A multi-source data fusion distributed energy collaborative optimization scheduling method

Abstract

The application belongs to the technical field of energy scheduling, and particularly relates to a multi-source data fusion distributed energy collaborative optimization scheduling method, which is based on historical scheduling performance analysis, constructs an energy uncertainty graph taking nodes to represent various energy subsystems and edge weights to represent the uncertainty transmission relationship between different energy subsystems, realizes data-driven description of dynamic risk interaction among multiple energies, performs path analysis on the energy uncertainty graph, identifies risk transmission paths from each energy subsystem node to the overall output node, calculates the cumulative uncertainty index of each risk transmission path, combines the cumulative uncertainty index statistical distribution characteristics, screens key uncertainty sources to modify the initial scheduling instruction constructed according to the net surplus power of various energies, and finally issues the scheme for execution after ensuring the reliability of the scheme through the feasibility check and iteration mechanism, thereby effectively improving the reliability of the distributed energy collaborative optimization scheduling.

Description

Technical Field

[0001] This invention belongs to the field of energy dispatching technology, specifically a distributed energy collaborative optimization dispatching method that integrates multi-source data. Background Technology

[0002] With the increasing penetration of renewable energy and the widespread integration of distributed energy sources, modern power systems are exhibiting typical characteristics of multi-energy complementarity and coordinated operation. Coordinated and optimized dispatching of heterogeneous energy sources such as wind power, photovoltaics, energy storage, and gas turbines, leveraging their complementary characteristics to mitigate power output fluctuations, has become a key means to improve the reliability and economy of power grid operation.

[0003] Currently, existing technical solutions in this field mainly follow the following technical paths: First, they focus on improving the accuracy of short-term power prediction for a single energy source, aiming to reduce prediction errors at the source. Second, they construct optimization models based on the prediction results, targeting operating costs or efficiency, and incorporate the physical and operational constraints of various devices into the models to obtain scheduling plans. Furthermore, some studies introduce probability distributions or fuzzy intervals to quantitatively describe uncertainties such as prediction errors, and employ robust optimization or stochastic programming methods to generate scheduling strategies with a certain degree of anti-interference capability.

[0004] However, in-depth analysis reveals that existing technological solutions still have the following limitations when dealing with the complex uncertainties and interactions of multi-energy systems: First, at the uncertainty modeling level, there is a lack of characterization of the dynamic interactions and risk transmission mechanisms among multiple energy sources. Existing methods typically model the uncertainties of various energy sources independently or simply superimpose them, failing to analyze and quantify from historical data how the output deviation of one energy source triggers and transmits to other energy sources in a dynamic process. This energy correlation based on static assumptions or simple correlations cannot accurately reflect the true coupled risk network within the energy subsystem, leading to biases in the assessment of overall uncertainty.

[0005] Second, data-driven precision in identifying key risk sources has been lacking. Existing methods mostly rely on expert experience to set importance or employ computationally complex global sensitivity analysis. They lack a mechanism to automatically learn from historical deviation events and identify the key sources and transmission paths that have the most significant impact on the overall system risk. This makes it difficult for scheduling decisions to focus on truly high-risk links, resulting in low efficiency in risk prevention and control.

[0006] Third, at the level of scheduling strategy generation, the correction methods are crude and it is difficult to balance security and economy. Faced with uncertainty, existing strategies often adopt global re-optimization or conservative adjustments based on fixed rules. They cannot make preventive and differentiated precise power corrections for identified key risk transmission links. This easily leads to scheduling schemes that are either too conservative and sacrifice economy, or cannot effectively suppress potential cascading risks due to insufficient targeting. It is difficult to balance the robustness and optimization of scheduling strategies. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, this invention provides a distributed energy collaborative optimization scheduling method based on multi-source data fusion, which can effectively solve the problems mentioned in the prior art.

[0008] The objective of this invention can be achieved through the following technical solution: a distributed energy collaborative optimization scheduling method with multi-source data fusion, comprising: obtaining the real-time available power and local predicted load of various energy sources in the current scheduling cycle, and constructing an initial scheduling instruction according to the net surplus power.

[0009] Based on historical scheduling performance analysis, an energy uncertainty graph containing various energy sources is constructed. The energy uncertainty graph uses nodes to represent various energy subsystems and edge weights to characterize the uncertainty transmission relationship between different energy subsystems.

[0010] Path analysis is performed on the energy uncertainty map to identify risk transmission paths from each energy subsystem node to the overall output node, and the cumulative uncertainty index of each risk transmission path is calculated.

[0011] Based on the cumulative uncertainty indicators and their statistical distribution characteristics of each risk transmission path, key uncertainty sources are screened and a list of key uncertainty sources is generated.

[0012] Based on the list of key uncertainty sources, the scheduling parameters of the corresponding energy subsystems in the initial scheduling instructions are corrected to generate corrected scheduling instructions, which are then sent to the execution units of each energy subsystem.

[0013] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) Based on the scheduling deviation sequence of each energy subsystem, the present invention quantifies the dynamic process of how the output deviation of a certain energy subsystem triggers and is transmitted to other sub-energy systems through time-series causal analysis, and constructs an energy uncertainty map in which nodes represent energy subsystems and edge weights characterize the uncertainty transmission relationship, effectively solving the defects of the prior art in multi-energy uncertainty modeling that lacks dynamic interaction and risk transmission mechanism characterization.

[0014] (2) This invention identifies all risk transmission paths leading to the overall output node by performing path analysis on the energy uncertainty map and calculates their cumulative uncertainty index. Subsequently, instead of subjectively setting thresholds, it screens key uncertainty sources based on the statistical distribution characteristics of the cumulative uncertainty index of all paths and comprehensively considers the cumulative frequency of nodes in high-risk paths. This avoids relying on subjective experience or complex global analysis and can objectively and efficiently locate the uncertainty sources that contribute most significantly to the overall risk of the system from historical data.

[0015] (3) This invention only makes targeted corrections to the energy subsystems of the list of key uncertainty sources, determines the power correction direction based on the main response direction in the historical transmission, and generates differentiated power correction coefficients based on the comprehensive risk contribution. It proposes a preventive and differentiated scheduling correction strategy for key risk transmission links, and ensures the feasibility of the scheme through a feasibility verification and iterative correction mechanism. Attached Figure Description

[0016] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the implementation steps of the method of the present invention.

[0018] Figure 2 This is a logical schematic diagram illustrating the construction of an energy uncertainty map encompassing various energy sources for this invention.

[0019] Figure 3 This is a schematic diagram illustrating the logic of how the revised scheduling instructions are issued to the execution units of each energy subsystem, and how the scheduling execution effect is evaluated through closed-loop feedback and the dynamic updating of the map. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 As shown, the present invention provides a distributed energy collaborative optimization scheduling method based on multi-source data fusion, including: S1. Obtaining the real-time available power and local predicted load of various energy sources in the current scheduling cycle, and constructing an initial scheduling instruction according to the net surplus power.

[0022] The aforementioned real-time available power is defined as the maximum power that various energy subsystems can actually output within the current scheduling cycle, based on their real-time operating status and physical conditions. This value reflects the power generation or supply capacity of the equipment in the current scheduling cycle, and can be obtained by retrieving logs or real-time databases from the monitoring and data acquisition systems deployed in each energy subsystem.

[0023] Local forecast load is defined as the total predicted power load that the local scheduling area or system itself needs to undertake and satisfy within the current scheduling cycle. It can be obtained by retrieving the logs of the load forecasting system or calling its forecasting interface.

[0024] In a preferred embodiment of the present invention, the initial scheduling instruction construction process includes: subtracting the real-time available power from the local predicted load to obtain the net surplus power of each type of energy in the current scheduling cycle.

[0025] By loading the scheduling constraints and energy conversion efficiency of controllable devices in each energy subsystem, the original net surplus power is corrected.

[0026] The optimization objective is to minimize the total operating cost of the current scheduling cycle, which includes at least fuel cost, operation and maintenance cost, and start-up and shutdown cost. To meet the net scheduling load demand, the planned scheduling power of each type of energy in the current scheduling cycle is calculated, and the initial scheduling instruction is generated by encapsulating it according to a predefined instruction format.

[0027] It should be noted that the above-described solution process for the planned scheduling power belongs to a classic optimization problem in this field and can be implemented using existing mature technologies. Specifically, based on the net surplus power, equipment constraints, and total operating cost, an optimization scheduling mathematical model is constructed with the objective of minimizing the total operating cost of the current scheduling cycle, such as a mixed-integer linear programming model. The mathematical expressions of the objective function and constraints of this model are well known to those skilled in the art. After obtaining this mathematical model, it can be solved directly using mature commercial or open-source optimization solvers, such as CPLEX, Gurobi, or scientific computing software built-in functions such as MATLAB's intlinprog function, to obtain the planned scheduling power for each type of energy. The improvement of this invention over the prior art lies not in the construction and solution method of the optimization model itself, but in the subsequent key parameter correction strategy based on uncertainty spectrum analysis provided for the optimization model. Therefore, the detailed solution process is not specifically given.

[0028] S2. Based on historical scheduling performance analysis, construct an energy uncertainty graph containing various types of energy. The energy uncertainty graph represents various energy subsystems through nodes and characterizes the uncertainty transmission relationship between different energy subsystems through edge weights.

[0029] Reference Figure 2As shown, in a preferred embodiment of the present invention, the construction of an energy uncertainty map including various types of energy is carried out as follows: the planned scheduling power and actual scheduling power of each energy subsystem in multiple historical scheduling cycles are obtained, the difference between the actual scheduling power and the planned scheduling power is used as the scheduling deviation value, and the scheduling deviation values ​​of the energy subsystem in multiple historical scheduling cycles are arranged in chronological order to generate a scheduling deviation sequence for each energy subsystem.

[0030] Each energy subsystem is taken as an initial deviation source in turn. For each event in the scheduling deviation sequence that exceeds the preset scheduling tolerance error, the preset scheduling tolerance error can be determined according to the percentage of the rated power of each energy subsystem. For example, 2%-5% of the rated power can be selected. Time-series causal analysis is performed to determine the influence transmission characteristics when each energy subsystem is taken as an initial deviation source.

[0031] The time-series causal analysis process includes: for any two energy subsystems A and B, calculating the cross-correlation function between the scheduling adjustment and the scheduling deviation of the other energy subsystem under multiple time lags within the same scheduling period when a scheduling deviation occurs in one of the energy subsystems. The calculation formula is as follows: .

[0032] in, The scheduling deviation sequence of energy subsystem A, This refers to the time sequence number for each historical scheduling cycle. , This refers to the sequence length, and also to the total number of historical scheduling cycles. This refers to the sequence of scheduling adjustments for energy subsystem B relative to energy subsystem A under multiple time lags within the same scheduling period. For a given time lag, the range of values ​​is generally [value range missing]. , For the maximum lag step size, These are the mean values ​​in the corresponding sequences of energy subsystems A and B, respectively.

[0033] The numerator of this calculation formula represents the decentralized dot product of the scheduling deviation sequence of energy subsystem A and the scheduling adjustment sequence of energy subsystem B relative to energy subsystem A under multiple time lags within the same scheduling period, and represents the covariance of the two sequences under a certain time lag.

[0034] The denominator represents the product of the standard deviations of the two sequences, used to normalize the correlation coefficient to the interval [-1, 1].

[0035] This calculation formula characterizes the normalized cross-correlation function, which can eliminate the influence of differences in sequence amplitude and reflect the similarity of two sequences in shape and trend. When the cross-correlation function value is 1, it means that the scheduling adjustment of energy subsystem B is completely positively correlated with the scheduling deviation of energy subsystem A. When the cross-correlation function value is -1, it means that the scheduling adjustment of energy subsystem B is completely negatively correlated with the scheduling deviation of energy subsystem A. When the cross-correlation function value is 0, it means that there is no linear correlation.

[0036] Iterate through all the cross-correlation function values ​​corresponding to time lags, and filter out the peak value with the largest absolute value and its corresponding time lag. When the peak value is greater than the preset correlation threshold, it is determined that the transmission relationship between the two energy subsystems exists; otherwise, it is determined that it does not exist.

[0037] It should be noted that the above correlation threshold is pre-set based on statistical significance. For a given total number of historical scheduling cycles, if we assume that the two sequences are independent, the cross-correlation coefficient approximately follows a mean of 0 and a standard deviation of [missing value]. Since it follows a normal distribution, for a given significance level of 95%, the correlation threshold can be set as the ratio of the significance level to the standard deviation.

[0038] If the determination is found to exist, the sign of the peak value is determined as the response direction. Specifically, if the peak value is positive, it indicates that the two energy subsystems are changing in the same direction, which is positive conduction; if it is negative, it is negative conduction.

[0039] The time lag corresponding to the peak value is taken as the response lag, which represents the time interval required for energy subsystem B to make adjustments after a deviation occurs in energy subsystem A.

[0040] Regression analysis is used to quantify the ratio of adjustment to deviation, with deviation A as the quantifier. As the independent variable, B at the response lag time Adjustment amount Using as the dependent variable, perform a linear fit, and use the slope of the obtained linear fit as the response magnitude.

[0041] Thus, the conduction parameters, including response direction, response delay, and response amplitude, are determined.

[0042] In all historical scheduling events, the number of events in which a transmission relationship exists is counted. That is, in each historical scheduling event, if cross-correlation analysis determines that a transmission relationship exists from A to B, it is counted once. The occurrence frequency is defined as the ratio of the number of events in which a transmission relationship exists to the total number of analyses of historical scheduling events, and this is used to count the occurrence frequency of transmission relationships between any two energy subsystems.

[0043] Stable transmission relationships with both occurrence frequency and average response amplitude higher than the corresponding confidence level are screened. The occurrence frequency corresponding to the confidence level can be tested using a binomial distribution test. The specific process is as follows: the null hypothesis is that the occurrence of the transmission relationship is random. The minimum frequency threshold that needs to be reached at a given significance level of 95% is calculated using a binomial distribution.

[0044] The confidence level corresponding to the average response amplitude can be determined by retrieving the response amplitude distribution of all historical conduction relationships and setting the 75th percentile as the confidence level. This percentile is only a recommended example used to characterize a higher intensity standard. Those skilled in the art can customize other percentile values ​​as thresholds according to the actual requirements of conduction intensity in the scenario, in order to balance the stringency of the screening with the size of the conduction network.

[0045] Using all energy subsystems as nodes and all stable transmission relationships guided by them as initial deviation sources as directed edges, a set of uncertainty transmission paths is constructed. The edge weight of each directed edge is obtained by setting a penalty factor based on the average response time delay, thus correcting the base weight calculated based on the average response amplitude and occurrence frequency.

[0046] As an example of edge weight calculation, the average response delay, average response amplitude, and occurrence frequency are first normalized. Then, the three normalized results are linearly weighted and fused. The linear weight assigned to the average response delay carries a negative sign to represent the penalty nature. The weights of the other three normalized results should be set based on their relative importance in representing the risk of uncertainty transmission. Specifically, the average response amplitude, which directly represents the transmission intensity, can be assigned the largest weight, the occurrence frequency, which represents the commonness of the transmission relationship, can be assigned the second largest weight, and the average response delay, which represents the transmission speed, can be assigned the smallest weight. For example, the average response amplitude weight is 0.5, the occurrence frequency weight is 0.3, and the average response delay weight is -0.2.

[0047] For each energy subsystem node that serves as an initial source of deviation, the corresponding set of uncertainty transmission paths is reconstructed based on its influence transmission characteristics. Each uncertainty transmission path records an ordered sequence of nodes that start from the initial node, pass through one or more intermediate nodes, and ultimately affect the overall output node.

[0048] By integrating all uncertainty transmission paths, an energy uncertainty map is obtained.

[0049] Based on the scheduling deviation sequence of each energy subsystem, this invention quantifies the dynamic process of how the output deviation of a certain energy subsystem triggers and is transmitted to other sub-energy systems through time-series causal analysis. It constructs an energy uncertainty graph in which nodes represent energy subsystems and edge weights characterize uncertainty transmission relationships, effectively solving the shortcomings of existing technologies in multi-energy uncertainty modeling that lack dynamic interaction and risk transmission mechanisms.

[0050] S3. Perform path analysis on the energy uncertainty map to identify the risk transmission paths from each energy subsystem node to the overall output node, and calculate the cumulative uncertainty index of each risk transmission path.

[0051] In a preferred embodiment of the present invention, the risk transmission path identification process includes: obtaining the fitting change slope and mean of the scheduling deviation sequence of each energy subsystem; screening energy subsystems with inherent uncertain properties according to predefined screening rules, marking them as first-class uncertainty sources, and including them in the set to be given priority attention.

[0052] It should be noted that the above-predefined initial screening rules are as follows: after normalizing the slope of the fitted change and the mean of the scheduling deviation, the sums are obtained to obtain the comprehensive uncertainty index of the scheduling deviation sequence of each energy subsystem. An unsupervised clustering algorithm, such as the K-means algorithm with a set number of clusters of 2, is used to cluster the comprehensive uncertainty index of all energy subsystems, and each subsystem is divided into two categories.

[0053] Calculate the center values ​​of the two clusters, identify the cluster with the higher center value as the high uncertainty group, and mark all energy subsystems within this group as the first type of uncertainty source.

[0054] For the remaining energy subsystems not marked by the initial screening rules, the current external environmental impact parameters that are strongly correlated with them are obtained in real time, based on the type of the energy subsystem.

[0055] For example, if the photovoltaic or wind power energy subsystem is weather-sensitive, it is necessary to monitor the measured values ​​of irradiance, cloud cover and ambient temperature for the photovoltaic energy subsystem, and the measured values ​​of wind speed or wind direction for the wind power energy subsystem.

[0056] The hydropower energy subsystem is hydrologically dependent, requiring monitoring of real-time inflow, reservoir water level, short-term precipitation forecast data, and downstream discharge from upstream power stations.

[0057] The energy subsystems of gas turbines or oil-fired units are fuel-dependent / market-sensitive, requiring the collection of real-time market price indices and fuel supply information for the primary fuels.

[0058] In addition, some energy subsystems may be load-sensitive, requiring the collection of real-time load mutation rates in the power supply area and voltage fluctuation rates at key network nodes.

[0059] According to the operating specifications of the energy subsystem, the operating specifications define the safe allowable range of various external environmental impact parameters that are strongly correlated with the energy subsystem during operation. Only when the current monitoring value of any strongly correlated external environmental impact parameter exceeds the safe allowable range is it determined that the current external environmental impact parameter constitutes a disturbance that causes its operating state to deviate from the expected benchmark. In this case, the energy subsystem is marked as a second type of uncertainty source and included in the set of key concerns.

[0060] Using the overall output node as the sink node, starting from all energy subsystem nodes within the set to be focused on, search for all directed paths that can reach the sink node, which serve as risk transmission paths. The cumulative uncertainty index of each risk transmission path is the cumulative calculation result of the edge weights of all edges on the path.

[0061] S4. Based on the cumulative uncertainty indicators and their statistical distribution characteristics of each risk transmission path, screen key uncertainty sources and generate a list of key uncertainty sources.

[0062] In a preferred embodiment of the present invention, the screening of key uncertainty sources includes: sorting all risk transmission paths in descending order according to the cumulative uncertainty index.

[0063] Calculate the interquartile range of the cumulative uncertainty index in the sorted path sequence, and use the sum of the third quartile and the interquartile range of a preset multiple as the key uncertainty contribution threshold.

[0064] Traverse the path sequence and mark the paths whose cumulative uncertainty index is greater than or equal to the key uncertainty contribution threshold as high-risk paths.

[0065] Integrate the energy subsystem nodes that appear in all high-risk paths and their cumulative frequency of occurrence.

[0066] The cumulative occurrence frequency of all energy subsystem nodes is regarded as a set of observation data. The mean and standard deviation of the cumulative occurrence frequency are calculated. The mean is subtracted from the cumulative occurrence frequency of each energy subsystem node, and then divided by the standard deviation to obtain the deviation statistic of the cumulative occurrence frequency of each energy subsystem node.

[0067] Energy subsystem nodes whose deviation statistics exceed the critical value obtained by looking up the standard normal distribution table based on a pre-set reliability level are identified as significant outliers and included in the list of key uncertainty sources. For example, if the two-sided critical value of ±1.96 corresponding to a statistical confidence level of 95% is selected, and the deviation statistics of the cumulative frequency of an energy subsystem node are greater than +1.96 or less than -1.96, and the probability that it falls outside the 95% range of all data is less than 5%, it can be considered a statistically significant outlier, and the energy subsystem node is identified as a significant outlier.

[0068] This invention, through path analysis of the energy uncertainty map, identifies all risk transmission paths leading to the overall output node and calculates their cumulative uncertainty indices. Subsequently, instead of subjectively setting thresholds, it selects key uncertainty sources based on the statistical distribution characteristics of the cumulative uncertainty indices of all paths and comprehensively considers the cumulative frequency of nodes in high-risk paths. This avoids reliance on subjective experience or complex global analysis, and can objectively and efficiently locate the uncertainty sources that contribute most significantly to the overall system risk from historical data.

[0069] S5. Based on the list of key uncertainty sources, the scheduling parameters of the corresponding energy subsystems in the initial scheduling instructions are corrected to generate corrected scheduling instructions, which are then sent to the execution units of each energy subsystem.

[0070] In a preferred embodiment of the present invention, the step of correcting the scheduling parameters of the corresponding energy subsystem in the initial scheduling instruction includes: parsing the initial scheduling instruction and extracting the planned scheduling power corresponding to each energy subsystem in the list of key uncertainty sources.

[0071] Retrieve high-risk paths originating from each energy subsystem node in the list of key uncertainty sources. For each high-risk path, calculate a single metric value for the risk contribution of that path to the starting node. This metric value is obtained by dividing the cumulative uncertainty index of that path by the number of edges of that path. Sum the single metric values ​​of the high-risk paths originating from each energy subsystem node in the list of key uncertainty sources to calculate the comprehensive risk contribution of each energy subsystem in the list of key uncertainty sources.

[0072] Based on the main response direction of each energy subsystem in the list of key uncertainty sources when it is the initial source of deviation in the historical time series causal analysis, if the main response direction is positive, the power correction direction is determined to be downward, with room for upward adjustment to cope with its possible negative deviation; if the main response direction is negative, the power correction direction is determined to be upward, with room for upward adjustment to cope with its possible positive deviation.

[0073] Based on the power correction direction and the overall risk contribution, the power correction coefficient of each energy subsystem in the list of key uncertainty sources is calculated. As an example, the formula for calculating the power correction coefficient is as follows: .

[0074] in, For the calculated list of key uncertainty sources, the first... Power correction coefficient for each energy subsystem The numbers are assigned to each energy subsystem in the list of critical uncertainty sources. , This is the sign function, and its value is determined by the power correction direction. Specifically, if the power correction direction is downward, then... If the power correction direction is upward, then , These are the maximum and minimum values ​​of the combined risk contribution of all energy subsystems in the list of key uncertainty sources. The first in the list of key uncertainty sources The overall risk contribution of each energy subsystem The preset overall risk buffer coefficient is used to control the overall preventive adjustment range to uncertainty. It is mainly achieved by conducting offline simulations on a large number of historical scheduling cycles in the early stage of development, analyzing the impact of different values ​​on the actual operating cost of the system and the reduction ratio of deviation exceeding the limit events, fitting the simulation curves of the two, and taking the value mapped by the intersection of the two curves as the recommended value for deployment.

[0075] The first in the list of key uncertainty sources The normalized comprehensive risk contribution of each energy subsystem represents the first... The relative risk ranking of each energy subsystem among all current key sources of uncertainty is used to ensure that the correction amount is comparable and adaptive under different scheduling cycles and risk levels.

[0076] The above formula for calculating the power correction coefficient integrates three decision-making factors—correction direction, overall risk preference, and relative risk magnitude of the energy subsystem—into a mathematical expression, thereby quantifying the qualitative scheduling strategy.

[0077] The corrected planned scheduling power is obtained by multiplying the power correction factor of each energy subsystem in the list of key uncertainty sources with its planned scheduling power.

[0078] Planned power allocation for other energy subsystems within the list of non-critical uncertainty sources will remain unchanged.

[0079] In a preferred embodiment of the present invention, the step of correcting the scheduling parameters of the corresponding energy subsystem in the initial scheduling instruction further includes: performing a feasibility verification on the corrected planned scheduling power of each energy subsystem in the list of key uncertainty sources: node-level static constraint verification: for the corrected planned scheduling power of each energy subsystem, verify whether it meets its own technical output upper and lower limits, as well as the maximum allowable ramp rate constraint relative to the actual output of the previous scheduling period.

[0080] System-level static constraint verification: Verify whether the sum of the revised planned scheduling power of all energy subsystems meets the system net load balance requirements of the current scheduling cycle, and the allowable deviation does not exceed the rated power balance upper limit.

[0081] Network-level transmission risk verification: Based on the energy uncertainty map and the corrected planned scheduling power, the propagation impact of potential deviations from key uncertainty sources along their high-risk transmission paths is simulated and calculated to verify whether they will trigger systemic over-limit risks. The systemic over-limit risk refers to the risk that, after the simulated deviation transmission, the virtual operating state of any node in the path, resulting from the superposition of the planned scheduling power and the simulated deviation, exceeds its safe operating range.

[0082] If all the above checks pass, they will be integrated into the corrected scheduling instructions.

[0083] If the verification fails, an iterative correction mechanism is initiated to locate one or more energy subsystems that caused the conflict and determine the conflict type to recalculate the power correction amount until the feasibility verification conditions are met or the preset maximum number of iterations is reached.

[0084] The specific process for determining the conflict type and recalculating the power correction is as follows: If it does not meet the node-level static constraint verification item, it is classified as a conflict source, and its corrected planned scheduling power is directly adjusted to its technical constraint boundary and this value is fixed.

[0085] If the system does not meet the system-level static constraint verification, it is classified as a Class II conflict source, and the system power rebalancing iteration is initiated: calculate the difference between the sum of the planned scheduling power after correction of all current key uncertainty sources and the system net load demand. If the difference is greater than 0, it means that the planned power of some nodes needs to be fine-tuned downward. If it is less than 0, it means that the planned power of some nodes needs to be fine-tuned upward. Therefore, among the other key uncertainty sources except for Class I conflict sources, they are arranged in ascending order according to their comprehensive risk contribution, and their power correction coefficients are fine-tuned according to the preset compensation until the system power is rebalanced.

[0086] If the network-level transmission risk verification is not met, it is classified as a third type of conflict source, and risk transmission suppression iteration is initiated: among the source nodes that cause high-risk transmission and the key intermediate nodes in the path, select those whose power correction coefficients still have room for adjustment, strengthen the correction direction with a preset step size to enhance their ability to resist potential deviations, until the simulated transmission risk drops below the threshold. The preset step size can be set according to the percentage of the planned scheduling power of the node, for example, adjusting its planned value by 1% in each iteration.

[0087] Reference Figure 3 As shown, in a preferred embodiment of the present invention, after the modified scheduling instructions are issued to the execution units of each energy subsystem, the method further includes a step of closed-loop feedback evaluation and dynamic map update of the scheduling execution effect: real-time acquisition of the actual scheduling power of each energy subsystem during the execution of the modified scheduling instructions.

[0088] The actual scheduling power is compared with the expected target values ​​of the initial scheduling command and the revised scheduling command, and the scheduling execution error index corresponding to each key uncertainty source is calculated.

[0089] Based on the aforementioned scheduling execution error index, the edge weights of relevant nodes in the energy uncertainty graph are updated in reverse. One method for updating edge weights in reverse is as follows: for a given edge, if the actual deviation of its source node triggers propagation in the current cycle and the response of the target node conforms to historical patterns, then the edge weight is multiplied by an enhancement factor greater than 1, such as 1.05. If propagation is not triggered or the response is abnormal, then it is multiplied by a decay factor less than 1, such as 0.95. Simultaneously, upper and lower bound constraints are applied to the updated edge weights, and a moving average is used for smoothing to prevent drastic weight fluctuations caused by a single update, ensuring the stability of the graph evolution.

[0090] The update trajectory of edge weights is continuously recorded over multiple scheduling cycles. The average coefficient of variation of edge weights over multiple scheduling cycles is calculated to obtain the overall rate of change of the graph structure. If the overall rate of change exceeds a preset warning threshold, the graph reconstruction process is triggered, and a new round of time-series causal analysis and uncertainty transmission path reconstruction process based on the latest historical data is automatically started to generate an energy uncertainty graph synchronized with the current state. The preset warning threshold is obtained by calibrating the baseline rate of change sequence collected during the historical stable operation phase using a statistical method of mean plus three times the standard deviation.

[0091] This invention provides targeted modifications only to the energy subsystems of the key uncertainty source list. It determines the power correction direction based on the main response direction in historical transmission and generates differentiated power correction coefficients based on the comprehensive risk contribution. It proposes a preventive and differentiated scheduling correction strategy for key risk transmission links and ensures the feasibility of the solution through a feasibility verification and iterative correction mechanism.

[0092] It should be noted that during the initialization phase of this invention, in the absence of historical scheduling data, all possible connection edges can be assigned the same initial weight based on the physical connection topology and design parameters of the energy subsystem, or the initial weights can be set using expert experience. After accumulating operational data over several scheduling cycles, the system switches to a graph construction mode based on historical data analysis.

[0093] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A distributed energy collaborative optimization scheduling method based on multi-source data fusion, characterized in that, include: Obtain the real-time available power and local predicted load of various energy sources in the current scheduling cycle, and construct the initial scheduling instructions according to the net surplus power; Based on historical scheduling performance analysis, an energy uncertainty graph containing various energy sources is constructed. The energy uncertainty graph represents various energy subsystems through nodes and characterizes the uncertainty transmission relationship between different energy subsystems through edge weights. The construction of the energy uncertainty map including various energy sources is implemented as follows: The planned and actual scheduling power of each energy subsystem is obtained within multiple historical scheduling periods, and a scheduling deviation sequence of each energy subsystem is generated. Each energy subsystem is taken as an initial deviation source in turn. For each event in the scheduling deviation sequence that exceeds the preset scheduling tolerance error, a time-series causal analysis is performed to determine the influence transmission characteristics when each energy subsystem is taken as an initial deviation source. For each energy subsystem node that serves as the initial source of deviation, the corresponding set of uncertainty transmission paths is reconstructed based on its influence transmission characteristics. Each uncertainty transmission path records an ordered sequence of nodes that start from the energy subsystem node that serves as the initial source of deviation, pass through one or more intermediate nodes, and ultimately affect the overall output node. By integrating all uncertainty transmission paths, an energy uncertainty map is obtained; Path analysis is performed on the energy uncertainty map to identify risk transmission paths from each energy subsystem node to the overall output node, and the cumulative uncertainty index of each risk transmission path is calculated. Based on the cumulative uncertainty indicators and their statistical distribution characteristics of each risk transmission path, key uncertainty sources are screened and a list of key uncertainty sources is generated. The screening of key uncertainty sources includes: Sort all risk transmission paths in descending order based on cumulative uncertainty indicators; Calculate the interquartile range of the cumulative uncertainty index in the sorted path sequence, and use the sum of the third quartile and the interquartile range of a preset multiple as the key uncertainty contribution threshold. Traverse the path sequence and mark the paths whose cumulative uncertainty index is greater than or equal to the key uncertainty contribution threshold as high-risk paths; Integrate the energy subsystem nodes and their cumulative frequency of occurrence in all high-risk paths; Energy subsystem nodes that have appeared more frequently than a preset significant level will be included in the list of key uncertainty sources. Based on the list of key uncertainty sources, the scheduling parameters of the corresponding energy subsystems in the initial scheduling instructions are corrected to generate corrected scheduling instructions, which are then sent to the execution units of each energy subsystem.

2. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 1, characterized in that, The initial scheduling instruction construction process includes: The net surplus power of each type of energy is obtained by subtracting the real-time available power from the local predicted load in the current scheduling cycle. By loading the scheduling constraints and energy conversion efficiency of controllable devices in each energy subsystem, the original net surplus power is corrected. With the goal of minimizing the total operating cost of the current scheduling cycle, the planned scheduling power of each type of energy in the current scheduling cycle is calculated, and the initial scheduling instructions are generated by encapsulating them according to a predefined instruction format.

3. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 1, characterized in that, The time-series causal analysis process includes: For any two energy subsystems, calculate the cross-correlation function between the scheduling adjustment amount and the scheduling deviation of the other energy subsystem under multiple time lags in the same scheduling period when one energy subsystem has a scheduling deviation. Based on the peak value of the cross-correlation function, determine whether the transmission relationship between the two energy subsystems exists. If it is determined that the relationship exists, determine the transmission parameters including the response direction, response time lag and response amplitude. By aggregating the analysis results of all historical scheduling events, the frequency of occurrence of transmission relationships between any two energy subsystems is statistically analyzed, and stable transmission relationships with both occurrence frequency and average response amplitude higher than the corresponding confidence level are selected. Using all energy subsystems as nodes and all stable transmission relationships guided by them as initial deviation sources as directed edges, a set of uncertain transmission paths is constructed. The edge weight of each directed edge is obtained by setting a penalty factor through the average response time delay and correcting the basic weight calculated based on the average response amplitude and occurrence frequency.

4. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 1, characterized in that, The risk transmission path identification process includes: Obtain the fitting change slope and mean scheduling deviation of the scheduling deviation sequence of each energy subsystem. According to the predefined initial screening rules, select energy subsystems with inherent uncertain properties, mark them as the first type of uncertainty source, and include them in the set of key attention. For the remaining energy subsystems not marked by the initial screening rules, the current external environmental impact parameters that are strongly correlated with them are acquired in real time. By comparing them with the operating specifications of the energy subsystem, it is determined whether the current external environmental impact parameters constitute a disturbance that causes its operating state to deviate from the expected benchmark. If so, the energy subsystem is marked as a second type of uncertainty source and included in the set of key concerns. Using the overall output node as the sink node, all directed paths that can reach the sink node are searched starting from all energy subsystem nodes within the set to be focused on, serving as risk transmission paths.

5. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 1, characterized in that, The cumulative uncertainty index of each risk transmission path is the cumulative calculation result of the edge weights of all edges on the path.

6. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 1, characterized in that, The modification of the corresponding energy subsystem scheduling parameters in the initial scheduling instruction includes: Analyze the initial scheduling instructions and extract the planned scheduling power corresponding to each energy subsystem in the list of key uncertainty sources; Retrieve high-risk paths originating from each energy subsystem node in the list of key uncertainty sources, and calculate the comprehensive risk contribution of each energy subsystem in the list of key uncertainty sources by using the cumulative uncertainty index and the number of path edges of the high-risk paths. Based on the main response direction of each energy subsystem in the list of key uncertainty sources when it is used as the initial source of deviation in time-series causal analysis, the power correction direction is determined. Based on the power correction direction and the comprehensive risk contribution, the power correction coefficient of each energy subsystem in the list of key uncertainty sources is calculated, and multiplied by its planned scheduling power to obtain the corrected planned scheduling power; Planned power allocation for other energy subsystems within the list of non-critical uncertainty sources will remain unchanged.

7. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 6, characterized in that, The modification of the corresponding energy subsystem scheduling parameters in the initial scheduling instruction also includes: Feasibility verification is performed on the revised planned scheduling power of each energy subsystem in the list of key uncertainty sources. If the verification passes, the revised scheduling instructions are integrated into the revised instructions. If the verification fails, an iterative correction mechanism is initiated to locate one or more energy subsystems that caused the conflict and determine the conflict type to recalculate the power correction amount until the feasibility verification conditions are met or the preset maximum number of iterations is reached.

8. The distributed energy collaborative optimization scheduling method based on multi-source data fusion according to claim 1, characterized in that, After issuing the revised scheduling instructions to the execution units of each energy subsystem, the process also includes closed-loop feedback evaluation of the scheduling execution effect and dynamic updating of the map. Real-time acquisition of the actual dispatch power of each energy subsystem during the execution of the revised dispatch instructions; The actual scheduling power is compared with the expected target values ​​of the initial scheduling command and the revised scheduling command, and the scheduling execution error index corresponding to each key uncertainty source is calculated. Based on the scheduling execution error index, the edge weights of relevant nodes in the energy uncertainty graph are updated in reverse. The update trajectory of edge weights is continuously recorded within multiple scheduling cycles, and the overall change rate of the graph structure is calculated. If the overall change rate exceeds the preset warning threshold, the graph reconstruction process is triggered, and a new round of time-series causal analysis and uncertainty transmission path reconstruction process based on the latest historical data is automatically started to generate an energy uncertainty graph synchronized with the current state.

Images