Distributed energy storage multi-scenario configuration method considering bidirectional power flow

Abstract

The present application relates to the technical field of optimal configuration of power distribution network energy storage, and specifically discloses a distributed energy storage multi-scenario configuration method considering bidirectional power flow, which collects source-load time series data sets of new energy output, load and upper interactive power composition of each node of the power distribution network; extracts space-time coupling rules to construct a multi-dimensional bearing capacity index set, normalizes the weight to obtain a comprehensive bearing capacity coefficient, and corrects the bidirectional power flow margin of each node to obtain a bearing capacity boundary value; according to the boundary value and the source-load data, a plurality of operation scenarios are constructed to generate charge-discharge power time series, and through correlation decoupling, a decoupled power sequence of each energy storage unit is obtained; a multi-dimensional comprehensive benefit index is established, and the capacity and power are taken as decision variables for iterative optimization, and the optimal configuration scheme of the comprehensive benefit is output. The present application provides a reliable operation boundary constraint for energy storage configuration, and provides a quantifiable decision basis for distributed energy storage capacity planning under different operation requirements.

Description

Technical Field

[0001] This invention relates to the field of energy storage optimization configuration technology for distribution networks, and specifically to a multi-scenario configuration method for distributed energy storage that considers bidirectional power flow. Background Technology

[0002] In the power distribution network sector, with the increasing penetration rate of distributed power sources, the power flow of the distribution network is gradually changing from the traditional unidirectional radial pattern to a bidirectional power flow characteristic, significantly increasing the uncertainty on both the source and load sides. Distributed energy storage systems, due to their fast response speed and flexible power regulation, have become an important means to improve the operating status of the distribution network and promote the local consumption of new energy sources.

[0003] Under bidirectional power flow conditions in the distribution network, the actual adjustable margin of each node changes dynamically due to the differences in the spatiotemporal coupling characteristics between renewable energy output and load. Existing methods, which use a unified fixed boundary, cannot accurately reflect the true carrying capacity of each node. At the same time, under the multi-scenario operation requirements, there is a lack of effective correlation assessment and decoupling processing between the charging and discharging power sequences of energy storage, which causes the power commands of multiple energy storage units to couple and interfere with each other, resulting in the final configuration result failing to take into account the actual needs of different operation scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a distributed energy storage multi-scenario configuration method that takes into account bidirectional power flow, so as to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A multi-scenario configuration method for distributed energy storage considering bidirectional power flow includes the following steps:

[0007] S1, collect the new energy output time-series data and load time-series data of each access node of the target distribution network, as well as the power interaction data of the upper-level power grid, to form a source-load time-series dataset;

[0008] S2. Based on the source-load time series dataset, extract the spatiotemporal coupling characteristics of new energy output and load at each node, and construct a multi-dimensional carrying capacity index set based on the spatiotemporal coupling characteristics. The multi-dimensional carrying capacity index set is used to characterize the distribution network's ability to accept distributed power sources.

[0009] S3 normalizes and weights the multi-dimensional carrying capacity index set, solves the comprehensive carrying capacity coefficient of the distribution network, and corrects the bidirectional power flow margin of each node based on the comprehensive carrying capacity coefficient to obtain the carrying capacity boundary value of each node.

[0010] S4. Based on the bearing capacity boundary values ​​and source load time series datasets of each node, construct multiple typical operating scenarios for distributed energy storage, and generate multi-scenario charging and discharging power time series sets according to the priority order of each scenario. Then, perform correlation calculation and decorrelation processing on the multi-scenario charging and discharging power time series sets to obtain the decoupled operating power sequence of each energy storage unit.

[0011] S5. Based on the decoupled operating power sequence and source-load time series dataset, establish a multi-dimensional comprehensive benefit evaluation index set, perform weighted aggregation on the multi-dimensional comprehensive benefit evaluation index set, use energy storage configuration capacity and configuration power as decision variables, adopt adaptive perturbation search for iterative optimization, and output the configuration capacity and configuration power of each energy storage unit that maximize comprehensive benefits.

[0012] As a further aspect of the present invention: the construction of a multi-dimensional bearing capacity index set based on the spatiotemporal coupling characteristics specifically includes:

[0013] The source-load time series dataset is divided into preset time windows. The hourly changes in the output of new energy sources and the hourly changes in the load of each node within each time window are calculated to obtain the output fluctuation sequence and load fluctuation sequence of each node.

[0014] Based on the electrical distance between distribution network nodes, the output fluctuation sequence of adjacent nodes is superimposed and correlated with the load fluctuation sequence to generate a spatiotemporal coupling coefficient matrix for characterizing the intensity of bidirectional power transmission.

[0015] Based on the spatiotemporal coupling coefficient matrix, combined with the voltage safe operating range and line thermal stability limit of each node, the upper and lower limits of the acceptable power of each node are calculated and summarized into a multi-dimensional load-bearing capacity index set.

[0016] As a further aspect of the present invention: the generation of the spatiotemporal coupling coefficient matrix for characterizing the bidirectional power transfer intensity specifically includes:

[0017] Extract the fluctuation amount at the same time in the output fluctuation sequence and load fluctuation sequence of each node, and subtract the output fluctuation amount and load fluctuation amount of the same node at the same time to obtain the net fluctuation time series sequence of each node.

[0018] The spatial coupling weight is determined based on the electrical distance between distribution network nodes. The spatial coupling weight is then used to cross-weight the net fluctuation time series of adjacent nodes to obtain the associated fluctuation series of each node affected by adjacent nodes.

[0019] Based on the node connection relationship, the associated fluctuation sequences of each node are aligned and arranged according to their corresponding time positions to form a spatiotemporal coupling coefficient matrix.

[0020] As a further aspect of the present invention: obtaining the bearing capacity boundary values ​​of each node specifically includes:

[0021] Extract the real-time values ​​of each index in the multi-dimensional bearing capacity index set at multiple consecutive sampling times, and use the relative displacement of the real-time value of each index with respect to its historical fluctuation range as the normalized value.

[0022] Calculate the consistency of the changing trends among the normalized value sequences of each indicator, and determine the differentiated weighting values ​​of each indicator based on the consistency of the changing trends, so that the differentiated weighting values ​​are negatively correlated with the consistency.

[0023] The comprehensive carrying capacity coefficient obtained after weighted aggregation is used as the common scaling ratio of the bidirectional power flow adjustable margin of each node. The scaling direction is determined according to the current voltage offset direction of each node, and the upper and lower limits of the power flow adjustable margin of each node are adjusted in reverse to form the carrying capacity boundary value of each node.

[0024] As a further aspect of the present invention: the calculation of the consistency of the changing trends among the normalized value sequences of each indicator specifically includes:

[0025] Select the sequence with the largest median fluctuation from the normalized value sequences of all indicators as the benchmark change sequence;

[0026] The normalized value sequence of each remaining indicator is compared with the benchmark change sequence at each time step to determine whether it fluctuates in the same direction. The proportion of the time when each sequence changes in the same direction as the benchmark change sequence is used as the initial value of the trend consistency of each indicator.

[0027] The initial values ​​of trend consistency are sorted in ascending order, and weights are assigned in decreasing order to make the contribution of indicators with higher trend consistency smaller in subsequent aggregation.

[0028] As a further aspect of the present invention: obtaining the decoupled operating power sequence of each energy storage unit specifically includes:

[0029] Extract the upper and lower limits of adjustable power from the bearing capacity boundary values ​​of each node, and combine them with the net load sequence in the source load time series dataset. Divide the operating period into surplus period, deficit period and stable period according to the positive and negative polarity and amplitude of the net load.

[0030] Based on surplus periods, deficit periods, and stable periods, adjacent periods with the same attribute are merged into scene segments. The scene segments are then sorted in descending order according to the total energy deficit or surplus of each scene segment to determine the priority of each scene segment.

[0031] According to the priority order, the charging and discharging power of the corresponding energy storage unit is allocated to each scene segment, and the power transition is smoothed at the junction of adjacent scene segments by means of moving average, forming a multi-scene charging and discharging power time series set;

[0032] The correlation degree of synchronous change between each pair of sequences in the multi-scenario charging and discharging power time series is calculated. The power components of the sequence pairs with the correlation degree of synchronous change exceeding the threshold are separated and regrouped so that the correlation degree of synchronous change between the remaining sequences is reduced to below the threshold, thus obtaining the decoupled operation power sequence of each energy storage unit.

[0033] As a further aspect of the present invention: the smoothing of power transition at the junction of adjacent scene pieces using a moving average method to form a multi-scene charging and discharging power timing set specifically includes:

[0034] Identify the transition moment between every two adjacent scene pieces, and define the transition window by extending a preset number of sampling points forward and backward from the transition moment as the center;

[0035] Within the transition window, the power sequence at the end of the previous scene piece and the power sequence at the beginning of the next scene piece are extracted. They are assigned linearly decaying weights according to their distance from the transition time, and the weighted average is calculated time by time to obtain the replacement power value at each time within the transition window.

[0036] The original power values ​​within the transition window are replaced with the corresponding replacement power values ​​in a sequential manner, and the results are merged to form a continuous multi-scenario charging and discharging power time series.

[0037] As a further aspect of the present invention: the output that optimizes the overall efficiency of each energy storage unit's configuration capacity and power specifically includes:

[0038] Based on the bearing capacity boundary value and decoupled operation power sequence of each energy storage unit, the initial range of configuration capacity and configuration power of each energy storage unit is determined, and multiple candidate solutions are randomly generated within the range of values;

[0039] Calculate the comprehensive benefit value of each candidate solution under the multidimensional comprehensive benefit evaluation index set, and select the candidate solution with the largest comprehensive benefit value as the current dominant solution;

[0040] Centered on the current dominant solution, a neighborhood candidate solution group is generated according to the current search radius. Candidate solutions in the neighborhood candidate solution group with a comprehensive benefit value greater than the current dominant solution are updated as the new dominant solution, and the search radius is reduced by a preset ratio.

[0041] If the current dominant solution is not updated after multiple iterations, the search radius is restored to the initial value and the process proceeds to generating a neighborhood candidate solution group until the search radius converges to a preset threshold. The final dominant solution is then output as the optimal configuration capacity and configuration power for each energy storage unit.

[0042] As a further aspect of the present invention: the generation of neighborhood candidate solution groups specifically includes:

[0043] Obtain the values ​​of each decision variable in the current dominant solution. Using the current value of each decision variable as the center point, multiply the current search radius by the span of the initial value range of each decision variable to obtain the directional search step size corresponding to each decision variable.

[0044] Offset the directional search step size along the positive and negative directions of each decision variable to generate marginal candidate solutions when each decision variable changes individually. Then, offset each decision variable by 1 / 2 of the directional search step size to generate intermediate candidate solutions.

[0045] All edge candidate solutions and intermediate candidate solutions are combined to form a neighborhood candidate solution group.

[0046] The beneficial effects of this invention are:

[0047] (1) By extracting the spatiotemporal coupling characteristics of new energy output and load of each node, a multi-dimensional carrying capacity index set is constructed and the comprehensive carrying capacity coefficient is solved. Based on this coefficient, the upper and lower limits of the adjustable bidirectional power flow of each node are differentiated and reversed, so that the carrying capacity boundary value can be dynamically and adaptively adjusted with the direction of node voltage offset. This avoids the problem of the distribution network operating status exceeding the limit due to the bidirectional flow of power flow after the distributed power source is connected, and provides reliable operating boundary constraints for energy storage configuration.

[0048] (2) By constructing multiple typical operating scenarios and generating multi-scenario charging and discharging power time series sets according to priority, and then performing correlation calculation and decorrelation processing on the time series sets, the charging and discharging power sequences of each energy storage unit are independent in time, avoiding the coupling conflict of power commands between multiple scenarios; combined with the adaptive perturbation search with the goal of optimal comprehensive benefits, the configuration capacity and configuration power of each energy storage unit are directly output, providing a quantifiable decision basis for distributed energy storage capacity planning under different operating requirements. Attached Figure Description

[0049] The invention will now be further described with reference to the accompanying drawings.

[0050] Figure 1 This is a flowchart of the method of the present invention;

[0051] Figure 2 This is a flowchart illustrating the process of constructing a multi-dimensional bearing capacity index set based on the spatiotemporal coupling characteristics in this invention.

[0052] Figure 3 This is a flowchart illustrating the process of obtaining the bearing capacity boundary values ​​of each node in this invention. Detailed Implementation

[0053] 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.

[0054] Please see Figure 1 As shown, this invention is a multi-scenario configuration method for distributed energy storage considering bidirectional power flow, comprising the following steps:

[0055] S1, collect the new energy output time-series data and load time-series data of each access node of the target distribution network, as well as the power interaction data of the upper-level power grid, to form a source-load time-series dataset;

[0056] S2. Based on the source-load time series dataset, extract the spatiotemporal coupling characteristics of new energy output and load at each node, and construct a multi-dimensional carrying capacity index set based on the spatiotemporal coupling characteristics. The multi-dimensional carrying capacity index set is used to characterize the distribution network's ability to accept distributed power sources.

[0057] S3 normalizes and weights the multi-dimensional carrying capacity index set, solves the comprehensive carrying capacity coefficient of the distribution network, and corrects the bidirectional power flow margin of each node based on the comprehensive carrying capacity coefficient to obtain the carrying capacity boundary value of each node.

[0058] S4. Based on the bearing capacity boundary values ​​and source load time series datasets of each node, construct multiple typical operating scenarios for distributed energy storage, and generate multi-scenario charging and discharging power time series sets according to the priority order of each scenario. Then, perform correlation calculation and decorrelation processing on the multi-scenario charging and discharging power time series sets to obtain the decoupled operating power sequence of each energy storage unit.

[0059] S5. Based on the decoupled operating power sequence and source-load time series dataset, establish a multi-dimensional comprehensive benefit evaluation index set, perform weighted aggregation on the multi-dimensional comprehensive benefit evaluation index set, use energy storage configuration capacity and configuration power as decision variables, adopt adaptive perturbation search for iterative optimization, and output the configuration capacity and configuration power of each energy storage unit that maximize comprehensive benefits.

[0060] In S1, the time-series data of renewable energy output and load at each access node of the target distribution network, as well as the power interaction data of the upstream power grid, are collected to form a source-load time-series dataset, which specifically includes:

[0061] At each wind power grid-connected node and photovoltaic power grid-connected node in the target distribution network, a bidirectional power acquisition terminal is deployed. The acquisition terminal is connected to the grid line through current transformers and voltage transformers. It continuously records the instantaneous value of the active power of the new energy grid-connected nodes at a sampling frequency of once per second, and stores the time stamp information of the sampling time together to form the new energy output time sequence data of each access node.

[0062] At each load access node of the target distribution network, a bidirectional power acquisition terminal is also deployed to continuously record the instantaneous value of active power consumed by each load node according to the same sampling frequency and time scale reference, forming load time series data.

[0063] At the grid connection point where the target distribution network connects to the upper-level grid, a gate power acquisition terminal is deployed. This terminal collects the instantaneous value of active power injected from the upper-level grid to the distribution network or sent back from the distribution network to the upper-level grid through the current transformer and voltage transformer at the grid connection point. The data is recorded according to the same sampling frequency and unified time scale as the aforementioned acquisition terminal to form the upper-level grid interactive power data.

[0064] All the aforementioned acquisition terminals upload the acquisition time and corresponding instantaneous active power value to the data aggregation server via fiber optic communication network. The data aggregation server uses the sampling time as the primary key to align the time-series data of renewable energy output from each access node, the time-series data of load from each load node, and the power interaction data from the upper-level power grid according to time sequence, and merge them into a source-load time-series dataset covering the complete sampling period for subsequent steps.

[0065] Please see Figure 2 As shown, in S2, based on the source-load time-series dataset, the spatiotemporal coupling characteristics of new energy output and load at each node are extracted. A multi-dimensional carrying capacity index set is then constructed based on these characteristics. This multi-dimensional carrying capacity index set is used to characterize the distribution network's ability to accept distributed power sources, specifically including:

[0066] The source-load time-series dataset is divided according to a preset time window. The preset time window length is set to 15 minutes, meaning that data from 900 consecutive sampling moments constitute one time window. The data aggregation server reads the renewable energy output time-series data and load time-series data of each access node from the source-load time-series dataset, and divides the entire sampling period into several consecutive and non-overlapping time windows according to the chronological order of sampling moments, with a step size of 900 sampling points.

[0067] Within each time window, for each access node, the instantaneous value of renewable energy output at the end of the time window is subtracted from the instantaneous value of renewable energy output at the beginning of the time window to obtain the hourly change in renewable energy output of the node within the time window; the instantaneous value of load active power at the end of the time window is subtracted from the instantaneous value of load active power at the beginning of the time window to obtain the hourly change in load of the node within the time window.

[0068] By iterating through all time windows and all access nodes in this manner, the hourly changes in renewable energy output and load of each access node within each time window are obtained. The hourly changes in renewable energy output of the same access node across all time windows are arranged in chronological order to form the output fluctuation sequence of that node; the hourly changes in load of the same access node across all time windows are arranged in chronological order to form the load fluctuation sequence of that node.

[0069] Based on the electrical distance between distribution network nodes, the output fluctuation sequence and load fluctuation sequence of adjacent nodes are superimposed and correlated. The line impedance value between any two nodes is read from the distribution network topology, and the sum of the per-unit values ​​of all line impedance values ​​between the two nodes is taken as the electrical distance between the two nodes.

[0070] For each access node, first extract the fluctuations within the same time window from its output fluctuation sequence and load fluctuation sequence. Subtract the load fluctuation from the output fluctuation within the same time window to obtain the net fluctuation of that node within that time window. Arrange the net fluctuations of all time windows in chronological order to form the net fluctuation time series sequence of that node.

[0071] For any given node, obtain all adjacent nodes directly connected to that node via power distribution lines. Take the reciprocal of the electrical distance between the current node and each adjacent node to obtain the reciprocal of the distance between each adjacent node. Sum the reciprocals of the distances between all adjacent nodes to obtain the total sum of the reciprocals of the distances.

[0072] Divide the reciprocal of the distance to each adjacent node by the sum of the reciprocals of those distances to obtain the spatial coupling weight for each adjacent node. Use these spatial coupling weights to cross-weight the net volatility time series of the current node: for each time window, multiply the net volatility of the current node in that time window by 1, then subtract the sum of the net volatility of all adjacent nodes in that time window multiplied by their respective spatial coupling weights. This yields the associated volatility of the current node in that time window due to the influence of its adjacent nodes.

[0073] Arrange the associated fluctuations of all time windows in chronological order to form the associated fluctuation sequence of the current node. Iterate through all nodes in this way to obtain the associated fluctuation sequence of each node.

[0074] According to the node connection relationship of the distribution network, the associated fluctuation sequences of all nodes are aligned window by window in chronological order, so that each time window corresponds to a row and each node corresponds to a column. The associated fluctuation of each node in the time window is filled into the corresponding position to form a spatiotemporal coupling coefficient matrix.

[0075] Read the voltage safe operating range of each node from the distribution network dispatching system. The upper limit of the voltage safe operating range is 1.05 times the rated voltage, and the lower limit is 0.95 times the rated voltage. Read the thermal stability limit of each line. The thermal stability limit is 0.8 times the rated transmission capacity of the line.

[0076] For each node, multiply each associated fluctuation in the column corresponding to that node in the spatiotemporal coupling coefficient matrix by the upper and lower limits of the aforementioned voltage safe operating range to obtain the upper and lower voltage limit constraints for that node in that time window; multiply the associated fluctuation of that node by the thermal stability limit of the line connected to that node and divide by the rated transmission capacity of that line to obtain the thermal stability constraint value for that node in that time window.

[0077] The minimum value among the upper voltage limit constraint, lower voltage limit constraint, and thermal stability constraint within each time window is taken as the upper limit of the acceptable power for that node, and the lower voltage limit constraint is taken as the lower limit of the acceptable power for that node. The upper and lower limits of the acceptable power for each node within each time window are summarized according to time sequence and node number to form a multi-dimensional bearing capacity index set.

[0078] Please see Figure 3 As shown, in S3, the multi-dimensional carrying capacity index set is normalized and weighted to solve for the comprehensive carrying capacity coefficient of the distribution network. Based on the comprehensive carrying capacity coefficient, the bidirectional power flow margin of each node is corrected to obtain the carrying capacity boundary value of each node. Specifically, this includes:

[0079] The real-time values ​​of each indicator at multiple consecutive sampling times are extracted from a set of multi-dimensional carrying capacity indicators. The real-time values ​​of all sampling times over the past 24 hours are taken from the current time to form the real-time value sequence of each indicator.

[0080] For each indicator, the maximum value among all real-time values ​​of the indicator in the past 24 hours is used as the upper bound of the historical fluctuation range, and the minimum value is used as the lower bound of the historical fluctuation range.

[0081] For the real-time value of the indicator at the current moment, calculate its relative displacement with respect to the historical fluctuation range. Specifically, subtract the lower bound of the historical fluctuation range from the real-time value to obtain the numerator; subtract the lower bound of the historical fluctuation range from the upper bound to obtain the denominator; divide the numerator by the denominator, and the quotient is the normalized value of the indicator at the current moment.

[0082] If the real-time value is less than the lower bound of the historical fluctuation range, the normalized value is 0; if the real-time value is greater than the upper bound of the historical fluctuation range, the normalized value is 1. The normalized value of each indicator is obtained by processing its real-time value in this way. The calculation expression is as follows:

[0083] ;

[0084] in, For the first The normalized value of each indicator. For the first The real-time value of each indicator at the current moment. For the first The lower bound of the historical fluctuation range of each indicator. For the first The upper limit of the historical fluctuation range of each indicator.

[0085] Calculate the consistency of the changing trends among the normalized value sequences of each indicator. Read the normalized values ​​of each indicator at all sampling times over the past 24 hours from the data aggregation server to form the normalized value sequence of each indicator.

[0086] For each indicator, calculate the absolute value of the difference between two adjacent sampling times in its normalized value sequence, sort all the absolute values ​​of adjacent differences from largest to smallest, and take the median to obtain the median fluctuation range of the indicator.

[0087] The normalized value sequence of the indicator with the largest median fluctuation amplitude is selected as the benchmark change sequence. For each remaining indicator, its normalized value sequence is compared with the benchmark change sequence time by time: starting from the second sampling time, if the direction of change of the normalized value of the indicator at the current time relative to the previous time is the same as the direction of change of the benchmark change sequence at the current time relative to the previous time, it is recorded as a change in the same direction.

[0088] The initial value of the trend consistency of this indicator is obtained by counting the number of times the indicator changes in the same direction as the baseline change sequence, dividing by the total number of times in the sequence minus one. The initial values ​​of the trend consistency of all remaining indicators are calculated in the same way.

[0089] Sort the initial trend consistency values ​​of all indicators (including those corresponding to the baseline change sequence, with an initial consistency value of 1) in ascending order, assigning each indicator a sorting number. Set a total of... The index is numbered as follows: For smaller metrics, the formula for calculating differentiated weights is as follows:

[0090] ;

[0091] in, For the sequence number is the The weighting values ​​corresponding to smaller indicators, This represents the total number of indicators. As can be seen from the calculation expression for differentiated weighting, the lower the trend consistency (i.e., the smaller the ranking number), the larger the weighting value of the indicator, thus achieving a negative correlation between weighting value and consistency. The sum of the weighting values ​​of all indicators equals 1.

[0092] The normalized value of each indicator is multiplied by its corresponding differentiated weighted value to obtain the weighted value of each indicator. The weighted values ​​of all indicators are then summed to obtain the comprehensive carrying capacity coefficient. This coefficient is a dimensionless value between 0 and 1.

[0093] The bidirectional power flow margin of each node is corrected based on the comprehensive carrying capacity coefficient. The current measured voltage value of each node is read from the distribution network dispatching system, and the voltage offset direction of each node is calculated: if the measured voltage value of the node is greater than the rated voltage, the offset direction is recorded as positive; if it is less than the rated voltage, the offset direction is recorded as negative; if it is equal to the rated voltage, the offset direction is recorded as zero.

[0094] For each node, obtain the current bidirectional power flow adjustable upper and lower limits for that node. The upper limit represents the maximum power value that the node is allowed to inject into the distribution network, and the lower limit represents the maximum power value that the node is allowed to absorb from the distribution network.

[0095] If the voltage offset direction of the node is positive, the comprehensive bearing capacity coefficient is used as a reduction ratio, the upper limit is multiplied by the comprehensive bearing capacity coefficient to obtain a new upper limit, and the lower limit is multiplied by the comprehensive bearing capacity coefficient to obtain a new lower limit; if the offset direction is negative, the comprehensive bearing capacity coefficient is used as a expansion ratio, the upper limit is divided by the comprehensive bearing capacity coefficient to obtain a new upper limit, and the lower limit is divided by the comprehensive bearing capacity coefficient to obtain a new lower limit; if the offset direction is zero, the original upper and lower limits remain unchanged.

[0096] After the aforementioned reverse adjustment, the updated adjustable upper and lower limits of the power flow at each node become the bearing capacity boundary values ​​for that node. The bearing capacity boundary values ​​of all nodes are then aggregated for use in subsequent steps.

[0097] In S4, based on the bearing capacity boundary values ​​and source load time series datasets of each node, multiple typical operating scenarios of distributed energy storage are constructed. Then, multi-scenario charging and discharging power time series sets are generated according to the priority order of each scenario. Finally, correlation calculation and decorrelation processing are performed on the multi-scenario charging and discharging power time series sets to obtain the decoupled operating power sequence of each energy storage unit, specifically including:

[0098] Identify the connection point between any two adjacent scene slices. The data aggregation server traverses the temporal arrangement of all scene slices. For any two temporally adjacent scene slices, the boundary point between the last sampling time of the preceding scene slice and the first sampling time of the following scene slice is determined as the connection point between the two scene slices.

[0099] Centered on this transition moment, a range of 11 sampling points is defined, extending forward by 5 sampling points and backward by 5 sampling points. Extending forward means retrospectively moving 5 sampling points towards the historical timeline, while extending backward means advancing 5 sampling points towards the future timeline.

[0100] If the forward expansion exceeds the starting boundary of the previous scene piece, then the starting boundary of that scene piece is taken as the actual starting point of the forward expansion; if the backward expansion exceeds the ending boundary of the next scene piece, then the ending boundary of that scene piece is taken as the actual ending point of the backward expansion.

[0101] The weighted average is calculated time-by-time within the transition window. The power sequence at the end of the previous scene slice is extracted from the multi-scene charge and discharge power time series set. This sequence consists of the charge and discharge power values ​​corresponding to all sampling points belonging to the previous scene slice within the transition window. The power sequence at the beginning of the next scene slice is extracted. This sequence consists of the charge and discharge power values ​​corresponding to all sampling points belonging to the next scene slice within the transition window.

[0102] For each sampling moment within the transition window, a linear attenuation weight is assigned based on the time distance between that sampling moment and the transition moment. Specifically, for each sampling point before the transition moment, the difference in the number of sampling points between that sampling point and the transition moment is calculated. This difference is then divided by the total number of forward-extended sampling points (5) to obtain the attenuation coefficient of the foreground scene power value. This attenuation coefficient is 0 at the transition moment and 1 at the furthest point of the forward extension. The intermediate sampling points are determined by linear interpolation. The weight of the foreground scene power value at that moment is equal to 1 minus this attenuation coefficient.

[0103] For each sampling point after the connection time, calculate the difference in the number of sampling points from the connection time. Divide the difference in the number of sampling points by the total number of sampling points extended backward, 5, to obtain the growth coefficient of the power value of the subsequent scene. This growth coefficient is 0 at the connection time and 1 at the farthest point of the backward extension. The intermediate sampling points are determined by linear interpolation. The weight of the power value of the subsequent scene at that time is equal to this growth coefficient.

[0104] At the transition moment itself, the weights of the preceding scene power value and the following scene power value are both 0.5. The permutation power value at each sampling moment within the transition window is obtained by multiplying the preceding scene power value by its weight at that moment, and then adding the following scene power value by its weight at that moment.

[0105] The charging and discharging power values ​​are represented by positive numbers to indicate the discharge direction and negative numbers to indicate the charging direction. The original positive and negative signs are retained during the weighted calculation. Therefore, the replacement power value shows a gradual transition from the power value of the original scene piece to the power value of the adjacent scene piece within the transition window.

[0106] The charging and discharging power values ​​at each sampling time within the transition window are replaced sequentially with the replacement power values ​​at the corresponding time, while the charging and discharging power values ​​at other sampling times outside the transition window remain unchanged.

[0107] The power values ​​of all scene slices are concatenated in chronological order to form a complete multi-scene charging and discharging power time series, which undergoes the smooth transition processing described above at all junctions of adjacent scene slices. Each sampling moment in this time series corresponds to a definite power value, and there are no abrupt power jumps at the junctions of adjacent scene slices.

[0108] In S5, a multi-dimensional comprehensive benefit evaluation index set is established based on the decoupled operating power sequence and source-load time series dataset. This index set is then weighted and aggregated. Using energy storage configuration capacity and configuration power as decision variables, an adaptive perturbation search is employed for iterative optimization. The result is the output of the optimal configuration capacity and configuration power for each energy storage unit that maximizes the comprehensive benefit. Specifically, this includes:

[0109] In each iteration of the optimization process, once the current dominant solution is determined, the data aggregation server obtains the specific values ​​of all decision variables contained in the current dominant solution.

[0110] The decision variables include the configured capacity and configured power values ​​of each energy storage unit. Assuming the total number of energy storage units participating in the optimization is M, the total number of decision variables is 2M, where the first M decision variables correspond to the configured capacity values ​​of each energy storage unit, and the last M decision variables correspond to the configured power values ​​of each energy storage unit.

[0111] For each decision variable, read its upper and lower limits from the initial value range determined in the first stage of step five, calculate the difference between the upper and lower limits, and obtain the span of the initial value range of the decision variable.

[0112] Simultaneously, the current search radius used in the current iteration is read. The initial value of this search radius is 0.5, which is reduced or restored in subsequent iterations based on updates to the search state. For any decision variable, the initial range of values ​​for that decision variable is multiplied by the current search radius, and the product is used as the directional search step size for that decision variable.

[0113] For each decision variable, generate marginal candidate solutions according to the directional search step size corresponding to that decision variable.

[0114] For each decision variable, two marginal candidate solutions are generated: The first marginal candidate solution is based on the current dominant solution, but the value of the decision variable is shifted positively by one directional search step, that is, the new value of the decision variable is equal to the current value plus the directional search step, while the values ​​of the other decision variables remain exactly the same as the current dominant solution; The second marginal candidate solution is based on the current dominant solution, but the value of the decision variable is shifted negatively by one directional search step, that is, the new value of the decision variable is equal to the current value minus the directional search step, while the values ​​of the other decision variables remain exactly the same as the current dominant solution.

[0115] If, after a positive shift, the new value of the decision variable exceeds the upper limit of its initial value range, then the new value is truncated to that upper limit. If, after a negative shift, the new value of the decision variable is lower than the lower limit of its initial value range, then the new value is truncated to that lower limit. This process is repeated for all 2M decision variables to obtain 4M marginal candidate solutions.

[0116] A new intermediate candidate solution is generated: the new values ​​of all decision variables in this intermediate candidate solution are based on the corresponding decision variable values ​​in the current dominant solution, while offset by half of the directional search step size of the corresponding decision variable. That is, for each decision variable, if its value in the current dominant solution is V, then the value of that decision variable in the intermediate candidate solution is V plus 0.5 times the directional search step size. If the offset exceeds the initial value range, truncation is also performed.

[0117] All marginal and intermediate candidate solutions are aggregated and merged to form a neighborhood candidate solution group. This neighborhood candidate solution group contains (4M plus 1) candidate solutions, each consisting of 2M decision variable values, and each candidate solution falls within the initial value range of each decision variable. This neighborhood candidate solution group is provided to subsequent steps to calculate the comprehensive benefit value of each candidate solution and compare it with the current dominant solution to determine whether to update the dominant solution.

[0118] The working principle of this invention is as follows: A source-load time-series dataset is constructed by collecting time-series data of renewable energy output, load, and power interaction data from each access node of the target distribution network. Then, the spatiotemporal coupling characteristics of renewable energy output and load at each node are extracted from this dataset. Based on these spatiotemporal coupling characteristics, a multi-dimensional carrying capacity index set is constructed to characterize the distribution network's ability to accept distributed generation. This multi-dimensional carrying capacity index set is then normalized and weighted to solve for the comprehensive carrying capacity coefficient of the distribution network. Based on this comprehensive carrying capacity coefficient, the bidirectional power flow margin of each node is corrected to obtain the carrying capacity boundary of each node. The system first calculates the load capacity boundary values ​​of each node and the source-load time series dataset, then constructs multiple typical operating scenarios for distributed energy storage based on the load capacity boundary values ​​of each node and the source-load time series dataset. It then generates a multi-scenario charging and discharging power time series set according to the priority of each scenario, performs correlation calculation and decorrelation processing on this multi-scenario charging and discharging power time series set to obtain the decoupled operating power sequence of each energy storage unit. Finally, it establishes a multi-dimensional comprehensive benefit evaluation index set based on this decoupled operating power sequence and the source-load time series dataset, and performs weighted aggregation on it. Using the energy storage configuration capacity and configuration power as decision variables, it uses adaptive perturbation search for iterative optimization, outputting the configuration capacity and configuration power of each energy storage unit that maximizes the comprehensive benefit.

[0119] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A multi-scenario configuration method for distributed energy storage considering bidirectional power flow, characterized in that, Includes the following steps: S1, collect the new energy output time-series data and load time-series data of each access node of the target distribution network, as well as the power interaction data of the upper-level power grid, to form a source-load time-series dataset; S2. Based on the source-load time series dataset, extract the spatiotemporal coupling characteristics of new energy output and load at each node, and construct a multi-dimensional carrying capacity index set based on the spatiotemporal coupling characteristics. The multi-dimensional carrying capacity index set is used to characterize the distribution network's ability to accept distributed power sources. S3 normalizes and weights the multi-dimensional carrying capacity index set, solves the comprehensive carrying capacity coefficient of the distribution network, and corrects the bidirectional power flow margin of each node based on the comprehensive carrying capacity coefficient to obtain the carrying capacity boundary value of each node. S4. Based on the bearing capacity boundary values ​​and source load time series datasets of each node, construct multiple typical operating scenarios for distributed energy storage, and generate multi-scenario charging and discharging power time series sets according to the priority order of each scenario. Then, perform correlation calculation and decorrelation processing on the multi-scenario charging and discharging power time series sets to obtain the decoupled operating power sequence of each energy storage unit. S5. Based on the decoupled operating power sequence and source-load time series dataset, establish a multi-dimensional comprehensive benefit evaluation index set, perform weighted aggregation on the multi-dimensional comprehensive benefit evaluation index set, use energy storage configuration capacity and configuration power as decision variables, adopt adaptive perturbation search for iterative optimization, and output the configuration capacity and configuration power of each energy storage unit that maximize comprehensive benefits.

2. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 1, characterized in that, The multi-dimensional bearing capacity index set constructed based on the spatiotemporal coupling characteristics specifically includes: The source-load time series dataset is divided into preset time windows. The hourly changes in the output of new energy sources and the hourly changes in the load of each node within each time window are calculated to obtain the output fluctuation sequence and load fluctuation sequence of each node. Based on the electrical distance between distribution network nodes, the output fluctuation sequence of adjacent nodes is superimposed and correlated with the load fluctuation sequence to generate a spatiotemporal coupling coefficient matrix for characterizing the intensity of bidirectional power transmission. Based on the spatiotemporal coupling coefficient matrix, combined with the voltage safe operating range and line thermal stability limit of each node, the upper and lower limits of the acceptable power of each node are calculated and summarized into a multi-dimensional load-bearing capacity index set.

3. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 2, characterized in that, The generation of the spatiotemporal coupling coefficient matrix used to characterize the strength of bidirectional power transfer specifically includes: Extract the fluctuation amount at the same time in the output fluctuation sequence and load fluctuation sequence of each node, and subtract the output fluctuation amount and load fluctuation amount of the same node at the same time to obtain the net fluctuation time series sequence of each node. The spatial coupling weight is determined based on the electrical distance between distribution network nodes. The spatial coupling weight is then used to cross-weight the net fluctuation time series of adjacent nodes to obtain the associated fluctuation series of each node affected by adjacent nodes. Based on the node connection relationship, the associated fluctuation sequences of each node are aligned and arranged according to their corresponding time positions to form a spatiotemporal coupling coefficient matrix.

4. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 1, characterized in that, The process of obtaining the bearing capacity boundary values ​​for each node specifically includes: Extract the real-time values ​​of each index in the multi-dimensional bearing capacity index set at multiple consecutive sampling times, and use the relative displacement of the real-time value of each index with respect to its historical fluctuation range as the normalized value. Calculate the consistency of the changing trends among the normalized value sequences of each indicator, and determine the differentiated weighting values ​​of each indicator based on the consistency of the changing trends, so that the differentiated weighting values ​​are negatively correlated with the consistency. The comprehensive carrying capacity coefficient obtained after weighted aggregation is used as the common scaling ratio of the bidirectional power flow adjustable margin of each node. The scaling direction is determined according to the current voltage offset direction of each node, and the upper and lower limits of the power flow adjustable margin of each node are adjusted in reverse to form the carrying capacity boundary value of each node.

5. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 4, characterized in that, The calculation of the consistency of the changing trends among the normalized value sequences of each indicator specifically includes: Select the sequence with the largest median fluctuation from the normalized value sequences of all indicators as the benchmark change sequence; The normalized value sequence of each remaining indicator is compared with the benchmark change sequence at each time step to determine whether it fluctuates in the same direction. The proportion of the time when each normalized value sequence changes in the same direction as the benchmark change sequence is calculated to be used as the initial value of the trend consistency of each indicator. The initial values ​​of trend consistency are sorted in ascending order, and weights are assigned in decreasing order to make the contribution of indicators with higher trend consistency smaller in subsequent aggregation.

6. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 1, characterized in that, The process of obtaining the decoupled operating power sequence of each energy storage unit specifically includes: Extract the upper and lower limits of adjustable power from the bearing capacity boundary values ​​of each node, and combine them with the net load sequence in the source load time series dataset. Divide the operating period into surplus period, deficit period and stable period according to the positive and negative polarity and amplitude of the net load. Based on surplus periods, deficit periods, and stable periods, adjacent periods with the same attribute are merged into scene segments. The scene segments are then sorted in descending order according to the total energy deficit or surplus of each scene segment to determine the priority of each scene segment. According to the priority order, the charging and discharging power of the corresponding energy storage unit is allocated to each scene segment, and the power transition is smoothed at the junction of adjacent scene segments by means of moving average, forming a multi-scene charging and discharging power time series set; The correlation degree of synchronous change between each pair of sequences in the multi-scenario charging and discharging power time series is calculated. The power components of the sequence pairs with the correlation degree of synchronous change exceeding the threshold are separated and regrouped so that the correlation degree of synchronous change between the remaining sequences is reduced to below the threshold, thus obtaining the decoupled operation power sequence of each energy storage unit.

7. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 6, characterized in that, The process of smoothing power transitions at the junctions of adjacent scene segments using a moving average method to form a multi-scene charging and discharging power timing set specifically includes: Identify the transition moment between every two adjacent scene pieces, and define the transition window by extending a preset number of sampling points forward and backward from the transition moment as the center; Within the transition window, the power sequence at the end of the previous scene piece and the power sequence at the beginning of the next scene piece are extracted. They are assigned linearly decaying weights according to their distance from the transition time, and the weighted average is calculated time by time to obtain the replacement power value at each time within the transition window. The original power values ​​within the transition window are replaced with the corresponding replacement power values ​​in a sequential manner, and the results are merged to form a continuous multi-scenario charging and discharging power time series.

8. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 1, characterized in that, The output optimizes the overall efficiency of each energy storage unit's configuration capacity and power, specifically including: Based on the load-bearing capacity boundary value and decoupled operation power sequence of each energy storage unit, the initial range of configuration capacity and configuration power of each energy storage unit is determined, and multiple candidate solutions are randomly generated within the initial range. Calculate the comprehensive benefit value of each candidate solution under the multidimensional comprehensive benefit evaluation index set, and select the candidate solution with the largest comprehensive benefit value as the current dominant solution; Centered on the current dominant solution, a neighborhood candidate solution group is generated according to the current search radius. Candidate solutions in the neighborhood candidate solution group with a comprehensive benefit value greater than the current dominant solution are updated as the new dominant solution, and the search radius is reduced by a preset ratio. If the current dominant solution is not updated after multiple iterations, the search radius is restored to the initial value and the process proceeds to generating a neighborhood candidate solution group until the search radius converges to a preset threshold. The final dominant solution is then output as the optimal configuration capacity and configuration power for each energy storage unit.

9. The distributed energy storage multi-scenario configuration method considering bidirectional power flow as described in claim 8, characterized in that, The generation of neighborhood candidate solution groups specifically includes: Obtain the values ​​of each decision variable in the current dominant solution. Using the current value of each decision variable as the center point, multiply the current search radius by the span of the initial value range of each decision variable to obtain the directional search step size corresponding to each decision variable. Offset the directional search step size along the positive and negative directions of each decision variable to generate marginal candidate solutions when each decision variable changes individually. Then, offset each decision variable by 1 / 2 of the directional search step size to generate intermediate candidate solutions. All edge candidate solutions and intermediate candidate solutions are combined to form a neighborhood candidate solution group.

Images