A power grid state evaluation method and system applying multi-source data fusion

Abstract

The application discloses a kind of power grid state evaluation method and system of application multi-source data fusion, it is related to data analysis technical field, including the expansion of the shaft section to be evaluated in combination with power grid dynamic characteristics and correlation degree, the power grid state under the shaft section to be evaluated is often related with the data under previous time, in combination with power grid dynamic characteristics and correlation degree to determine a reasonable forward time node, data expansion is carried out to the shaft section to be evaluated, the comprehensiveness and rationality of data are guaranteed, and reliable foundation is provided for subsequent multi-source information feature extraction.The uncertain information is fused by combination rule, and the problem of misjudgment of boundary state by traditional threshold method is solved.The conflict between multiple sources of information can be handled by evidence theory, and the evaluation result is dominated by single data source by discount coefficient and dynamic evidence weight.The accuracy and adaptability of power grid state evaluation are improved, and the rationality of power grid state monitoring and the stability and safety of power grid operating state are guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and particularly relates to a power grid state evaluation method and system applying multi-source data fusion. BACKGROUND

[0002] With the evolution of new power systems to "double high" (high proportion of renewable energy, high proportion of power electronic equipment) and "double random" (load randomness, new energy output randomness), the power grid operation state presents strong uncertainty, multi-time and space scale coupling and dynamic evolution characteristics. The traditional single data source evaluation method (such as topology analysis based on SCADA measurement and phasor measurement based on PMU) is difficult to accurately capture hidden faults, parameter drift and cascading risks under complex working conditions due to single data dimension and poor anti-interference. Multi-source data fusion technology integrates the complementary features of heterogeneous data sources (such as wide-area measurement data, device sensor signals, meteorological environment information, and historical operation and maintenance records) and combines evidence theory, Bayesian network, deep learning and other algorithms to break through the perception limitations of single data source, realize cross-domain association modeling and dynamic evaluation of power grid state.

[0003] In the prior art, the correlation between multi-source data and the correlation of data over time are not considered, resulting in poor accuracy and adaptability of power grid state evaluation, which cannot guarantee the rationality of power grid state monitoring, and makes the power grid task efficiency poor.

[0004] Therefore, how to improve the accuracy and adaptability of power grid state evaluation is a technical problem to be solved at present. SUMMARY

[0005] The present application relates to the technical field of data analysis, and particularly relates to a power grid state evaluation method and system applying multi-source data fusion.

[0006] Collecting multi-source information of the power grid, establishing a power grid multi-source information time axis according to the time stamp of the multi-source information, determining a time period to be evaluated, and marking the axis section to be evaluated on the power grid multi-source information time axis;

[0007] Determining the dynamic characteristics of the power grid in the axis section to be evaluated, analyzing the correlation degree of the multi-source information in the axis section to be evaluated on the power grid multi-source information time axis, and expanding the axis section to be evaluated in combination with the dynamic characteristics of the power grid and the correlation degree;

[0008] Defining the power grid state advantage and disadvantage levels according to the power grid multi-source information time axis, and extracting multi-source information features based on the expanded axis section to be evaluated;

[0009] According to the grid state advantage and disadvantage grade, an identification framework of the evidence theory is constructed, and the basic probability of the evidence theory is constructed, multi-source information features are input into the evidence theory, and the grid state advantage and disadvantage grade in the to-be-evaluated time period is identified.

[0010] In some embodiments of the present application, the grid dynamic characteristics in the to-be-evaluated shaft section are determined, including,

[0011] The grid dynamic characteristics include two dynamic modes of transient mode and steady mode.

[0012] The time domain features and the frequency domain features in the to-be-evaluated shaft section are extracted, the dynamic mode to which the time domain features and the frequency domain features belong is judged according to the ranges where the time domain features and the frequency domain features are located, and the dynamic value of the dynamic mode is determined in combination with the time domain features and the frequency domain features.

[0013] In some embodiments of the present application, the correlation degree of the multi-source information in the to-be-evaluated shaft section on the grid multi-source information time axis is analyzed, including,

[0014] The multi-source information is classified by data categories, the autocorrelation function between the to-be-evaluated shaft section and the same data of the shaft section at the previous time is calculated on the basis of the grid multi-source information time axis, and the autocorrelation curve graph changing with time is drawn, the median value and the maximum value are determined on the autocorrelation curve graph, the strong correlation interval on the autocorrelation curve graph is determined, and the autocorrelation time node is determined according to the strong correlation interval.

[0015] The autocorrelation time nodes of each category of data are integrated to obtain an autocorrelation time node set, and a first forward time point is determined according to the autocorrelation time node set.

[0016] The dependent variable elements are determined on the to-be-evaluated shaft section, the dependent variable elements are data or events, the corresponding independent variables of each dependent variable element are determined, the cross-correlation function between the independent variables and the dependent variable elements is calculated based on the shaft section range between the first forward time point and the to-be-evaluated shaft section, and the cross-correlation curve graph changing with time is drawn.

[0017] The mode, the original average value, the standard deviation and the maximum value are counted on the cross-correlation curve graph, the average value of the mode and the maximum value is calculated, the multiple is determined based on the average value, the threshold value is determined in combination with the original average value, the standard deviation and the multiple.

[0018] The strong correlation area on the cross-correlation curve graph is divided by the threshold value, and the area of the strong correlation area is calculated to determine the correlation degree corresponding to each dependent variable element.

[0019] In some embodiments of the present application, the to-be-evaluated shaft section is expanded in combination with the grid dynamic characteristics and the correlation degree, including,

[0020] According to the dynamic mode type, the dynamic value and the correlation degree, a second forward time point is determined, and the shaft section to be evaluated is expanded according to the second forward time point.

[0021] In some embodiments of the present application, the power grid state advantage and disadvantage grade is defined according to the power grid multi-source information time axis, including,

[0022] The multi-source information includes power generation side data, power transmission side data and power distribution side data, and the power generation side data, the power transmission side data and the power distribution side data are extracted on the power grid multi-source information time axis.

[0023] The balance index and the power grid feature are extracted according to the power generation side data, the power transmission side data and the power distribution side data.

[0024] The power grid feature is input into a preset power grid unbalanced state model, and the initial state degree of the power grid is output, and the power grid state advantage and disadvantage grade is defined according to the initial state degree of the power grid and the balance index.

[0025] The initial state degree of the power grid describes the state condition in the unbalanced direction of the power grid, and the balance index describes the state condition in the balanced direction of the power grid.

[0026] In some embodiments of the present application, the recognition framework of the evidence theory is constructed according to the power grid state advantage and disadvantage grade, including,

[0027] All values of the power grid state advantage and disadvantage grade are determined, and the proposition set of the evidence theory is allocated one by one according to the value range of the power grid state advantage and disadvantage grade, and the proposition set includes all power grid state advantage and disadvantage grades.

[0028] In some embodiments of the present application, the basic probability of the evidence theory is constructed, including,

[0029] The multi-source features are normalized, the multi-source features are divided into multiple evidences, the membership degree function of each evidence and the proposition is defined, and the original BPA is calculated.

[0030] The conflict evidence is determined by analyzing multiple evidences, the KL divergence of the conflict evidence is calculated, the discount coefficient is determined according to the KL divergence, the original BPA is adjusted according to the discount coefficient, and thus the BPA of each evidence is determined.

[0031] In some embodiments of the present application, the multi-source information features are input into the evidence theory, including,

[0032] The real-time noise of each evidence is counted, the cross degree and mutual support degree between multiple evidences are calculated, the fusion weight of each evidence in the evidence fusion process is dynamically determined based on the real-time noise, the cross degree and the mutual support degree, the fusion between multiple evidences is realized, and the proposition result is output.

[0033] Correspondingly, a power grid state evaluation system applying multi-source data fusion comprises,

[0034] A first module is configured to collect multi-source information of the power grid, establish a power grid multi-source information timeline according to time stamps of the multi-source information, determine a time period to be evaluated, and mark an axis section to be evaluated on the power grid multi-source information timeline;

[0035] A second module is configured to determine power grid dynamic characteristics under the axis section to be evaluated, analyze a correlation degree of the multi-source information on the axis section to be evaluated on the power grid multi-source information timeline, and expand the axis section to be evaluated in combination with the power grid dynamic characteristics and the correlation degree;

[0036] A third module is configured to define power grid state advantage-disadvantage levels according to the power grid multi-source information timeline, and extract multi-source information features on the basis of the expanded axis section to be evaluated.

[0037] A fourth module is configured to construct a recognition framework of evidence theory according to the power grid state advantage-disadvantage levels, construct basic probabilities of the evidence theory, input the multi-source information features into the evidence theory, and recognize the power grid state advantage-disadvantage levels under the time period to be evaluated.

[0038] Compared with the prior art, the present application has the following beneficial effects:

[0039] 1. The axis section to be evaluated is expanded in combination with the power grid dynamic characteristics and the correlation degree. The power grid state under the axis section to be evaluated is often related to data at previous times. A reasonable forward time node is determined in combination with the power grid dynamic characteristics and the correlation degree, data of the axis section to be evaluated is expanded, the comprehensiveness and rationality of the data are ensured, and a reliable foundation is provided for subsequent multi-source information feature extraction. The power grid state advantage-disadvantage levels are defined. The balance direction and the unbalance direction of the power grid are considered to comprehensively define the power grid state advantage-disadvantage levels, the feasibility and pertinence of the evidence theory implementation are ensured, and the actual operation state of the power grid is adapted.

[0040] 2. The basic probabilities of the evidence theory are constructed, the multi-source information features are input into the evidence theory, uncertain information is fused through a Dempster combination rule, and the misjudgment problem of a traditional threshold method for a boundary state is solved. The evidence theory can process conflicts between multi-source information. Through a discount coefficient and a dynamic evidence weight, a single data source is avoided from dominating the evaluation result, the decision robustness under complex working conditions is improved, a quantifiable state confidence reference is provided for dispatchers, the accuracy and adaptability of the power grid state evaluation are improved, the rationality of the power grid state monitoring is ensured, and the stability and safety of the power grid operation state are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 A flowchart of a power grid state evaluation method applying multi-source data fusion is provided.

[0042] Figure 2 A structural schematic diagram of a power grid state evaluation system using multi-source data fusion is provided for the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0044] Reference Figure 1 A power grid state evaluation method using multi-source data fusion includes the following steps:

[0045] Step S101, collect multi-source information of the power grid, establish a power grid multi-source information time axis according to the time stamps of the multi-source information, determine a time period to be evaluated, and mark the axis section to be evaluated on the power grid multi-source information time axis.

[0046] In this embodiment, the multi-source information includes three types of data of the power generation side data, the power transmission side data and the power distribution side data of the power grid, for example, collecting SCADA measurement (voltage, current), PMU phasor data (frequency, phase angle), weather data (temperature, humidity), equipment sensing signals (transformer oil temperature, circuit breaker state) and the like of the power grid. The multi-source data is sorted according to the time stamp, the power grid multi-source information time axis is constructed, the time period to be evaluated (such as 10:00-10:10) is determined according to the dispatching demand, and the time axis is marked as the axis section to be evaluated.

[0047] It should be noted that the prior art needs to analyze the power grid state at 10:00-10:10, and may only analyze the data of the time period to be evaluated (10:00-10:10), but the power grid data has causality and timeliness, i.e. the power grid state at 10:00-10:10 may be related to or affected by the power grid situation before 10:00, which needs to be considered to ensure the reliability of the power grid state evaluation. The axis section to be evaluated is expanded to solve these problems.

[0048] Step S102, determine the power grid dynamic characteristics under the axis section to be evaluated, analyze the correlation degree of the multi-source information on the axis section to be evaluated on the power grid multi-source information time axis, and expand the axis section to be evaluated in combination with the power grid dynamic characteristics and the correlation degree.

[0049] In this embodiment, the power grid dynamic characteristics and the correlation degree of the multi-source information (the correlation of the data before the time and the data of the axis section to be evaluated) will affect the expansion requirement, and the determination of the expansion time point is considered.

[0050] In some embodiments of the present application, the power grid dynamic characteristics under the axis section to be evaluated include,

[0051] The power grid dynamic characteristics include transient mode and steady mode;

[0052] The time domain features and the frequency domain features of the to-be-evaluated shaft section are extracted, the dynamic mode to which the time domain features and the frequency domain features belong is judged according to the ranges in which the time domain features and the frequency domain features are located, and the dynamic value of the dynamic mode is determined in combination with the time domain features and the frequency domain features.

[0053] In this embodiment, the time domain features (mean value, variance) and the frequency domain features (FFT harmonic components) of the data in the to-be-evaluated shaft section are extracted, the transient mode is a non-stationary stage in which the power grid transits from an initial state to a new balance when the power grid is subjected to external disturbances (such as short circuit, new energy fluctuation) or internal parameter mutation, and the steady mode is a state in which the power grid reaches a new balance after normal operation or disturbance, the electrical quantity fluctuates within a permitted range, and the system structure is stable. The dynamic mode to which the time domain features and the frequency domain features belong is judged according to the ranges in which the time domain features and the frequency domain features are located, for example, if the time domain variance is greater than a threshold value and the frequency domain high-order harmonic proportion is greater than 10%, it is determined that the transient mode (such as short circuit fault) is reached. If the time domain variance is less than the threshold value and the frequency domain fundamental wave proportion is greater than 90%, it is determined that the steady mode (such as normal operation) is reached. Transient mode: dynamic value = time domain variance x frequency domain high-order harmonic proportion. Steady mode: dynamic value = 1 / (time domain variance + frequency domain fundamental wave proportion). The dynamic value describes the quantitative degree of the power grid dynamic.

[0054] In some embodiments of the present application, the correlation degree of the multi-source information on the to-be-evaluated shaft section on the time axis of the power grid multi-source information is analyzed, including,

[0055] The multi-source information is classified according to data categories, the autocorrelation function between the same data of the to-be-evaluated shaft section and the shaft section at a previous time on the basis of the time axis of the power grid multi-source information is calculated, and the autocorrelation curve graph changing with time is drawn. The median value and the maximum value are determined on the autocorrelation curve graph, so as to determine the strong correlation interval on the autocorrelation curve graph, and the autocorrelation time node is determined according to the strong correlation interval.

[0056] The autocorrelation time nodes of each category of data are integrated to obtain a set of autocorrelation time nodes, and a first forward time point is determined according to the set of autocorrelation time nodes.

[0057] The dependent variable elements are determined on the to-be-evaluated shaft section, the dependent variable elements are data or events, and the corresponding independent variables of each dependent variable element are determined. The cross-correlation function between the independent variables and the dependent variable elements is calculated based on the shaft section range between the first forward time point and the to-be-evaluated shaft section, and the cross-correlation curve graph changing with time is drawn.

[0058] The mode, the original average value, the standard deviation and the maximum value are counted on the cross-correlation graph, the average value of the mode and the maximum value is calculated, the multiple is determined based on the average value, and the threshold value is determined in combination with the original average value, the standard deviation and the multiple;

[0059] The strong correlation region on the cross-correlation graph is divided by the threshold value, the area of the strong correlation region is calculated to determine the correlation degree corresponding to each dependent variable element.

[0060] In this embodiment, the autocorrelation function (ACF) of the to-be-evaluated shaft section and the same type of data in the previous period (for example, ten minutes) is calculated, and the autocorrelation curve is drawn. The part between the median value and the maximum value is taken as the strong correlation interval, the strong correlation interval (such as ACF>0.7) is determined, and the autocorrelation time node set (such as 10:00, 10:15) is extracted. A first forward time point is determined according to the autocorrelation time node set, the autocorrelation time node closest to the front in the set is taken as the first forward time point, which ensures the maximum coverage degree, avoids the lack or loss of relevant data, and filters out the strong correlation interval, which can effectively reduce the data processing amount and ensure the analysis effect.

[0061] In this embodiment, after analyzing the autocorrelation of the same type of data at different times, the cross-correlation of the to-be-evaluated shaft section and the previous time data (the first forward time point) needs to be considered. The relationship between the independent variable and the dependent variable can be determined according to the power grid model theory or historical data. The dependent variable is the target data / event directly observed or calculated in the to-be-evaluated shaft section, and is the core object of analysis. For example, the voltage sag amplitude (dependent variable: voltage amplitude), wind farm power fluctuation rate (dependent variable: output power), fault trigger signal (dependent variable: protection device action event). The independent variable is a potential driving factor that affects the change of the dependent variable, which may come from historical data or external input. For example, new energy output (independent variable: photovoltaic / wind power), load switching (independent variable: user-side power consumption), environmental disturbance (independent variable: wind speed / light intensity), etc. The cross-correlation function (CCF) is calculated, the average value of the mode and the maximum value on the graph is calculated, different multiples correspond to different average values, the threshold value is determined in combination with the original average value, the standard deviation and the multiple, and the threshold value = average value ± 3 times standard deviation. Each dependent variable element corresponds to an association relationship, i.e. the correlation degree, which is calculated by trapezoidal numerical integration, parameterized curve fitting and other methods to map the correlation degree.

[0062] In some embodiments of the present application, the to-be-evaluated shaft section is expanded in combination with the dynamic characteristics of the power grid and the correlation degree, including,

[0063] The correlation degrees corresponding to all dependent variable elements are integrated, the second forward time point is determined according to the dynamic mode type, the dynamic value and the correlation degree, and the to-be-evaluated shaft section is expanded according to the second forward time point.

[0064] In this embodiment, the second forward time point is determined according to the dynamic mode type, the dynamic value and the correlation degree, an initial time point is determined according to the correlation degree, a correction factor is determined according to the dynamic value, the mapping relationship of different dynamic mode types is different, and the initial time point is corrected through the correction factor (correction factor * initial time point) to obtain the second forward time point, so as to expand the to-be-evaluated shaft section.

[0065] In step S103, the grid state advantage and disadvantage grade is defined according to the grid multi-source information time axis, and the multi-source information features are extracted on the basis of the expanded to-be-evaluated shaft section.

[0066] In this embodiment, the grid state advantage and disadvantage is comprehensively evaluated from two directions, one direction is the balance condition in the grid, and the other direction is the state under non-balance (the condition or scenario that the balance is good but the state is not good, for example, the latent period of hidden fault, the strong uncertainty impact of new energy, the critical state of device overload, etc.). The latent period of hidden fault is the internal insulation aging of the device (such as partial discharge of the transformer), poor contact of the line, but no tripping is caused, but the power / voltage / frequency deviation meets the standard. The strong uncertainty impact of new energy is the sudden drop of wind power / photovoltaic output (such as a 30% drop within 10 seconds), but the real-time balance index is not exceeded through the AGC rapid adjustment to maintain power balance.

[0067] In some embodiments of the present application, the grid state advantage and disadvantage grade is defined according to the grid multi-source information time axis, including,

[0068] The multi-source information includes the power generation side data, the power transmission side data and the power distribution side data of the grid, and the power generation side data, the power transmission side data and the power distribution side data of the grid are extracted on the grid multi-source information time axis;

[0069] The balance index and the grid feature are extracted according to the power generation side data, the power transmission side data and the power distribution side data of the grid;

[0070] The grid feature is input into a preset grid non-balance state model, and an initial state degree of the grid is output, and the grid state advantage and disadvantage grade is defined according to the initial state degree of the grid and the balance index;

[0071] The initial state degree of the grid describes the state condition under the grid non-balance direction, and the balance index describes the state condition under the grid balance direction.

[0072] In this embodiment, the grid non-balance state model can be a dynamic system model, a machine learning model, etc., and the balance index includes a voltage balance index, a power balance index and a frequency balance index, and specifically,

[0073] The voltage balance index

[0074] Voltage magnitude deviation: Range of deviation of three-phase voltage magnitude from rated value.

[0075] Voltage unbalance: Ratio of negative or zero sequence voltage component to positive sequence voltage component.

[0076] Voltage fluctuation rate: Maximum rate of change of voltage magnitude per unit time.

[0077] Power balance index

[0078] Active / reactive power balance rate: Deviation rate of actual output power from planned power.

[0079] Power factor: Ratio of active power to apparent power.

[0080] Flow distribution rationality: Comparison of line load rate and thermal limit.

[0081] Frequency balance index

[0082] Frequency deviation: Difference between actual frequency and rated frequency (50Hz or 60Hz).

[0083] Frequency fluctuation rate: Maximum rate of change of frequency per unit time.

[0084] Grid characteristics

[0085] Topology structure characteristics

[0086] Mesh structure: Radial, ring or mesh structure.

[0087] Node type: Generation node (PV / PQ node), load node, balance node.

[0088] Line parameter: Resistance, reactance, admittance.

[0089] Operating state characteristics

[0090] Load rate: Ratio of line or transformer load to rated capacity.

[0091] Flow direction: Direction and size of power flow.

[0092] Switching state: On / off state of circuit breaker, disconnector.

[0093] Equipment state characteristics

[0094] Equipment health: Evaluation of equipment degradation based on online monitoring data.

[0095] Failure probability: Equipment failure prediction based on historical data.

[0096] Maintenance record: Time and content of the last maintenance of the equipment.

[0097] Dynamic response characteristics

[0098] Frequency response capability: active regulation capability when system frequency changes.

[0099] Voltage response capability: reactive regulation capability when system voltage changes.

[0100] Transient stability: dynamic recovery capability after fault.

[0101] In this embodiment, the formula for defining the grid state good-bad level according to the initial state degree and the balance index of the power grid is as follows:

[0102]

[0103] wherein, is the grid state good-bad level, is the initial state degree of the power grid, is the number of balance indexes, is the combined weight of the th balance index, is the size of the th balance index, is the maximum value in is the first constant, is the second constant, represents the correction of the maximum balance to the balance average value, the value range is between 0.879-1.211, the first constant is to balance the size of the correction function, and the second constant is to balance the size of the grid state good-bad level, and [] is the rounding symbol.

[0104] In step S104, the recognition framework of the evidence theory is constructed according to the grid state good-bad level, the basic probability of the evidence theory is constructed, the multi-source information characteristics are input into the evidence theory, and the grid state good-bad level in the time period to be evaluated is recognized.

[0105] In this embodiment, the DS (Dempster-Shafer) evidence theory is an effective tool that can handle uncertainty and incomplete information, and the basic probability assignment (BPA) is the core of the DS evidence theory, which represents the trust degree of each hypothesis (or proposition). When constructing the BPA, the grid characteristics, the balance index, and the expert knowledge need to be combined. The process is 1 to construct the recognition framework-2 to construct the basic probability assignment (BPA)-3 to synthesize the evidence-4 to recognize the state.

[0106] In some embodiments of the present application, the recognition framework of the evidence theory is constructed according to the grid state good-bad level, which includes,

[0107] ​​All values of the power grid state advantage grade are determined, and the proposition set of the evidence theory is allocated one by one according to the value range of the power grid state advantage grade. The proposition set includes all power grid state advantage grades.

[0108] In this embodiment, the power grid state advantage grade is a proposition, and the proposition is the output result of the evidence theory.

[0109] In some embodiments of the present application, the basic probability of the evidence theory is not constructed, including,

[0110] The multi-source features are normalized, the multi-source features are divided into multiple evidences, the membership function of each evidence and the proposition is defined, and the original BPA is calculated.

[0111] The multiple evidences are analyzed to determine the conflict evidence, the KL divergence of the conflict evidence is calculated, the discount coefficient is determined according to the KL divergence, the original BPA is adjusted according to the discount coefficient, and thus the BPA of each evidence is determined.

[0112] In this embodiment, the normalized features are divided into multiple evidences, and the membership function of each evidence and the proposition is defined. For example, the evidences are divided according to the feature types, such as the load rate, the voltage unbalance degree, and the power factor. The membership function is defined using fuzzy logic or expert knowledge. The conflict evidence refers to the evidence with a large difference in support for the same proposition. The KL divergence is used to measure the difference between two probability distributions. The KL divergence between all pairs of evidences is calculated, the conflict evidence is identified, the discount coefficient is inversely proportional to the KL divergence, and the discount coefficient is specifically

[0113]

[0114] wherein, is the discount coefficient, is the adjustment coefficient, is the KL divergence, is the adjusted BPA, is the original BPA, is the total set BPA of the identification framework (i.e., “uncertain” or “unknown” distribution), reflecting the overall uncertainty of the evidence.

[0115] In some embodiments of the present application, the multi-source information features are input into the evidence theory, including,

[0116] The real-time noise of each evidence is counted, the cross degree and mutual support degree between the multiple evidences are calculated, the fusion weight of each evidence in the evidence fusion process is dynamically determined based on the real-time noise, the cross degree and the mutual support degree, the fusion between the multiple evidences is realized, and the proposition result is output.

[0117] ​In this embodiment, the noise of the evidence (i.e., data), the crossover degree between evidence, and the mutual support degree affect the weights in the evidence fusion process, and therefore are dynamically set. Evidence is divided according to the fusion method (different evidence combinations). For each evidence combination, the Euclidean distance (similarity case) between one piece of evidence and other evidence is calculated to determine the crossover degree. Mutual support degree is the degree to which evidence source 1 supports other evidence source 2, specifically the degree of overlap of its BPA (Balance of Propositions) on the principal propositions of other evidence source 2. The crossover degree and mutual support are weighted and summed. Different real-time noise levels correspond to different adjustment factors. The weighted summation result is adjusted (producted) by these adjustment factors to adjust the weights.

[0118] Understandably, in Dempster-Shafer Theory, the discounting coefficient and the combination weight are two different mechanisms, operating in the evidence preprocessing and fusion stages respectively. While their objectives are related, they are not inherently contradictory. The discounting coefficient, used in the evidence preprocessing stage, adjusts the basis of impact (BPA) of individual evidence sources before fusion, reducing the weight of highly conflicting or unreliable evidence. The combination weight, used in the evidence fusion stage, allocates the relative importance of each piece of evidence when combining multiple sources, adjusting the contribution of each source to the fusion formula and directly reflecting its influence on decision-making.

[0119] Correspondingly, a power grid condition assessment system that applies multi-source data fusion, such as... Figure 2 As shown, including,

[0120] The first module is used to collect multi-source information of the power grid, establish a time axis of multi-source information of the power grid according to the timestamps of the multi-source information, determine the time period to be evaluated, and mark the axis segment to be evaluated on the time axis of multi-source information of the power grid.

[0121] The second module is used to determine the dynamic characteristics of the power grid under the axis segment to be evaluated, analyze the correlation degree of multi-source information on the time axis of the power grid multi-source information under the axis segment to be evaluated, and expand the axis segment to be evaluated by combining the dynamic characteristics of the power grid and the correlation degree.

[0122] The third module is used to define the quality level of the power grid state based on the time axis of the multi-source information of the power grid, and to extract the features of the multi-source information based on the expanded axis segment to be evaluated.

[0123] The fourth module is used to construct an identification framework for the evidence theory based on the quality level of the power grid, and to construct the basic probability of the evidence theory. It inputs multi-source information features into the evidence theory to identify the quality level of the power grid under the time period to be evaluated.

[0124] Compared with the prior art, the beneficial effects of this invention are as follows:

[0125] 1. The power grid's dynamic characteristics and correlation levels are considered to expand the data for the axis segment under evaluation. The power grid state under the evaluated axis segment is often related to data from previous time periods. By combining the power grid's dynamic characteristics and correlation levels, a reasonable forward time node is determined to expand the data for the axis segment under evaluation, ensuring the comprehensiveness and rationality of the data and providing a reliable foundation for subsequent multi-source information feature extraction. A power grid state quality level is defined, considering both the balance and imbalance directions of the power grid to comprehensively define the quality level, ensuring the feasibility and relevance of the evidence theory and adapting it to the actual operating conditions of the power grid.

[0126] 2. A basic probability model for evidence theory is constructed. Multi-source information features are input into the evidence theory, and uncertain information is fused using Dempster's combination rule to address the problem of misjudgment of boundary states in traditional threshold methods. Evidence theory can handle conflicts between multi-source information. Through discount coefficients and dynamic evidence weights, it avoids a single data source dominating the evaluation results, improves the robustness of decision-making under complex operating conditions, and provides dispatchers with quantifiable state confidence references. This improves the accuracy and adaptability of power grid state assessment, ensuring the rationality of power grid state monitoring and the stability and safety of power grid operation.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0128] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing the present invention.

[0129] Those skilled in the art will understand that the modules in the system of the implementation scenario can be distributed throughout the system of the implementation scenario as described, or they can be modified to reside in one or more systems different from this implementation scenario. The modules of the above-mentioned implementation scenario can be merged into one module, or they can be further divided into multiple sub-modules.

[0130] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A power grid state evaluation method using multi-source data fusion, characterized in that, Comprising, collecting multi-source information of the power grid, establishing a power grid multi-source information timeline according to time stamps of the multi-source information, determining a time period to be evaluated, and marking an axis section to be evaluated on the power grid multi-source information timeline; determining power grid dynamic characteristics under the axis section to be evaluated, analyzing the correlation degree of the multi-source information on the axis section to be evaluated on the power grid multi-source information timeline, and expanding the axis section to be evaluated in combination with the power grid dynamic characteristics and the correlation degree; defining power grid state advantage and disadvantage levels according to the power grid multi-source information timeline, extracting multi-source information features on the basis of the expanded axis section to be evaluated; constructing a recognition framework of evidence theory according to the power grid state advantage and disadvantage levels, constructing basic probabilities of the evidence theory, inputting the multi-source information features into the evidence theory, and recognizing the power grid state advantage and disadvantage levels under the time period to be evaluated; wherein, determining the power grid dynamic characteristics under the axis section to be evaluated comprises, the power grid dynamic characteristics include two dynamic modes of transient mode and steady state mode; extracting time domain features and frequency domain features under the axis section to be evaluated, judging the dynamic mode according to the range where the time domain features and the frequency domain features are located, and determining the dynamic value of the dynamic mode in combination with the time domain features and the frequency domain features; analyzing the correlation degree of the multi-source information on the axis section to be evaluated on the power grid multi-source information timeline comprises, performing data category division on the multi-source information, calculating the autocorrelation function between the same data of the axis section to be evaluated and the axis section before the time on the basis of the power grid multi-source information timeline, and drawing an autocorrelation curve graph changing with time, determining the median value and the maximum value on the autocorrelation curve graph, thereby determining the strong correlation interval on the autocorrelation curve graph, and determining the autocorrelation time node according to the strong correlation interval; integrating the autocorrelation time nodes of each type of data to obtain a set of autocorrelation time nodes, and determining a first forward time point according to the set of autocorrelation time nodes; determining dependent variable elements on the axis section to be evaluated, the dependent variable elements being data or events, and determining the corresponding independent variables of each dependent variable element, calculating the cross-correlation function between the independent variables and the dependent variable elements based on the axis section range between the first forward time point and the axis section to be evaluated, and drawing a cross-correlation curve graph changing with time; statistically obtaining the mode, the original average value, the standard deviation and the maximum value on the cross-correlation curve graph, calculating the average value of the mode and the maximum value, determining a multiple based on the average value, and determining a threshold value in combination with the original average value, the standard deviation and the multiple; dividing the strong correlation region on the cross-correlation curve graph through the threshold value, calculating the area of the strong correlation region to determine the correlation degree corresponding to each dependent variable element; expanding the axis section to be evaluated in combination with the power grid dynamic characteristics and the correlation degree comprises, comprehensively integrating the correlation degrees corresponding to all dependent variable elements, determining a second forward time point according to the dynamic mode type, the dynamic value and the correlation degree, and expanding the axis section to be evaluated according to the second forward time point.

2. The power grid state estimation method using multi-source data fusion according to claim 1, characterized in that, defining the power grid state advantage and disadvantage levels according to the power grid multi-source information timeline comprises, the multi-source information includes power grid generation side data, power transmission side data and power distribution side data, and the power grid generation side data, the power transmission side data and the power distribution side data are extracted on the power grid multi-source information timeline; According to the power generation side data, power transmission side data and power distribution side data of the power grid, balance indicators and power grid characteristics are extracted; The power grid characteristics are input into a preset power grid unbalanced state model, and the initial state degree of the power grid is output, and the power grid state advantage grade is defined according to the initial state degree of the power grid and the balance indicators. The initial state degree of the power grid describes the state of the power grid in the unbalanced direction, and the balance indicators describe the state of the power grid in the balanced direction.

3. The method of claim 1, wherein, According to the power grid state advantage grade, an identification framework of evidence theory is constructed, including, All values of the power grid state advantage grade are determined, and the proposition set of evidence theory is allocated one by one according to the value range of the power grid state advantage grade. The proposition set includes all power grid state advantage grades.

4. The method of claim 3, wherein, And the basic probability of evidence theory is constructed, including, The multi-source characteristics are normalized, the multi-source characteristics are divided into multiple evidences, the membership function of each evidence and proposition is defined, and the original BPA is calculated. The conflict evidence is determined by analyzing multiple evidences, the KL divergence of the conflict evidence is calculated, the discount coefficient is determined according to the KL divergence, and the original BPA is adjusted according to the discount coefficient, so as to determine the BPA of each evidence.

5. The method for power grid state estimation using multi-source data fusion according to claim 4, characterized in that, Input multi-source information characteristics into evidence theory, Including, The real-time noise of each evidence is counted, the cross degree and mutual support degree between multiple evidences are calculated, the fusion weight of each evidence in the evidence fusion process is dynamically determined based on the real-time noise, cross degree and mutual support degree, the fusion between multiple evidences is realized, and the proposition result is output.

6. A power grid state assessment system using multi-source data fusion, characterized in that, The system for realizing the power grid state evaluation method using multi-source data fusion according to any one of claims 1-5, the system comprises, The first module is used for collecting multi-source information of the power grid, establishing a power grid multi-source information time axis according to the time stamp of the multi-source information, determining a time period to be evaluated, and marking the axis section to be evaluated on the power grid multi-source information time axis; The second module is used for determining the dynamic characteristics of the power grid in the axis section to be evaluated, analyzing the correlation degree of the multi-source information in the axis section to be evaluated on the power grid multi-source information time axis, and expanding the axis section to be evaluated in combination with the dynamic characteristics of the power grid and the correlation degree; The third module is used for defining the power grid state advantage grade according to the power grid multi-source information time axis, and extracting multi-source information characteristics based on the expanded axis section to be evaluated; The fourth module is used for constructing an identification framework of evidence theory according to the power grid state advantage grade, constructing the basic probability of evidence theory, inputting the multi-source information characteristics into the evidence theory, and identifying the power grid state advantage grade in the time period to be evaluated.

Images