feature layer fusion designs a multi-modal network and a self-attention mechanism for heterogeneous data; the decision layer fusion improves the accuracy under energy prediction conflicts. Multi-scale time correlation modeling captures different time scale characteristics, uncertainty quantization and propagation methods optimize non-Gaussian and multi-peak characteristics processing, and energy domain knowledge constraint fusion ensures that the results conform to the physical law; the hierarchical dynamic reliability evaluation framework constructs a measurement system from three dimensions of data, model and system. The reliability is quantified using the fusion data features, the prediction error distribution and probability calibration are considered to evaluate the model reliability, the system response reliability is evaluated from dynamic tracking and response stability, and the Bayesian network comprehensively evaluates the model to make the results more objective and comprehensive; the closed-loop feedback multi-objective dynamic correction framework realizes adaptive correction combined with various information, and the reliability-driven hierarchical correction strategy optimizes the model structure, parameters and real-time data; the multi-source data fusion feedback correction realizes self-optimization of the fusion process; the energy scene specialized correction algorithm provides special strategies for photovoltaic and wind power problems; the federated learning enhanced distributed correction framework realizes multi-site collaborative optimization and protects data privacy. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0069] Figure 1 is the overall flowchart of the method for distributed energy trusted perception based on multi-source data fusion according to the present application. DETAILED DESCRIPTION

[0070] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0071] Embodiment 1, refer to Figure 1 For the first embodiment of the present application, the embodiment provides a distributed energy trusted perception method based on multi-source data fusion, comprising:

[0072] S1: fusing the distributed energy data through a hierarchical fusion structure;

[0073] S2: based on the fused distributed energy data, using a hierarchical dynamic trustworthiness evaluation framework to construct a trustworthiness measurement system from three dimensions of data, model and system;

[0074] S3: based on the trustworthiness measurement system, using a closed-loop feedback multi-objective dynamic correction framework to combine the fusion data features and energy domain knowledge for adaptive correction.

[0075] It should be noted that through steps S1-S3, the embodiment effectively fuses the distributed energy data through a hierarchical fusion structure, solving the compatibility and efficiency problems of multi-source data in the integration process. On this basis, the hierarchical dynamic trustworthiness evaluation framework is used to construct a trustworthiness measurement system from three dimensions of data, model and system, comprehensively and accurately evaluating the trustworthiness of each element in the distributed energy system, providing a reliable basis for subsequent correction work; the use of a closed-loop feedback multi-objective dynamic correction framework, combined with fusion data features and energy domain knowledge for adaptive correction, further improves the accuracy and reliability of the state perception of the distributed energy system.

[0076] Embodiment 2, refer to Figure 1 For an embodiment of the present application, based on the previous embodiment, a distributed energy trusted perception method based on multi-source data fusion is provided, comprising:

[0077] In this embodiment, the fusion of distributed energy data through a hierarchical fusion structure in the above step S1 comprises:

[0078] It should be noted that before fusing the distributed energy data through a hierarchical fusion structure, the multi-dimensional data of different types of distributed energy are first collected. Taking a photovoltaic power generation system as an example, the collected data includes:

[0079] Energy output data: real-time power, voltage, current, frequency, etc.

[0080] Environmental parameters: solar irradiance, temperature, humidity, wind speed, etc.

[0081] Device status: inverter operating state, photovoltaic panel temperature, dust coverage, etc.

[0082] Historical operation data: power generation, efficiency, fault records, etc.

[0083] External information: weather forecast, electricity price information, grid dispatching instructions, etc.

[0084] The collected multi-dimensional data are preprocessed, mainly including:

[0085] Data cleaning: remove outliers and redundant data.

[0086] Data completion: use interpolation or machine learning methods to fill in missing data.

[0087] Data standardization: unify different types and dimensions of data to the same scale.

[0088] Anomaly detection: identify abnormal data based on statistical methods or machine learning algorithms.

[0089] The hierarchical fusion structure includes three layers: data layer fusion, feature layer fusion and decision layer fusion, and is equipped with a real-time adaptive adjustment mechanism. The specific algorithm is as follows:

[0090] Data layer fusion:

[0091] For the same type and dimension of data (such as multiple sensors measuring the same physical quantity), an improved state estimation equation is designed, which is expressed as:

[0092] x(k) = Φ(k|k-1)x(k-1) + G(k-1)w(k-1) + F(k-1)u(k-1) (1)

[0093] Where x(k) is the state vector of the energy system at time k (such as photovoltaic output power, energy storage SOC, etc.), x(k-1) is the state vector of the energy system at time k-1; Φ(k|k-1) is the state transition matrix, G(k-1) is the noise gain matrix, w(k-1) is the process noise, F(k-1) is the control matrix, and u(k-1) is the control input vector (such as dispatching instructions).

[0094] It should be noted that the introduction of the control input vector u(k-1) specific to distributed energy can effectively represent the influence of external dispatching instructions on the state of the energy system, solving the modeling deficiency problem of traditional Kalman filtering in handling controlled energy systems.

[0095] Feature layer fusion:

[0096] A multi-modal feature extraction and fusion method based on neural network is designed for different types of but interrelated data (such as irradiance, temperature and photovoltaic output power), which is represented as:

[0097] f i =CNN i (X i ) (2)

[0098] F=Attention(f1,f2,...,f n ) (3)

[0099] Wherein, X i is the i-th data, CNN i is the corresponding feature extraction network, f i is the extracted feature vector, Attention is a self-attention fusion function, and F is the fused feature.

[0100] It should be noted that the multi-modal feature extraction network structure is designed for the characteristics of energy data, and the self-attention mechanism is introduced to dynamically adjust the weight of different data sources, which solves the problem of insufficient processing capacity of heterogeneous energy data in traditional methods.

[0101] Decision layer fusion:

[0102] A decision layer fusion algorithm based on D-S evidence theory is designed for multiple prediction and evaluation results, which is represented as:

[0103]

[0104] Formula (4) and formula (5) respectively represent the basic probability assignment composition rule and the belief degree calculation formula in D-S evidence theory, wherein, m i (A) is the basic probability assignment of the i-th evidence source to the hypothesis A, w i is the weight of the evidence source, Bel(A) is the belief degree of the hypothesis A, and the summation range is all subsets B of A.

[0105] It should be noted that this step provides a hierarchical adaptive multi-source data fusion framework, which is specially designed for the characteristics of distributed energy systems, and solves the shortcomings of traditional fusion algorithms in dealing with the diversity, time-varying and uncertainty of energy data; In view of the uncertainty characteristics of distributed energy, the evidence conflict measure and dynamic weight adjustment mechanism are introduced, which significantly improves the fusion accuracy in the case of energy prediction conflict.

[0106] In the process of fusing distributed energy data through hierarchical fusion structure, a time correlation modeling method is proposed for the multi-scale time characteristics (second-level fluctuation, minute-level change, hour-level trend, day-level cycle) of distributed energy data. The multi-scale time correlation modeling equation is expressed as:

[0107]

[0108] where m is the different time scale (such as 1 minute, 10 minutes, 1 hour, etc.), τ i is the time interval corresponding to the time scale m, Φ i is the corresponding state transition matrix, α i is the time scale weight coefficient, determined by online learning, and w(k) is the process noise vector.

[0109] It should be noted that the multi-scale time correlation modeling breaks through the limitation of traditional fusion algorithm considering only single time scale, which can capture both short-term fluctuations and long-term trends of distributed energy, and improve the prediction accuracy.

[0110] In the process of fusing distributed energy data through hierarchical fusion structure, an uncertainty quantification and propagation method based on improved particle filtering is designed for the inherent randomness and uncertainty of distributed energy, expressed as:

[0111]

[0112] where, is the i-th particle, is the corresponding weight, and δ is the Dirac function.

[0113] It should be noted that the uncertainty quantification and propagation designs adaptive particle resampling strategy and multi-modal distribution modeling method, which is optimized for non-Gaussian, multi-peak characteristics of renewable energy such as photovoltaic and wind power, effectively preventing the weight degradation and particle poverty problems in traditional methods.

[0114] In the process of fusing distributed energy data through hierarchical fusion structure, the physical model and prior knowledge of distributed energy are introduced into the fusion process, and a knowledge-constrained correction mechanism is designed:

[0115]

[0116] where, represents the corrected system state estimation value, is the algorithm estimation value, is the estimation value based on the physical model, and λ is the dynamically adjusted weight coefficient.

[0117] It should be noted that the energy domain knowledge constraint fusion formalizes the professional knowledge in the energy field into constraint conditions into the data fusion process, and ensures that the fusion result conforms to the physical law of the energy system, such as photovoltaic power generation efficiency constraint, wind power curve constraint, etc.

[0118] In the embodiment, the step S2 of constructing the trustworthiness measurement system from the three dimensions of data, model and system based on the fused distributed energy data by using the hierarchical dynamic trustworthiness assessment framework includes:

[0119] The embodiment designs a hierarchical dynamic trustworthiness assessment framework (HDTAF), which fully utilizes the characteristic information of the fused data, constructs a trustworthiness measurement system from the three dimensions of data, model and system, and realizes accurate quantification of the trustworthiness of the distributed energy participating in the power auxiliary service.

[0120] Specifically, unlike the traditional simple weighted evaluation, the HDTAF constructs the trustworthiness evaluation index by using the characteristic information of the fused data:

[0121] Based on the uncertainty characteristics of the fused data, the information entropy is introduced to quantify the data trustworthiness:

[0122]

[0123] Wherein, H(X t ) is the information entropy of the fused data at time t, H max is the maximum possible entropy value, and p(x i ) is the probability density function of the data distribution, which is extracted from the fused data by kernel density estimation.

[0124] It should be noted that the application of information entropy theory to distributed energy trustworthiness evaluation can directly utilize the probability distribution characteristics of the fused data to quantify its uncertainty, and provides an information theory basis for trustworthiness evaluation.

[0125] Based on the time-frequency domain characteristics of the fused data, the multi-scale fluctuation feature trustworthiness index is introduced:

[0126]

[0127] Wherein, H(X t ) is the fluctuation variance of the fused data at time t in the jth time scale, which is extracted from the fused data by wavelet decomposition; is a reference variance threshold; w j is a time scale weight, satisfying

[0128] It should be noted that the introduction of the multi-scale fluctuation credibility index breaks through the limitations of traditional single time scale evaluation and can capture the stability characteristics of distributed energy at different time scales, especially suitable for new energy systems with multi-period characteristics.

[0129] Based on the deep features of the fusion data, an anomaly detection model based on deep autoencoder is designed:

[0130]

[0131] Where E() and D() are the encoder and decoder functions, respectively, which are trained by fusion data; is the reconstruction error, and θ is the normalization parameter.

[0132] It should be noted that the anomaly detection model based on deep autoencoder combines deep learning with credibility evaluation, which can automatically learn the normal operation mode of distributed energy, identify complex anomalies that are difficult to capture through simple rules, and improve the evaluation accuracy.

[0133] Based on the verification results of the prediction model based on fusion data, a dynamic evaluation method of model credibility is proposed, including:

[0134] The actual value and the predicted value are compared using the fusion data feedback to construct the prediction error distribution credibility:

[0135]

[0136] Where f err (e,t) is the prediction error probability density function at time t, which is extracted from the historical error sequence by kernel density estimation; δ is the acceptable error range.

[0137] It should be noted that the prediction error distribution credibility not only considers the size of the prediction error, but also considers the probability distribution characteristics of the error, which can effectively distinguish between random errors and systematic deviations, and provide more detailed basis for model credibility evaluation.

[0138] Based on the probability calibration of the prediction model verified by the fusion data, a probability calibration credibility index is proposed:

[0139]

[0140] Where F emp (p,t) is the empirical cumulative distribution function at time t, which represents the actual occurrence frequency when the prediction probability is p.

[0141] It should be noted that the introduction of the probability calibration theory into the credibility evaluation of distributed energy can evaluate the reliability of the prediction probability output by the model, and is especially suitable for power auxiliary service scenarios that require risk assessment.

[0142] Based on the fusion data feedback of the system scheduling response of the distributed energy, a system response credibility evaluation method is designed, including:

[0143] A dynamic tracking error credibility is introduced, which is expressed as:

[0144]

[0145] Wherein, r(τ) is a scheduling instruction reference trajectory, y(τ) is an actual response trajectory (extracted from fusion data), γ is a scaling parameter, and T is a sliding window length.

[0146] It should be noted that the control system tracking performance evaluation method is introduced into the credibility evaluation, which can accurately quantify the dynamic performance of the distributed energy response to the scheduling instruction, and provide quantitative basis for the credibility evaluation of auxiliary services.

[0147] Based on the spectral characteristics of the fusion data, a response stability credibility index is designed:

[0148]

[0149] Wherein, S y (f,t) is the power spectral density function of the response signal at time t (calculated by fusion data), [f min ,f max ] is the unstable frequency range of interest.

[0150] It should be noted that the system stability is evaluated by using the frequency domain characteristics of the response signal, which can effectively identify the potential unstable oscillation mode, and provide early warning information for the safety evaluation of power grid auxiliary services.

[0151] In the embodiment, the step S3 based on the credibility measurement system adopts a closed-loop feedback multi-objective dynamic correction framework, and combines the fusion data characteristics and energy domain knowledge for adaptive correction, including:

[0152] The embodiment proposes a closed-loop feedback multi-objective dynamic correction framework (CMDCF), which is not simply applied to conventional optimization algorithm, but organically combines credibility evaluation results, fusion data characteristics and energy domain knowledge, and constructs an adaptive correction mechanism specially for distributed energy characteristics.

[0153] Specifically, based on the credibility measurement system, a hierarchical correction strategy is designed, which takes differentiated correction measures for different credibility levels, including:

[0154] High-level basic model structure correction:

[0155] When the credibility evaluation shows that the system performance is in a long-term low level, trigger model structure correction:

[0156]

[0157] where, is the candidate model set, is the training loss function, c is the model complexity measure, Q total is the comprehensive credibility evaluation result, and a and β are weighting parameters.

[0158] It should be noted that the credibility evaluation result is directly introduced into the model selection process to form a feedback regulation mechanism, so that the model structure can be adaptively optimized according to the system performance, solving the poor adaptability problem caused by fixed model structure in traditional methods.

[0159] Middle-level model parameter correction:

[0160] When the specific dimension credibility indicator fluctuates, trigger corresponding parameter correction:

[0161]

[0162] where, t is the model parameter at time t, is the parameter gradient, Q dim (t) is the credibility value of the corresponding dimension, and f() is the adaptive mapping function.

[0163] It should be noted that an adaptive learning rate adjustment mechanism for credibility feedback is designed, so that the parameter update speed and the credibility level form an inverse relationship, accelerating parameter adjustment when the credibility is low and improving system response speed.

[0164] Low-level real-time data correction:

[0165] A real-time correction mechanism is designed for sudden abnormalities in the fusion data:

[0166]

[0167] where, is the original estimated value, is the corrected estimated value, and Δx(t) is the correction amount, Q anom (t) is the anomaly detection credibility.

[0168] It should be noted that the correction strength is dynamically associated with the abnormal credibility, realizing a soft correction strategy and avoiding the over-correction or under-correction problem in traditional methods.

[0169] The multi-source data fusion result is fed back to the fusion process itself to realize adaptive optimization, including:

[0170] Based on the performance of different data sources after fusion, the fusion weight is dynamically adjusted, which is expressed as:

[0171]

[0172] Where w i (t) is the fusion weight of the i-th data source, Q i (t) is the local credibility of the data source, is the average credibility, σ Q (t) is the credibility standard deviation, and γ is the learning rate.

[0173] It should be noted that a weight adjustment strategy similar to AdaBoost is designed, so that high credibility data sources obtain higher weights, and low credibility data sources gradually reduce the weight, realizing the self-optimization of the fusion process.

[0174] According to the fusion evaluation feedback, the parameters of the multi-scale time correlation model are dynamically adjusted, which is expressed as:

[0175]

[0176] Where α i (t) is the weight coefficient of the time scale τ i , is the sensitivity of credibility to weight, which is calculated by automatic differentiation.

[0177] It should be noted that dynamically adjusting the parameters of the multi-scale time correlation model establishes an explicit correlation between credibility and multi-scale model parameters, realizes gradient-based adaptive optimization, and solves the limitations of traditional fixed time scale models.

[0178] For the specific scenario of distributed energy, the distributed energy correction mechanism is proposed, including:

[0179] Photovoltaic cloud influence compensator:

[0180] For the photovoltaic output fluctuation caused by the change of the cloud layer, a compensation algorithm is designed, which is expressed as:

[0181]

[0182] Where P orig (t) and P corr (t) are the power values before and after correction at time t, G clear (t) is the clear sky irradiance model value, G meas (t) is the measured irradiance value, and C shad (t) is the cloud shadow compensation coefficient based on multi-source data fusion.

[0183] It should be noted that the physical model is combined with the data-driven method to provide a special optimization strategy for the cloud shadow problem specific to photovoltaic power generation.

[0184] Wind power cut-out protection correction:

[0185] For strong wind cut-out risk, a prediction correction algorithm is designed, represented as:

[0186]

[0187] Where Ψ() is the wind speed cut-out risk function, trained by fusion data; v pred (t) is the predicted wind speed, v cut is the cut-out wind speed, σ v (t) is the wind speed prediction uncertainty.

[0188] It should be noted that the safety constraints specific to the wind farm are integrated into the correction process, which significantly improves the prediction accuracy under extreme weather conditions.

[0189] The federal learning technology is introduced into the distributed energy correction mechanism to realize multi-site collaborative optimization under the premise of protecting data privacy, represented as:

[0190]

[0191] Where θ k is the local model parameter of the kth energy site, Δθ aggr is the aggregated gradient, and λ is the learning rate.

[0192] It should be noted that the introduction of federal learning technology into the distributed energy correction mechanism breaks through the limitations of traditional single-site correction, enabling collaborative optimization of multiple distributed energy sites in the region, while protecting sensitive data through differential privacy technology.

[0193] Embodiment 3, the above is a schematic scheme of the distributed energy trusted perception method based on multi-source data fusion of the present embodiment. It should be noted that the technical scheme of the distributed energy trusted perception system based on multi-source data fusion belongs to the same concept as the technical scheme of the distributed energy trusted perception method based on multi-source data fusion described above. The technical scheme of the distributed energy trusted perception system based on multi-source data fusion in the present embodiment is not described in detail. Details can be seen from the description of the technical scheme of the distributed energy trusted perception method based on multi-source data fusion described above.

[0194] The present embodiment also provides a system for a distributed energy trusted perception method based on multi-source data fusion, comprising:

[0195] The fusion module is configured to fuse the distributed energy data through a hierarchical fusion structure.

[0196] a credibility evaluation module configured to construct a credibility measurement system from three dimensions of data, model and system based on the fused distributed energy data by using a hierarchical dynamic credibility evaluation framework;

[0197] an adaptive correction module configured to perform adaptive correction by combining fusion data features and energy domain knowledge based on the credibility measurement system using a closed-loop feedback multi-objective dynamic correction framework;

[0198] For specific scenarios of distributed energy, a distributed energy correction mechanism is adopted, including a photovoltaic cloud influence compensator.

[0199] Specifically, for photovoltaic output fluctuation caused by cloud layer change, a compensation algorithm is designed as follows:

[0200]

[0201] wherein, P orig (t) and P corr (t) are power values before and after correction at time t, G clear (t) is a clear sky irradiance model value, G meas (t) is a measured irradiance value, and C shad (t) is a cloud shadow compensation coefficient based on multi-source data fusion.

[0202] The embodiment also provides a computing device suitable for the case of the distributed energy credibility perception method based on multi-source data fusion, comprising:

[0203] a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the distributed energy credibility perception method based on multi-source data fusion as proposed in the above embodiment.

[0204] The embodiment also provides a storage medium having a computer program stored thereon, and the program is executed by a processor to implement the distributed energy credibility perception method based on multi-source data fusion as proposed in the above embodiment.

[0205] The storage medium proposed in the embodiment and the distributed energy credibility perception method based on multi-source data fusion proposed in the above embodiment belong to the same inventive concept, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0206] Embodiment 4, referring to Table 1, provides a distributed energy credibility perception method based on multi-source data fusion as an embodiment of the application, and in order to verify the beneficial effects of the application, scientific demonstration is performed through simulation experiments.

[0207] A simulation environment based on MATLAB is constructed, and the actual operation data of a 10 MW distributed photovoltaic power station are selected for testing. The simulation process includes four main links: data preprocessing, multi-source data fusion, credibility evaluation, and dynamic correction. The prediction performance and accuracy of the credibility evaluation of the system under different weather conditions (sunny, cloudy, overcast, and weather mutation days) are analyzed. In the simulation, we apply an improved multi-source data fusion algorithm to process heterogeneous data from 7 irradiance sensors, 12 temperature sensors, and 3 power measurement points, use a hierarchical dynamic credibility evaluation framework to calculate the system credibility index, and optimize the prediction results through a closed-loop feedback multi-objective dynamic correction mechanism. The simulation results are shown in Table 1:

[0208] Table 1 Comparison of multi-weather indicators between the proposed method and traditional methods for a 10 MW distributed photovoltaic power station

[0209]

[0210]

[0211] The simulation analysis results show that compared with traditional methods, the proposed framework has significant advantages under various weather conditions, especially under complex weather conditions. It can achieve accurate power prediction and give reasonable credibility evaluation, providing a reliable guarantee for the participation of distributed energy in power system operation.

[0212] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and they should be included in the scope of the claims of the present application.

Claims

1. A distributed energy trusted sensing method based on multi-source data fusion, characterized in that, include: Distributed energy data is fused through a hierarchical fusion structure; Based on the fused distributed energy data, a credibility measurement system is constructed from three dimensions: data, model, and system using a hierarchical dynamic credibility assessment framework. Based on the credibility measurement system, a closed-loop feedback multi-objective dynamic correction framework is adopted, which combines fused data features and energy domain knowledge for adaptive correction. For specific scenarios involving distributed energy resources, distributed energy correction mechanisms are adopted, including photovoltaic cloud impact compensators. Specifically, to address the fluctuations in photovoltaic output caused by cloud cover changes, a compensation algorithm is designed as follows: Among them, P orig (t) and P corr (t) represents the power values ​​before and after correction at time t, respectively. clear (t) represents the clear-sky irradiance model value, G meas (t) represents the measured irradiance value, C shad (t) represents the cloud shadow compensation coefficient based on multi-source data fusion.

2. The distributed energy trusted sensing method based on multi-source data fusion as described in claim 1, characterized in that, The fusion of distributed energy data through a hierarchical fusion structure includes: Construct a three-layer fusion structure comprising a data layer, a feature layer, and a decision layer; Data layer fusion includes: For data of the same type and dimensions, an improved state estimation equation is designed, expressed as: x(k)=Φ(k|k-1)x(k-1)+G(k-1)w(k-1)+F(k-1)u(k-1) Where x(k) is the energy system state vector at time k, x(k-1) is the energy system state vector at time k-1; Φ(k|k-1) is the state transition matrix, G(k-1) is the noise gain matrix, w(k-1) is the process noise, F(k-1) is the control matrix, and u(k-1) is the control input vector; Feature layer fusion includes: For different types of interrelated data, a multimodal feature extraction and fusion method based on neural networks is designed, as follows: f i =CNN i (X i ) F=Attention(f1,f2,...,f n ) Among them, X i For the i-th class of data, CNN i For the corresponding feature extraction network, f i Let F be the extracted feature vector, Attention be the self-attention fusion function, and F be the fused features.

3. The distributed energy trusted sensing method based on multi-source data fusion as described in claim 2, characterized in that, The fusion of distributed energy data through a hierarchical fusion structure also includes: To address the multi-scale temporal characteristics of distributed energy data, a temporal correlation modeling method is proposed. The multi-scale temporal correlation modeling equation is expressed as: Where x(k) represents the energy system state vector at time k, m represents different time scales, and τ i Φ is the time interval corresponding to the time scale m. i Let α be the corresponding state transition matrix. i The time-scale weighting coefficients are determined through online learning, and w(k) is the process noise vector.

4. The distributed energy trusted sensing method based on multi-source data fusion as described in claim 3, characterized in that, The fusion of distributed energy data through a hierarchical fusion structure also includes: By incorporating the physical model and prior knowledge of distributed energy into the fusion process, a correction mechanism for knowledge constraints is designed, expressed as follows: in, This represents the corrected system state estimate. The algorithm estimates the value. λ is the estimated value based on the physical model, and λ is the dynamically adjusted weighting coefficient.

5. The distributed energy trusted sensing method based on multi-source data fusion as described in claim 4, characterized in that, The aforementioned credibility measurement system, based on the fused distributed energy data and utilizing a hierarchical dynamic credibility assessment framework, constructs a credibility measurement system from three dimensions: data, model, and system. Based on the uncertainty characteristics of fused data, information entropy is introduced to quantify data credibility. Based on the time-frequency domain features of fused data, a multi-scale fluctuation feature credibility index is introduced. Based on the deep features of the fused distributed energy data, an anomaly detection model based on a deep autoencoder is designed.

6. The distributed energy trusted sensing method based on multi-source data fusion as described in claim 5, characterized in that, The aforementioned adaptive correction based on a credibility measurement system, employing a closed-loop feedback multi-objective dynamic correction framework, and combining fused data features and energy domain knowledge includes: Based on a credibility measurement system, a hierarchical correction strategy is designed, employing differentiated correction measures for different credibility levels, including: High-level basic model structural correction: When credibility assessments show that system performance remains at a low level for an extended period, model structure correction is triggered. Among them, M * The optimal model is ultimately selected. D is the set of candidate models. train Represents the training dataset. To train the loss function, Q is a measure of model complexity. total To comprehensively assess the credibility, α and β are trade-off parameters; Mid-level model parameter correction: When the credibility index of a specific dimension fluctuates, the corresponding parameter correction is triggered; Low-level real-time data correction: A real-time correction mechanism is designed to address sudden anomalies that occur in the fused data.

7. The distributed energy trusted sensing method based on multi-source data fusion as described in claim 6, characterized in that, The adaptive correction based on the credibility measurement system, employing a closed-loop feedback multi-objective dynamic correction framework, and combining fused data features and energy domain knowledge, also includes: Dynamically adjust the fusion weights based on the performance of the fused data sources; Based on the feedback from the fusion assessment, the parameters of the multi-scale time correlation model are dynamically adjusted.

8. A distributed energy trusted sensing system based on multi-source data fusion, employing the method described in any one of claims 1 to 7, characterized in that, include: The fusion module is used to fuse distributed energy data through a hierarchical fusion structure; The credibility assessment module is used to construct a credibility measurement system from three dimensions—data, model, and system—based on the fused distributed energy data and using a hierarchical dynamic credibility assessment framework. The adaptive correction module is used to perform adaptive correction based on a credibility measurement system, using a closed-loop feedback multi-objective dynamic correction framework, and combining fused data features and energy domain knowledge. For specific scenarios involving distributed energy resources, distributed energy correction mechanisms are adopted, including photovoltaic cloud impact compensators. Specifically, to address the fluctuations in photovoltaic output caused by cloud cover changes, a compensation algorithm is designed as follows: Among them, P orig (t) and P corr (t) represents the power values ​​before and after correction at time t, respectively. clear (t) represents the clear-sky irradiance model value, G meas (t) represents the measured irradiance value, C shad (t) represents the cloud shadow compensation coefficient based on multi-source data fusion.

9. A computing device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images