Method for operation optimization of main and distribution collaborative power distribution network based on data analysis

Abstract

The application discloses a main-distribution coordination power grid operation optimization method based on data analysis and belongs to the technical field of power grid operation control. The method collects the bus voltage on each side of a main grid transformer, tie line power, and the output of a distributed power source of a distribution grid, the state of charge of energy storage, and the interruptible capacity of a flexible load, and constructs a boundary power exchange characteristic vector of the main-distribution grid; according to the correlation degree of the characteristic vector and the over-limit risk of each node voltage of the distribution grid, a boundary coupling node set is screened, a virtual power injection equivalent device is arranged at the set, and the support capacity of the main grid to the distribution grid is quantified as an adjustable power interval of the boundary node; the adjustable power interval and the regulation dead zone of each distributed power source in the distribution grid are subjected to intersection operation to generate an initial feasible region of main-distribution coordination regulation and control; and the on-load voltage regulating switch position of the main grid transformer, the capacity of the capacitor bank, and the charging and discharging power of the energy storage of the distribution grid and the flexible load reduction amount are subjected to joint iterative optimization with the feasible region as a constraint, so that the regulation and control reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of distribution network operation control technology, specifically to a method for optimizing the operation of a distribution network based on data analysis and coordination between the main and distribution systems. Background Technology

[0002] In optimizing the coordinated operation of distribution networks, it is necessary to coordinate the on-load tap changer and capacitor switching of main grid transformers with adjustable resources such as energy storage and flexible loads on the distribution network side to address issues such as voltage exceedance caused by the high proportion of distributed power sources. Traditional methods often simplify the main grid as an infinite power source or use fixed equivalent parameters, making it difficult to accurately reflect the correlation between power exchange at the boundary between the main grid and the distribution network and the risk of voltage exceedance. When determining the boundary nodes for coordinated control, there is often reliance on electrical distance experience or traversal of all boundary nodes, resulting in dimensional redundancy in the optimization model, excessive computational burden, and easy introduction of weakly correlated nodes that interfere with the control effect. At the same time, the support capacity of the main grid for the distribution network is often set as a fixed power exchange range, failing to be dynamically quantified based on real-time operating data and transmission margin. This results in the ineffective utilization of the main grid's flexible resources or the occurrence of power exceedance after control due to overly optimistic ranges.

[0003] Distributed power inverters within distribution networks typically have a dead zone, during which their output power does not respond to external commands. Current main-distribution coordinated optimization methods often ignore this practical constraint, directly treating the rated output range of equipment as the feasible space. This leads to some optimization commands falling into the dead zone and becoming unexecutable, inducing repeated adjustments or exacerbating power fluctuations. Furthermore, the on-load tap changer positions of main grid transformers and capacitor bank switching are discrete variables, while the energy storage charging and discharging power and flexible load reduction in the distribution network are continuous variables. When solving these variables together, existing methods often employ a unified continuous relaxation or a single heuristic algorithm, failing to perform direction-sensitive exploratory optimization for discrete variables, and failing to embed feedback on active power imbalances caused by discrete main grid adjustments in the continuous variable search. This makes it difficult to balance solution stability and coordinated effects. In distribution network operation, there exists a true adjustable range determined by the main grid boundary support and the dead zone within the distribution network. The lack of effective construction and utilization of this range causes optimization results to deviate from actual execution conditions. Therefore, it is urgent to solve how to accurately extract the dominant boundary coupling nodes from the perspective of data correlation and dynamically quantify the main network support range, and how to integrate the distributed power supply regulation dead zone into the collaborative feasible domain and implement joint optimization that takes into account the characteristics of discrete and continuous variables. Summary of the Invention

[0004] This invention provides a data analysis-based method for optimizing the operation of a main-distribution coordinated distribution network. By constructing a boundary power exchange feature vector and using correlation to screen dominant boundary coupling nodes, the method uses virtual power injection to quantify the main grid's support capability. Then, the method intersects the quantization interval with the distributed power generation regulation dead zone to generate an initial feasible region. Under this constraint, the method iteratively optimizes the discrete variables of the main grid and the continuous variables of the distribution network, thereby achieving complete matching of the control strategy with the actual adjustable range of the equipment and efficient and reliable main-distribution coordinated operation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a method for optimizing the operation of a main distribution network based on data analysis, comprising: By collecting data on the bus voltages on each side of the main grid transformer, the active and reactive power of the tie lines, and the output of distributed power sources, the state of charge of energy storage, and the interruptible capacity of flexible loads at each node of the distribution network, a power exchange feature vector at the boundary between the main grid and the distribution network is constructed. This feature vector can fully characterize the electrical quantity interaction state and temporal variation trend between the main grid and the distribution network at the boundary, providing a high-dimensional information foundation for subsequent collaborative analysis.

[0006] Based on the correlation between the power exchange feature vector at the main distribution network boundary and the voltage exceedance risk at each node of the distribution network, a set of boundary coupling nodes that dominate the coordinated operation of the main and distribution networks is selected. By quantifying the correlation between the boundary power exchange characteristics and the voltage security risks within the distribution network, the set of key nodes that have the greatest impact on the operation of the distribution network can be accurately identified from numerous boundary nodes, effectively reducing the scale of coordinated control and improving the efficiency of subsequent optimization calculations.

[0007] A virtual power injection equivalent device is set at the set of boundary coupling nodes to quantify the support capacity of the main grid to the distribution network into the adjustable power range of the boundary nodes. By using joint modeling of virtual generators and virtual loads, the power support capacity of the main grid is explicitly expressed in the distribution network model, so that the regulation potential of the main grid is transformed into boundary constraints that can be directly used for distribution network optimization.

[0008] The adjustable power range is intersected with the regulation dead zones of each distributed power source within the distribution network to generate the initial feasible region for coordinated control of the main and distribution systems. By calculating the overlap between the external support range and the inherent insensitive regulation zone within the internal system, power supply devices that are unable to respond to control commands due to dead zone limitations are automatically excluded, ensuring that every control variable within the initial feasible region is practically executable.

[0009] Using the initial feasible region as a constraint, a joint iterative optimization is performed on the on-load tap changer positions of the main grid transformers, the switching capacity of capacitor banks, the charging and discharging power of each energy storage unit in the distribution network, and the reduction of flexible loads, to output a coordinated operation control strategy for the main grid and distribution network. This joint optimization can simultaneously take into account the timing coordination relationship between the discrete regulation equipment of the main grid and the continuous regulation resources of the distribution network, avoiding the problems of oscillation or mutual cancellation of control effects caused by inconsistent regulation pace between the two types of equipment, and ultimately obtaining a globally coordinated operation scheme.

[0010] As a technical solution of this invention, when constructing the power exchange feature vector at the boundary of the main distribution network, the active power, reactive power, and voltage amplitude of the low-voltage side bus of the main grid transformer are aligned by timestamps and then concatenated into an original feature sequence. The difference value and moving standard deviation of this sequence over three consecutive scheduling cycles are then calculated and appended to the end of the original feature sequence. Introducing the difference value and moving standard deviation can capture the abrupt changes and volatility of the power exchange process, ensuring that the feature vector not only contains static information of the current section but also reflects recent dynamic changes, thus improving the accuracy of predicting voltage exceedance risks in the distribution network.

[0011] As a technical solution of this invention, when screening the set of boundary coupled nodes, the mutual information value between each element in the power exchange feature vector of the main distribution network boundary and the voltage over-limit risk index of each node in the distribution network is calculated. The distribution network boundary nodes corresponding to the elements whose mutual information values ​​exceed a preset threshold are marked as coupled nodes. Then, redundant nodes that are simultaneously covered by at least two distributed power sources within the same main transformer power supply area are removed from the coupled nodes. Mutual information can effectively measure the nonlinear statistical dependence between feature elements and voltage over-limit risk, and is more suitable for complex distribution network scenarios than traditional correlation coefficients. Removing redundant nodes avoids the repeated calculation of control resources caused by multiple coverage by distributed power sources.

[0012] As a technical solution of this invention, when setting up a virtual power injection equivalent device at the boundary coupling node set, a virtual generator and a virtual load are bound to each node respectively. The upper limit of the active power output of the virtual generator is set equal to the maximum transmittable active power of the main grid tie line at that node, and the upper limit of the active power absorption of the virtual load is set equal to the maximum load demand of the downstream distribution network of that node. Then, the output or absorption states of the virtual generator and the virtual load are mutually exclusive and locked. The mutual exclusion locking mechanism ensures that at any given time, the boundary node only behaves as a power injection node or a power absorption node, which is completely consistent with the characteristic that the same tie line cannot simultaneously send and receive power in the actual physical system, thus avoiding the occurrence of physically unrealizable bidirectional simultaneous power flow in the optimization results.

[0013] As a technical solution of this invention, when generating the initial feasible region through intersection operation, the active and reactive power regulation dead zone boundary values ​​of each distributed power source are read from the inverter control parameters of each distributed power source in the distribution network. The adjustable power range is projected onto the output coordinate system of each distributed power source, and the overlap length between the projected range and the regulation dead zone boundary value is calculated. Distributed power sources with an overlap length greater than zero are marked as effective units that can participate in the main grid-distribution coordinated control. Then, the output regulation ranges of all effective units are intersected. This solution can accurately calculate the spatial overlap between the multi-dimensional space of the main grid support range and the constraint space formed by the dead zones of each distributed power source, ensuring that the power allocation scheme represented by any point in the initial feasible region can be simultaneously satisfied by the transmission capacity of the main grid tie line and the regulation margin of each distributed power source.

[0014] As a technical solution of this invention, during joint iterative optimization, the on-load tap changer positions of the main grid transformer and the switching capacity of the capacitor bank are encoded as the first set of decision variables, and the charging and discharging power of each energy storage unit in the distribution network and the reduction amount of flexible load are encoded as the second set of decision variables. With the first set of decision variables fixed, the particle swarm optimization algorithm is used to optimize the second set of decision variables. Then, with the second set of decision variables fixed, the discrete gradient descent method is used to optimize the first set of decision variables. These two optimization steps are executed alternately until the difference in the objective function between the two iterations is less than the convergence threshold. This alternating solution architecture treats discrete and continuous variables separately, each employing the optimization algorithm best suited to its mathematical characteristics. This avoids the computational complexity explosion problem of mixed integer programming and ensures that the discrete regulation on the main grid side and the continuous regulation on the distribution network side gradually approach the global optimum within a unified iterative framework.

[0015] Preferably, when using the particle swarm optimization algorithm to optimize the second set of decision variables, the charging and discharging power of each energy storage unit and the reduction amount of each flexible load are concatenated into a particle position vector. The value range of each dimension of the particle position vector is set to the projection interval of the initial feasible region on the corresponding dimension. An active power imbalance penalty term caused by the change of the main grid transformer tap position is introduced into the speed update formula, and the sum of the penalty term and the voltage deviation of the distribution network nodes is used as the particle fitness. Explicitly introducing the active power imbalance penalty term into the fitness function can drive the particle swarm to actively avoid schemes that cause excessive power mismatch between the main grid and the distribution network during the search process, significantly improving the optimization efficiency and the actual executability of the final solution.

[0016] Preferably, when optimizing the first set of decision variables using the discrete gradient descent method, the on-load tap changer positions of the transformer are discretized into integer variables with a difference of one between adjacent positions, and the switching capacity of the capacitor bank is discretized into integer variables that are integer multiples of the single-bank capacity. The sensitivity of the objective function to unit changes in each integer variable is calculated. Each integer variable is then tested sequentially in descending order of sensitivity, moving one unit towards reducing the objective function. If the voltage at each node of the distribution network does not exceed the limit after the test, the step is retained; otherwise, the state before the step is returned. This single-step testing strategy based on sensitivity ranking gives the optimization process of discrete variables a clear direction and controllability. Each test only triggers a limited number of power flow calculations, meeting the strict time requirements of online applications.

[0017] As a technical solution of this invention, when outputting the main distribution coordinated operation control strategy, the on-load tap changer positions of the main grid transformer and the switching capacity of the capacitor bank obtained through joint iterative optimization are converted into a remote control command sequence executable by the SCADA system. The charging and discharging power of each energy storage unit in the distribution network is decomposed into the form of active power target value plus dead zone hysteresis band, and the flexible load reduction is decomposed into a load level sequence cut off in each round. Then, the remote control command sequence, active power target value plus dead zone hysteresis band, and load level sequence are packaged into a download frame for different execution terminals. Adaptive command formats are generated for different execution devices, breaking down the protocol barriers between the optimization calculation results and the field control terminal, enabling the main distribution coordinated operation control strategy to be directly issued and executed without the need for manual translation in the intermediate links.

[0018] Preferably, when the charging and discharging power of each energy storage unit in the distribution network is decomposed into the form of an active power target value plus a dead zone hysteresis band, an upper dead zone bandwidth and a lower dead zone bandwidth are set for each energy storage unit. A hysteresis interval is constructed with the active power target value as the center, the upper dead zone bandwidth as the upper boundary, and the lower dead zone bandwidth as the lower boundary. When the current actual power of the energy storage unit is lower than the lower boundary of the hysteresis interval, a charging command is sent; when the current actual power of the energy storage unit is higher than the upper boundary of the hysteresis interval, a discharging command is sent; when the current actual power of the energy storage unit is within the hysteresis interval, the charging and discharging state of the previous moment remains unchanged. The setting of the hysteresis interval effectively prevents the energy storage unit from frequently switching between charging and discharging states due to measurement noise or small power fluctuations, thereby improving the service life and operational stability of the energy storage equipment.

[0019] The technical effects and advantages provided by the present invention in the above technical solution are as follows: A boundary power exchange feature vector with additional differential values ​​and sliding standard deviations is constructed from the bus voltages on each side of the main grid transformer, the active and reactive power of the tie lines, the output of distributed generation sources in the distribution network, the state of charge of energy storage, and the interruptible capacity of flexible loads. The mutual information values ​​between each element of this vector and the voltage exceedance risk indicators of each node in the distribution network are calculated. Boundary nodes with mutual information values ​​exceeding a preset threshold are marked as coupled nodes, and redundant nodes simultaneously covered by at least two distributed generation sources within the same main transformer power supply area are removed, resulting in a set of boundary coupled nodes. A virtual power injection equivalent device is set for each node in the set, binding virtual generators and virtual loads. The upper limit of the active power output of the virtual generator is set to the maximum active power that the main grid tie line can transmit at that node, and the upper limit of the active power absorption of the virtual load is set to the maximum load demand of the downstream distribution network, with both being forcibly mutually exclusive and locked. This method can extract the coupled nodes that have a significant impact on voltage over-limit using only boundary data correlation analysis, and accurately quantify the real-time support capability of the main network into the adjustable power range of the nodes. It does not require detailed modeling of the main network, thus avoiding boundary quantification deviations caused by incomplete or simplified main network models. At the same time, it removes redundant nodes and compresses the collaborative optimization space dimension.

[0020] The obtained adjustable power range of the boundary nodes is projected onto the output coordinate system of each distributed power source in the distribution network. The active and reactive power regulation dead zone boundary values ​​in the control parameters of each power source inverter are read, and the overlap length between the projected range and the dead zone boundary value is calculated. Distributed power sources with an overlap length greater than zero are identified as effective units that can participate in the main distribution coordinated control. Then, the intersection of the output adjustment ranges of all effective units is taken to generate the initial feasible region of main distribution coordinated control. This feasible region excludes output areas that are inaccessible due to inverter dead zones, ensuring that subsequent optimization searches always occur within the actual adjustable range of the equipment. Under this constraint, the on-load tap changer position of the main grid transformer, the switching capacity of the capacitor bank, the charging and discharging power of the distribution network energy storage, and the reduction of flexible loads are iteratively optimized alternately. Energy storage power and flexible load reduction are encoded as particle position vectors, with each dimension limited to the projection range of the initial feasible region. A penalty term for active power imbalance caused by changes in the main grid transformer tap position is introduced into the particle swarm velocity update. The sum of this penalty term and the node voltage deviation is used as the fitness. The top quarter of elite particles are retained and Gaussian mutation is performed, driving continuous variable particles to actively evolve in a direction that compensates for voltage deviations and offsets main grid regulation disturbances. For discrete decision variables, transformer tap positions are discretized into integer variables with a difference of one between adjacent tap positions, and capacitor switching capacities are discretized into integer variables that are integer multiples of a single capacity. The sensitivity to unit changes is calculated, and the steps towards reducing the objective function are attempted sequentially from largest to smallest sensitivity. The variable is retained only if the node voltage does not exceed the limit after each step; otherwise, it backs down. This alternating mode allows discrete variables to progressively optimize while ensuring voltage safety, while continuous variables collaboratively respond to power surges caused by main grid regulation during the particle swarm search. Each instruction in the resulting collaborative control strategy originates from the actual adjustable range of the equipment, eliminating unexecutable instructions and ineffective oscillations caused by regulation dead zones. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0022] Figure 1 This is a flowchart of a data analysis-based method for optimizing the operation of a coordinated distribution network. Figure 2 This is a schematic diagram of the equivalent device for virtual power injection at boundary coupling nodes; Figure 3 This is a flowchart of the effective unit selection and initial feasible domain generation of distributed power sources based on the overlap of power projection and dead zone; Figure 4 It is a distribution diagram of the mutual information values ​​between the power exchange feature vector elements at the main distribution network boundary and the voltage over-limit risk of distribution network nodes; Figure 5 This is a schematic diagram of the adjustable power range of the virtual power injection equivalent device at the boundary coupling node; Figure 6 This is a schematic diagram of the active charge and discharge target values ​​and dead zone / hysteresis band of the energy storage unit. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0024] See Figure 1 This invention provides a data analysis-based method for optimizing the operation of a main grid and distribution network in a coordinated manner. This method collects operational data from the main grid and distribution network, constructs a power exchange feature vector at the boundary of the main grid and distribution network, and selects a set of boundary coupling nodes based on the correlation between this vector and the voltage over-limit risk of distribution network nodes. Virtual power injection equivalent devices are set at the boundary coupling nodes to quantify the main grid's support capacity into an adjustable power range. An initial feasible region is generated through intersection calculations with the dead zone of distributed power generation. This region is then used as a constraint to jointly iteratively optimize the on-load tap changer position of transformers, the switching capacity of capacitor banks, the charging and discharging power of energy storage units, and the reduction of flexible loads. Finally, a coordinated operation control strategy for the main grid and distribution network is output.

[0025] Example 1:

[0026] In practice, the system collects data on the bus voltages on each side of the main grid transformer, the active and reactive power of the tie lines, the output of distributed power sources, the state of charge of energy storage, and the interruptible capacity of flexible loads at each node of the distribution network. This data is obtained from the dispatch automation system, the distribution network energy management system, and the load control system, with timestamp alignment granularity consistent with the dispatch cycle.

[0027] The active power, reactive power, and voltage amplitude of the low-voltage side bus of the main grid transformer are aligned by timestamps and then concatenated to form the original feature sequence. Within a scheduling cycle, the active power, reactive power, and voltage amplitude of the low-voltage side bus of the main grid transformer each form a vector with a length equal to the number of sampling points. These three vectors are concatenated sequentially to form the original feature sequence corresponding to that scheduling cycle.

[0028] For the original feature sequence, calculate the difference and moving standard deviation of the original feature sequence over three consecutive scheduling periods. For each scalar position in the original feature sequence, subtract the value of the same scalar position in the previous scheduling period from the value of the current scheduling period at that scalar position to obtain a difference value. Repeat this operation to obtain two difference value sequences over three scheduling periods: the difference value of the current scheduling period relative to the previous scheduling period, and the difference value of the previous scheduling period relative to the scheduling period before that. The moving standard deviation is calculated as follows: using the current scheduling period as a reference, for each scalar position in the original feature sequence, take the value of the current scheduling period at that scalar position, the value of the previous scheduling period at that scalar position, and the value of the scheduling period before that at that scalar position. These three values ​​form a sample set, and the standard deviation of this sample set is calculated to obtain the moving standard deviation sequence.

[0029] The difference values ​​and the moving standard deviation are appended to the end of the original feature sequence to form the power exchange feature vector at the main distribution network boundary. During appending, the scalar values ​​of the two difference value sequences are concatenated sequentially after the original feature sequence according to their scalar positions, followed by the moving standard deviation sequence. The resulting power exchange feature vector at the main distribution network boundary has a dimension equal to four times the dimension of the original feature sequence.

[0030] Calculate the mutual information value between each element in the power exchange feature vector at the main distribution network boundary and the voltage exceedance risk index of each node in the distribution network. The voltage exceedance risk index of a distribution network node is defined as the proportion of the duration during which the voltage amplitude of a distribution network node exceeds the upper or lower voltage limit within a specified time window. The formula for calculating the mutual information value is as follows:

[0031] in, The first element in the power exchange eigenvector at the boundary of the main and distribution networks represents the... A set of possible values ​​for each element. Indicates the distribution network number The set of values ​​for the node voltage over-limit risk indicator. express A specific value, express A specific value, for and Simultaneous joint probability estimates, for The estimated marginal probability of occurrence, for The marginal probability estimate of the occurrence. The above probability estimate was obtained by frequency statistics on historical sample data.

[0032] The distribution network boundary nodes corresponding to elements whose mutual information values ​​exceed a preset threshold are marked as coupled nodes. The preset threshold is 0.15, which is determined by the analysis of the subject operating characteristic curve of the ability to distinguish between the normal operation state and the voltage over-limit state of the distribution network, and is obtained by maximizing the Youden index.

[0033] Redundant nodes are removed from the coupled nodes if they are simultaneously covered by at least two distributed generation sources within the same main transformer power supply area. The determination method is as follows: Obtain the distribution network topology and the location of the distributed generation sources. Starting from each distributed generation source, search along the distribution network lines for its power supply reach. If a coupled node is simultaneously within the power supply reach of at least two distributed generation sources, then that coupled node is marked as a redundant node and removed. The remaining coupled nodes after removing redundant nodes constitute the boundary coupled node set.

[0034] See Figure 4 In the figure, the horizontal axis represents the index of the main distribution network boundary power exchange feature vector element, ranging from 0 to approximately 1200, and the vertical axis represents the mutual information value. Gray dots represent elements with mutual information values ​​below a preset threshold of 0.15, while black dots represent elements with mutual information values ​​above the threshold of 0.15. The threshold is drawn as a dashed line at the 0.15 position on the vertical axis. It can be seen that the mutual information values ​​of most elements are distributed between 0 and 0.1, significantly lower than the threshold, indicating a weak correlation between these feature elements and the voltage exceedance risk at various nodes in the distribution network. However, a few elements have mutual information values ​​significantly higher than the 0.15 threshold, reaching a maximum of approximately 0.5, indicating a strong statistical dependence between these feature elements and the voltage exceedance risk at distribution network nodes. According to the mutual information value screening steps in Example 1, the main distribution network boundary nodes corresponding to elements above the threshold in the figure are marked as coupling nodes, which are then used for subsequent boundary coupling node set construction and virtual power injection equivalent device setup. The distribution trend of mutual information values ​​in the figure clearly demonstrates the rationality of the threshold of 0.15. This threshold effectively distinguishes key feature elements that are significantly related to voltage over-limit risk, and helps to achieve accurate identification of boundary coupling nodes.

[0035] Example 2:

[0036] See Figure 2 Virtual power injection equivalent devices are set up at the selected set of boundary coupled nodes. The virtual power injection equivalent devices are implemented through software modules. In the distribution network state estimation and power flow calculation framework, a virtual generator object and a virtual load object are bound to each node in the set of boundary coupled nodes.

[0037] Both virtual generators and virtual loads are injected as nodes in the static power flow model and do not depend on the operation of physical equipment. The upper limit of the active power output of the virtual generator is set as the maximum transmissible active power of the main grid tie line at the boundary coupling node. The maximum transmissible active power is obtained by extracting the transmission path from the main grid tie line to the boundary coupling node, collecting the active power transmission values ​​corresponding to the thermal stability limit, voltage drop limit, and static power angle stability limit in the transmission path, and taking the minimum of the above three active power transmission values ​​as the maximum transmissible active power. The lower limit of the active power output of the virtual generator is set to zero. The method for extracting transmission paths starts from the boundary coupling node and traces upstream along the direction of active power flow, passing through distribution network feeders, interconnecting transformers, and main grid interconnecting lines, until reaching the high-voltage side busbar of the main grid transformer, forming a set of paths containing all series-connected electrical equipment. This set of paths is obtained by breadth-first topology search based on the real-time operating power grid public information model in the dispatch automation system. During the search process, only the electrical connection paths formed by circuit breakers and disconnecting switches that are currently in a closed state are retained. When there are multiple transmission paths operating in parallel, the one with the smallest equivalent impedance among all paths is taken as the dominant transmission path. The active power transmission value corresponding to the thermal stability limit is calculated as follows: Obtain the long-term allowable current carrying capacity of each series-connected electrical device in the transmission path at the reference ambient temperature. Based on the difference between the current ambient temperature and the reference ambient temperature, calculate the current carrying capacity correction factor at the current ambient temperature using the temperature correction formula in IEC 60287. Multiply the long-term allowable current carrying capacity of each device by the correction factor to obtain the current allowable current carrying capacity of each device. Take the minimum value among the current allowable current carrying capacities of each device as the path allowable current. Multiply the path allowable current by the product of the path rated line voltage and the square root of three to obtain the active power transmission value corresponding to the thermal stability limit. The active power transmission value corresponding to the voltage sag limit is calculated by setting the allowable lower limit of voltage sag at the boundary coupling node to 0.90 times the per-unit voltage. Using the high-voltage side bus voltage of the main grid transformer as the reference voltage, obtain the total equivalent resistance and total equivalent reactance of the transmission path, and obtain the reactive power demand of the boundary coupling node under the lower limit of voltage sag. Ignoring the line charging capacitor, solve for the maximum active power transmission that satisfies the node voltage constraint using the following formula:

[0038] in, This represents the active power transmission value corresponding to the voltage sag limit. The voltage amplitude of the high-voltage side busbar of the main grid transformer. This represents the lower limit of the allowable voltage sag at the boundary coupling node. and These are the total equivalent resistance and total equivalent reactance of the transmission path, respectively. The load power factor angle at the boundary coupling node is taken as the angle corresponding to the historical average power factor of that node. If the above solution yields no real solution, the critical active power value that makes the voltage drop exactly equal to the allowable lower limit is taken. This critical value is obtained by iteratively searching between zero and the short-circuit capacity of the transmission path using a bisection method. The active power transmission value corresponding to the static power angle stability limit is calculated by obtaining the total equivalent reactance of the transmission path. The voltage amplitude of the high-voltage side bus of the main grid transformer Voltage amplitude at boundary coupling nodes The product divided by the total equivalent reactance Then multiply by the preset power angle safety factor. Thus, the active power transmission value corresponding to the static power angle stability limit is obtained, that is:

[0039] in, Take the per-unit value of the voltage amplitude of the boundary coupling node under normal operating conditions, and the power angle safety factor. The value ranges from 0.70 to 0.85, and is determined by combining the upper limit of the linear segment of the power angle characteristic curve of the transmission path with the transient stability margin requirement; when the transmission path includes a transformer, the total equivalent reactance... The transformer short-circuit impedance needs to be converted to the same voltage reference before being added to the line reactance. After obtaining the active power transmission value corresponding to the thermal stability limit, the active power transmission value corresponding to the voltage drop limit, and the active power transmission value corresponding to the static power angle stability limit, the minimum value among the three is taken as the maximum transmittable active power, and the calculation process is dynamically updated as the topology and load change during the online operation of the distribution network.

[0040] The upper limit for active power absorption of virtual loads is set as the maximum load demand of the downstream distribution network of the boundary coupling node. The maximum load demand is obtained by statistically analyzing the historical annual maximum power values ​​of all loads in the downstream distribution network of the boundary coupling node, and then adding a distribution network loss correction factor to these annual maximum power values. The distribution network loss correction factor is taken as the highest loss rate of the downstream distribution network under historical heavy load conditions plus 1. The highest loss rate is determined by dividing the network loss by the maximum ratio of total active power load in historical power flow calculation data. The lower limit for active power absorption of virtual loads is set to zero.

[0041] The mutual exclusion locking of the output or absorption states of virtual generators and virtual loads is achieved by setting mutual exclusion constraints. The mutual exclusion constraints are as follows: at any given time, the product of the active power output of the virtual generator and the active power absorption value of the virtual load is zero; when the active power output of the virtual generator is greater than zero, the active power absorption value of the virtual load is forced to zero; and when the active power absorption value of the virtual load is greater than zero, the active power output of the virtual generator is forced to zero. During the power flow calculation iteration, a mutual exclusion check is performed after each state update. If the mutual exclusion constraint is broken, the smaller value is set to zero, and the output or absorption value corresponding to the larger value is retained.

[0042] Through the aforementioned virtual power injection equivalent device, the main grid's support capacity for the distribution network is quantified into the adjustable power range of the boundary nodes. The adjustable power range has the upper limit of the virtual generator's active power output as the positive boundary and the negative value of the upper limit of the virtual load's active power absorption as the negative boundary. The adjustable power of the boundary nodes continuously takes values ​​between the positive and negative boundaries, and positive injection and negative absorption will not occur simultaneously.

[0043] See Figure 5 In the figure, the horizontal axis represents the time sampling point, and the vertical axis represents the active power (in MW). The red solid line depicts the trend of the upper limit of the virtual generator's active power output (positive boundary) over time in Example 2, while the blue dashed line depicts the trend of the negative value (negative boundary) of the virtual load's active power absorption limit over time. The numerical ranges of the two curves are approximately between +9MW and +11MW and between -7MW and -9MW, respectively, showing the fluctuations of the maximum transmittable active power and maximum load demand at the boundary nodes over time.

[0044] The gray-filled area in the figure represents the adjustable power range of the boundary coupling node. This range consists of the upper limit of the active power output of the virtual generator (positive boundary) and the negative value of the upper limit of the active power absorption of the virtual load (negative boundary). The adjustable power range is continuous and takes values ​​between the positive and negative boundaries. The curve fluctuations reflect the impact of the dynamic changes in the thermal stability limit, voltage drop limit, and static power angle stability limit of the main grid tie line, as well as the downstream load demand and network loss correction coefficient of the distribution network on the adjustable power range.

[0045] As can be seen from the figure, the upper limit of the active power output of the virtual generator and the negative value of the upper limit of the active power absorption of the virtual load remain clearly separated throughout the time range, with no overlap. This indicates that the mutual exclusion constraint set by the virtual power injection equivalent device is satisfied, ensuring that the boundary node will not simultaneously exhibit positive injection and negative absorption states, which meets the design requirements of mutual exclusion locking between the virtual generator and virtual load states in Example 2.

[0046] Furthermore, within the time sampling range, the upper limit curve of the virtual generator's active power output exhibits periodic fluctuations, reflecting the dynamic characteristics of the main grid's support capacity at different time points. Similarly, the negative curve of the virtual load's active power absorption upper limit also shows a certain fluctuation trend, reflecting the time-varying characteristics of the distribution network's load demand. The change in the overall gray area width represents the difference in the size of the adjustable power range of the boundary coupling node at different time points, providing accurate input for subsequent intersection calculations based on this adjustable power range and the distributed generation's regulation dead zone.

[0047] In summary, the figure accurately reflects the dynamic active power boundary setting and time-varying characteristics of the virtual power injection equivalent device in Example 2, providing an intuitive data basis for the quantification of the main grid support capacity and the construction of the adjustable power range in the optimization of the main grid coordinated distribution network operation.

[0048] Example 3:

[0049] In specific implementation, please refer to Figure 3 The active power regulation dead zone boundary value and reactive power regulation dead zone boundary value of each distributed power source are read from the inverter control parameters of each distributed power source in the distribution network. The inverter control parameters are stored in the local controller of the distributed power source's grid-connected inverter or in the equipment ledger of the distribution network energy management system. For the first... A distributed power source, read the lower limit of the active power regulation dead zone. Active regulation dead zone upper limit value Lower limit of reactive power adjustment dead zone upper limit of reactive power adjustment dead zone The above four dead zone boundary values ​​are all expressed in per-unit form based on the rated capacity of the distributed power source, and the value range is [0,1].

[0050] The adjustable power range of the boundary nodes is projected onto the output coordinate system of each distributed power source. The adjustable power range includes active and reactive dimensions. During projection, the active output axis of the distributed power source is used as the projection target axis, mapping the active dimension range of the adjustable power range to the projection range; the reactive output axis of the distributed power source is used as the projection target axis, mapping the reactive dimension range of the adjustable power range to the projection range. The lower bound of the active power projection range is denoted as... The upper bound of the active projection interval is denoted as The lower bound of the reactive projection interval is denoted as The upper bound of the reactive projection interval is denoted as .

[0051] Calculate the overlap length between the projection interval and the adjustment dead zone boundary value. The overlap length of the active dimension is calculated using the following formula:

[0052] in, Indicates the first The overlap length between the active power projection range and the active power regulation dead zone boundary value of a distributed power source. This indicates taking the larger value between zero and the value within the parentheses. Indicates taking and The smaller value in Indicates taking and The larger value in the range. The overlap length of the reactive dimension. Using the same calculation method, replace the upper and lower bounds of the active power projection interval in the formula with the upper and lower bounds of the reactive power projection interval, and replace the upper and lower bounds of the active power regulation dead zone with the upper and lower bounds of the reactive power regulation dead zone.

[0053] Distributed power sources with an overlap length greater than zero are marked as valid units capable of participating in primary-distribution coordinated control. The determination criteria are: and If both are established, then the first A distributed power source is marked as a valid unit; if either of the two overlap lengths mentioned above is equal to zero, then the first... Distributed power sources are not marked as valid units.

[0054] The initial feasible region is generated by intersecting the output adjustment ranges of all effective units. The output adjustment range is defined as the adjustable range of active power output and reactive power output of effective units under the condition of satisfying the dead zone constraint. For the ... Each effective generating unit has an adjustable active power output range of [missing information]. The adjustable range of reactive power output is The intersection of the corresponding adjustable output intervals of all effective generating units is calculated dimension by dimension. The lower limit of each dimension in the intersection is the maximum value of the lower limits of the adjustable output intervals of all effective generating units in that dimension, and the upper limit of each dimension in the intersection is the minimum value of the upper limits of the adjustable output intervals of all effective generating units in that dimension. The resulting multi-dimensional intervals constitute the initial feasible region. The number of dimensions of the initial feasible region is equal to twice the number of effective generating units, corresponding to the active power output dimension and reactive power output dimension of each effective generating unit, respectively. If the overlap length between the active power projection interval and the active power regulation dead zone of all distributed generation sources is zero, or the overlap length between the reactive power projection interval and the reactive power regulation dead zone of all distributed generation sources is zero, then the set of effective generating units is an empty set. At this time, the main and distribution coordinated control automatically degrades to an operating mode in which only the main grid side equipment is regulated, that is, the on-load tap changer position of the main grid transformer and the switching capacity of the capacitor bank are used as the only control means, and the charging and discharging power of each energy storage unit and the reduction of flexible load on the distribution network side remain unchanged. The optimization problem in this degenerate mode degenerates into a single-layer optimization problem containing only the first set of decision variables, but it can still run normally within the overall algorithm framework.

[0055] Example 4:

[0056] In practical implementation, the on-load tap changer positions of the main grid transformers and the switching capacity of capacitor banks are coded as the first set of decision variables, while the charging and discharging power of each energy storage unit in the distribution network and the reduction of flexible loads are coded as the second set of decision variables. The on-load tap changer positions of the main grid transformers are represented as integer variables. express, The value range is from the minimum to the maximum tap position of the on-load tap changer of the main grid transformer, with a difference of one between adjacent tap positions. The switching capacity of the capacitor bank is an integer variable. express, The value ranges from zero to the largest integer between the total capacitance of the capacitor bank and the capacitance of a single capacitor bank. The actual value of the switching capacitance of the capacitor bank is... Multiply by the capacity of a single group. The charging and discharging power of each energy storage unit in the distribution network is a continuous variable. express, For the first The charging and discharging power of each energy storage unit is represented by a positive number indicating discharge and a negative number indicating charging. The flexible load reduction is a continuous variable. express, For the first The active power reduction of each flexible load. The first set of decision variables constitutes a vector. The second set of decision variables constitutes a vector. ,in The total number of energy storage units. This represents the total number of flexible loads.

[0057] With the first set of decision variables fixed, the particle swarm optimization algorithm is used to optimize the second set of decision variables. With the second set of decision variables fixed, the discrete gradient descent method is used to optimize the first set of decision variables. These two optimization steps are performed alternately until the difference in the objective function between two iterations is less than a convergence threshold, which is set to [a threshold value]. The convergence threshold is selected as one ten-thousandth of the typical value of the objective function, based on the magnitude of the objective function.

[0058] When optimizing the second set of decision variables using the particle swarm optimization algorithm, the charging / discharging power of each energy storage unit and the reduction amount of each flexible load are concatenated into a particle position vector. The particle swarm is composed of... Composed of individual particles, The setting is 50. The dimension of the position vector of each particle is... , No. The position vector of each particle is denoted as... The value range of each dimension of the particle position vector is set as the projection interval of the initial feasible region on the corresponding dimension. Specifically, the value range of the energy storage unit charging and discharging power dimension is the projection interval of the corresponding energy storage unit active power output dimension in the initial feasible region, and the value range of the flexible load reduction dimension is from zero to the maximum reducible power of the corresponding load node in the initial feasible region.

[0059] The particle swarm optimization algorithm's speed update formula incorporates a penalty term for active power imbalance caused by changes in the main grid transformer tap position. The speed update formula is as follows:

[0060] in, Indicates the first During the nth iteration The first particle velocity vector dimensional components, Indicates the first During the nth iteration The first particle velocity vector Dimensional components. The inertial weight is set to 0.7, which is a typical value that balances global search and local convergence, according to the standard configuration of the particle swarm optimization algorithm. and These are individual learning factors and social learning factors, respectively. The value is 1.5. The value is 1.5. Equal values ​​for both ensure that the individual optimum and the global optimum have the same guiding strength for particle motion. and These are independent random numbers uniformly distributed in the interval [0,1], generated independently in each iteration. For the first The best position in the history of the nth particle Dimensional components. The position of the global optimum Dimensional components. For the first During the nth iteration The position vector of the nth particle Dimensional components. The penalty coefficient is set to 0.05. This value is determined by the ratio of the imbalance caused by changes in different gears during offline simulation to the change in the objective function. The active power imbalance caused by the change in the tap position of the main grid transformer is calculated as follows: based on the tap position of the on-load tap changer of the main grid transformer in the first decision variable group. The change in active power injection on the main grid side caused by the change in the transformer turns ratio is calculated. The absolute value of the difference between this change in active power injection and the sum of the charging and discharging power of each energy storage unit and the sum of the reduction of each flexible load in the second decision variable group is used as the active power imbalance penalty term. The partial derivative term in the above speed update formula... In actual discrete iterative calculations, let the current particle position vector... The set of components corresponding to the energy storage charging and discharging power is as follows The set of components corresponding to the flexible load reduction is as follows and make the total adjustment amount Let the change in active power injection on the main grid side caused by the change in the tap position of the current main grid transformer be . The imbalance penalty term is: For the first The partial derivatives of the dimensional components are calculated using the subgradient formula:

[0061] in, For symbolic functions, The preset unbalanced dead zone threshold is set to a value equal to the power reference value. The penalty term is set to zero when the absolute value of the imbalance is less than or equal to the dead zone threshold. This is to prevent particle swarm oscillations caused by abrupt changes in the sign function near the equilibrium point, ensuring numerical stability and convergence smoothness of the iterative process. Simultaneously, the penalty coefficient... Its dimensions are consistent with those of power. Since it is a dimensionless quantity, the dimensions of all quantities in the velocity update formula are unified to the power dimension.

[0062] The particle fitness is composed of the active power imbalance penalty term and the voltage deviation at the distribution network nodes. The fitness calculation formula is as follows: ,in For the first The fitness of each particle The total number of distribution network nodes. For the first Per-unit value of voltage amplitude at each distribution network node The voltage reference value is set to 1.0 per unit. After calculating the fitness of all particles in each iteration, they are sorted by fitness from smallest to largest, and the top quartile of particles are retained as elite particles. The number of elite particles is... The position of the elite particle is subjected to Gaussian mutation. The mutation method is to add a Gaussian random perturbation with a mean of 0 and a standard deviation of 0.01 to each component of the elite particle's position vector. If the position component exceeds the value range of the corresponding dimension after perturbation, it is restricted to the boundary value.

[0063] When optimizing the first set of decision variables using the discrete gradient descent method, the on-load tap changer positions of the main grid transformer are... Discretize the capacitor bank into integer variables with an adjacent range difference of one, and then adjust the switching capacity of the capacitor bank accordingly. Discretize the data into a single set of integer variables that are integer multiples of the capacity. Calculate the sensitivity of the optimization objective function to unit changes in each integer variable. The sensitivity calculation method is as follows: [The text abruptly ends here, likely due to an incomplete sentence or missing information.] Calculate the change in the objective function by increasing or decreasing the tap position by one level, and take the average of the two changes as the unit change sensitivity of the on-load tap changer position of the main grid transformer. ; respectively switch the capacitor bank's capacity Adding and removing one group of capacitors, calculate the change in the objective function, and take the average of the two changes as the sensitivity to unit changes in the switching capacity of the capacitor bank. The objective function is consistent with the fitness function in the particle swarm optimization algorithm.

[0064] In order of sensitivity from highest to lowest, and Sort the data. For the integer variable with the highest sensitivity, step one unit in the direction that decreases the objective function. If the on-load tap changer of the main grid transformer is in position... It is the variable with the highest sensitivity, and Then Reduce by one unit; if Then Add one unit. If the capacitor bank's switching capacity... It is the variable with the highest sensitivity, and Then Reduce by one unit; if Then Increase by one unit. After stepping, perform distribution network power flow calculations to check if the voltage at each node of the distribution network exceeds its limit. The voltage limit exceeding the limit is determined when the voltage amplitude at any distribution network node is lower than the lower voltage limit or higher than the upper voltage limit. If the voltage at each node of the distribution network does not exceed its limit after stepping, the step is retained and the corresponding integer variable is updated to the value after the step; if the voltage at any distribution network node exceeds its limit, the state before stepping is reversed, and the integer variable remains unchanged. After completing the trial for the variable with the highest sensitivity, the same trial process is performed for the variable with the second highest sensitivity. After one round of trial is completed, the unit change sensitivity of the two integer variables is recalculated, and the next round of trial begins. When neither of the two integer variables has stepped in for two consecutive rounds of trial, the discrete gradient descent method terminates.

[0065] Example 5:

[0066] In practical implementation, the on-load tap changer positions and capacitor bank switching capacities obtained through joint iterative optimization of the main grid transformer are converted into a remote control command sequence executable by the SCADA system. For the on-load tap changer positions of the main grid transformer, the difference between the final tap value and the current tap value is used to obtain the tap change amount. An up-tap or down-tap command is generated based on the sign and absolute value of the tap change amount. Each command includes the controlled object identifier, operation direction, and number of operation steps, where the number of operation steps is the absolute value of the tap change amount. For the capacitor bank switching capacity, the difference between the final switching capacity and the current switching capacity is used to obtain the capacity change amount. An activation or deactivation command is generated based on the sign and absolute value of the capacity change amount. Each command includes the capacitor bank number, activation or deactivation status, and the number of operation groups, which is the integer obtained by dividing the capacity change amount by the capacity of a single group. The above commands are arranged in the order of tap regulation followed by reactive power compensation, forming a remote control command sequence. Each command in the remote control command sequence includes the operation time and the address of the actuator.

[0067] The charging and discharging power of each energy storage unit in the distribution network is decomposed into the form of active power target value plus dead zone hysteresis. An upward adjustment dead zone bandwidth is set for each energy storage unit. and reduce dead zone bandwidth ,in This refers to the energy storage unit serial number. Increase the dead-time bandwidth. The value is set to three percent of the rated power of the energy storage unit, and the dead-zone bandwidth is reduced. The value is taken as three percent of the rated power of the energy storage unit. This percentage is determined based on engineering experience regarding the power control accuracy of the energy storage converter and avoiding frequent charge-discharge switching. (Based on the active power target value) A hysteresis interval is constructed with the center as the upper boundary, the upper dead zone bandwidth as the upper boundary, and the lower dead zone bandwidth as the lower boundary. That is, the upper boundary of the hysteresis interval is... The lower boundary of the hysteresis interval is ,in The first obtained by joint iterative optimization The optimal charging and discharging power of each energy storage unit. The generation of energy storage unit charging and discharging commands follows the formula below:

[0068] in, Indicates the first The energy storage unit in the first The charging and discharging command status of each control cycle. Indicates the first The energy storage unit in the first The charging and discharging command status of each control cycle. Indicates the first The energy storage unit in the first The current actual active power of the energy storage unit in each control cycle, with positive values ​​indicating discharge to the grid and negative values ​​indicating charging from the grid. Below the lower boundary of the hysteresis interval When the energy storage unit is in charging mode, a charging command is sent to control it to switch to charging mode; when the actual power of the energy storage unit is... Above the upper boundary of the hysteresis interval When the energy storage unit is in a hysteresis state, a discharge command is sent to control it to switch to discharge mode; when the actual power of the energy storage unit is within the hysteresis range, the charge / discharge command status remains unchanged from the previous control cycle. The above description is unchanged. To avoid frequent switching between charging and discharging states by the energy storage unit due to measurement noise or small power fluctuations, this application constructs a hysteresis interval centered on the active power target value. The offset of the upper boundary of this hysteresis interval relative to the active power target value is defined as the upward dead zone bandwidth, and the offset of the lower boundary relative to the active power target value is defined as the downward dead zone bandwidth. The upward and downward dead zone bandwidths are not dead zones that do not respond to input signals in the traditional sense, but rather boundary offset parameters used to limit the width of the hysteresis interval. When the actual power of the energy storage unit is within this hysteresis interval, the charging and discharging command remains unchanged from the previous moment. At this time, power fluctuations within this hysteresis interval do not trigger new command outputs. Functionally, it is equivalent to an insensitive area at the command execution level. Therefore, this application describes the entire control band constructed based on this hysteresis interval as the dead zone hysteresis band. Furthermore, since the upper and lower boundary offsets of the hysteresis interval can be independently set according to the power control accuracy, response characteristics and on-site operation requirements of the energy storage unit, they are named the upper dead zone bandwidth and the lower dead zone bandwidth, respectively. Their specific values ​​are determined based on the power control accuracy of the energy storage converter and engineering experience to avoid frequent charge and discharge switching.

[0069] The flexible load reduction is broken down into a sequence of load levels to be cut off in each round. Flexible loads are sorted from lowest to highest priority, and the total flexible load reduction obtained through joint iterative optimization is allocated to each priority level. Starting with the lowest priority flexible load, the interruptible capacity of each load level is accumulated until the accumulated value is greater than or equal to the total flexible load reduction. The load levels involved are then defined as the cut-off level sequence. Each level in the cut-off level sequence corresponds to a load switch number and a cut-off command, with the command specifying the cut-off action type as remote control tripping.

[0070] In practice, the remote control command sequence, the active power target value plus energy storage charging and discharging parameters in the form of dead-zone hysteresis, and the load level sequence cut off in each round are packaged into a transmission frame for different execution terminals. The transmission frame for the on-load tap changer and capacitor bank execution terminals of the main grid transformer contains the remote control command sequence, and the frame structure consists of a start symbol, terminal address, number of commands, operation code and parameters for each command, and check code. The transmission frame for the energy storage unit controller contains the active power target value and the values ​​of upward and downward dead-zone bandwidth, and the frame structure consists of a start symbol, energy storage unit address, active power target value, upward and downward dead-zone bandwidth, and check code. The transmission frame for the flexible load control terminal contains the load level sequence, and the frame structure consists of a start symbol, terminal address, number of cut-off levels, load switch numbers and tripping commands for each level, and check code. All of the above transmission frames are transmitted to the corresponding execution terminals through the distribution network communication network.

[0071] See Figure 6 In the figure, the horizontal axis represents the control cycle, and the vertical axis represents the active power (in kW). The solid curve in the figure represents the actual active power change of the energy storage unit, the dashed line represents the active power target value, and the dotted line and the dotted line represent the upper boundary of the upper dead zone bandwidth and the lower boundary of the lower dead zone bandwidth, respectively. According to Example 5, the charging and discharging power of the energy storage unit is decomposed into the form of the active power target value plus the dead zone hysteresis band. The upper boundary of the upper dead zone and the lower boundary of the lower dead zone in the figure correspond to the upper and lower limits of the hysteresis interval.

[0072] As shown in the figure, the actual active power curve exhibits distinct energy storage (negative power) and discharge (positive power) processes at different stages throughout the entire control cycle, with the actual active power generally fluctuating within the hysteresis range. Specifically, when the active power target value is relatively high (positive), the actual active power gradually increases and approaches the target value, remaining below the upper boundary of the upward adjustment dead zone for most of the time, avoiding frequent switching between charging and discharging states. When the active power target value suddenly drops to a negative value, the actual active power decreases rapidly and remains above the lower boundary of the downward adjustment dead zone, similarly demonstrating the stabilizing effect of the hysteresis bandwidth.

[0073] Furthermore, there is a certain lag and buffer between the actual active power curve and the target active power value, indicating that the energy storage unit's charging and discharging commands are switched based on the comparison between the current actual power and the hysteresis interval boundary, avoiding frequent switching caused by small fluctuations in the target value. This characteristic is consistent with the generation logic of the energy storage unit's charging and discharging commands in Example 5, that is, the charging and discharging state only changes when the actual power exceeds the hysteresis interval, maintaining the stability of the energy storage system and the continuity of control.

[0074] In summary, the curves in the figure clearly reflect the effective regulation of the energy storage unit's charging and discharging behavior by the control strategy of the active power target value plus dead zone hysteresis in Example 5. This demonstrates the technical effect of the method in suppressing frequent charging and discharging switching and stabilizing energy storage power output in actual operation.

[0075] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for optimizing the operation of a main-distribution coordinated distribution network based on data analysis, characterized in that: include: Construct the power exchange feature vector at the boundary of the main distribution network; Based on the correlation between the power exchange feature vector at the boundary of the main distribution network and the voltage over-limit risk of each node in the distribution network, a set of boundary coupling nodes that dominate the coordinated operation of the main and distribution networks is selected. A virtual power injection equivalent device is set at the set of boundary coupling nodes to quantify the support capacity of the main network to the distribution network into the adjustable power range of the boundary nodes. The intersection operation of the adjustable power range and the adjustment dead zone of each distributed power source in the distribution network is performed to generate the initial feasible domain for coordinated control of the main and distribution networks. With the initial feasible domain as a constraint, the on-load tap changer position of the main network transformer, the switching capacity of the capacitor bank, the charging and discharging power of each energy storage unit in the distribution network, and the reduction of flexible load are jointly iteratively optimized to output the coordinated operation control strategy of the main and distribution networks.

2. The method for optimizing the operation of a main distribution network based on data analysis according to claim 1, characterized in that, The construction of the power exchange feature vector at the boundary of the main distribution network specifically includes: Collect bus voltages on each side of the main grid transformer, active and reactive power of tie lines, and output of distributed power sources, energy storage charge status, and interruptible capacity of flexible loads at each node of the distribution network. The active power, reactive power, and voltage amplitude of the low-voltage side bus of the main grid transformer are aligned by timestamp and then spliced ​​into an original feature sequence. The difference value and sliding standard deviation of the sequence within three consecutive scheduling cycles are then calculated. The difference value and sliding standard deviation are appended to the end of the original feature sequence to form the power exchange feature vector of the main distribution network boundary.

3. The method for optimizing the operation of a main distribution network based on data analysis according to claim 1, characterized in that, The selection of the set of boundary coupling nodes that dominate the primary and secondary cooperative operation specifically includes: Calculate the mutual information value between each element in the power exchange feature vector of the main distribution network boundary and the voltage over-limit risk index of each node in the distribution network. Mark the distribution network boundary node corresponding to the element whose mutual information value exceeds the preset threshold as a coupled node. Then, remove redundant nodes that are simultaneously covered by at least two distributed power sources in the same main transformer power supply area from the coupled nodes to obtain the set of boundary coupled nodes.

4. The method for optimizing the operation of a main distribution network based on data analysis according to claim 1, characterized in that, A virtual power injection equivalent device is set at the set of boundary coupling nodes, specifically including: For each node in the boundary coupling node set, bind a virtual generator and a virtual load respectively. Set the upper limit of the active power output of the virtual generator to be equal to the maximum active power that the main grid tie line can transmit at the node, and set the upper limit of the active power absorption of the virtual load to be equal to the maximum load demand of the downstream distribution network of the node. Then lock the output or absorption status of the virtual generator and the virtual load mutually exclusively.

5. The method for optimizing the operation of a main distribution network based on data analysis according to claim 1, characterized in that, The intersection calculation of the adjustable power range and the adjustment dead zones of each distributed power source within the distribution network specifically includes: Read the active and reactive power regulation dead zone boundary values ​​of each distributed power source from the inverter control parameters of each distributed power source in the distribution network, project the adjustable power range onto the output coordinate system of each distributed power source, calculate the overlap length between the projected range and the regulation dead zone boundary value, mark the distributed power sources with an overlap length greater than zero as effective units that can participate in the main and distribution coordinated control, and then perform an intersection operation on the output regulation range of all effective units to generate the initial feasible region.

6. The method for optimizing the operation of a main distribution network based on data analysis according to claim 1, characterized in that, The joint iterative optimization specifically includes: The on-load tap changer position of the main grid transformer and the switching capacity of the capacitor bank are encoded as the first decision variable group. The charging and discharging power of each energy storage unit in the distribution network and the reduction of flexible load are encoded as the second decision variable group. Under the condition that the first decision variable group is fixed, the particle swarm optimization algorithm is used to optimize the second decision variable group. Then, under the condition that the second decision variable group is fixed, the discrete gradient descent method is used to optimize the first decision variable group. The above two optimization steps are alternately executed until the difference of the objective function between the two iterations is less than the convergence threshold.

7. The method for optimizing the operation of a main distribution network based on data analysis according to claim 6, characterized in that, The optimization of the second set of decision variables using the particle swarm optimization algorithm specifically includes: The charging and discharging power of each energy storage unit and the reduction amount of each flexible load are concatenated into a particle position vector. The value range of each dimension of the particle position vector is set as the projection interval of the initial feasible domain on the corresponding dimension. An active power imbalance penalty term caused by the change of the main grid transformer tap is introduced into the speed update formula. The sum of the penalty term and the voltage deviation of the distribution network node is used as the particle fitness. The top quartile of particles are retained as elite particles according to their fitness and their positions are subjected to Gaussian mutation.

8. The method for optimizing the operation of a main distribution network based on data analysis according to claim 6, characterized in that, The optimization of the first set of decision variables using the discrete gradient descent method specifically includes: Discretize the on-load tap changer positions of the transformer into integer variables with a difference of one between adjacent positions, and discretize the switching capacity of the capacitor bank into integer variables that are integer multiples of the capacity of a single bank. Calculate the sensitivity of the objective function to the unit change of each integer variable, and try each integer variable in descending order of sensitivity, stepping one unit in the direction of decreasing the objective function. If the voltage of each node in the distribution network does not exceed the limit after the trial, the step is retained; otherwise, it is reverted to the state before the step.

9. The method for optimizing the operation of a main distribution network based on data analysis according to claim 1, characterized in that, The output master-slave collaborative operation control strategy specifically includes: The on-load tap changer positions of the main grid transformer and the switching capacity of the capacitor bank obtained by joint iterative optimization are converted into a remote control command sequence that can be executed by the SCADA system. The charging and discharging power of each energy storage unit in the distribution network is decomposed into the form of active power target value plus dead zone hysteresis band. The flexible load reduction amount is decomposed into a load level sequence cut off in each round. Then, the remote control command sequence, active power target value plus dead zone hysteresis band, and load level sequence are packaged into a sending frame for different execution terminals.

10. The method for optimizing the operation of a main distribution network based on data analysis according to claim 9, characterized in that, The method of decomposing the charging and discharging power of each energy storage unit in the distribution network into the form of active power target value plus dead zone hysteresis band specifically includes: An upper dead zone bandwidth and a lower dead zone bandwidth are set for each energy storage unit. A hysteresis interval is constructed with the active power target value as the center, the upper boundary of the upper dead zone bandwidth as the upper boundary, and the lower dead zone bandwidth as the lower boundary. When the current actual power of the energy storage unit is lower than the lower boundary of the hysteresis interval, a charging command is sent. When the current actual power of the energy storage unit is higher than the upper boundary of the hysteresis interval, a discharging command is sent. When the current actual power of the energy storage unit is within the hysteresis interval, the charging and discharging state of the previous moment remains unchanged.

Images