A power distribution area micro-grid operation optimization system

Abstract

This invention discloses a distribution substation microgrid operation optimization system, relating to the field of power grid optimization technology. It includes a topology identification module, a phase sequence optimization module, a margin assessment module, a power flow suppression module, a reactive power coordination module, and a loss verification module, all connected to the phase sequence optimization module. The phase sequence optimization module calculates the additional three-phase unbalanced losses based on the phase sequence adjustment scheme and sends a constraint correction signal to the phase sequence optimization module when the additional three-phase unbalanced losses increase. The thermal dynamic margin assessment module calculates the absorption space based on the dynamic curve of transformer winding hot spot temperature, considering the transformer's thermal inertia effect. As long as the hot spot temperature does not exceed the insulation heat resistance temperature, a certain absorption space is retained, avoiding conservative curtailment caused by static load rate assessment.

Description

Technical Field

[0001] This invention belongs to the field of power grid optimization technology, specifically a distribution area microgrid operation optimization system. Background Technology

[0002] With the large-scale integration of distributed photovoltaic power generation in distribution substations, microgrids in distribution substations are facing multiple technical challenges, including increased three-phase imbalance, limited transformer capacity, reverse power flow exceeding limits, node voltage deviation, and increased low-voltage line losses.

[0003] In existing technologies, the main methods for addressing three-phase imbalance in distribution transformer areas include three-phase imbalance compensation devices and intelligent phase-switching switches. Three-phase imbalance compensation devices suppress imbalance by injecting compensation current into the grid using power electronic devices, but they suffer from high costs and limited compensation capacity. Intelligent phase-switching switches reduce three-phase imbalance by adjusting the phase sequence of user loads, but existing switching schemes only optimize for minimizing three-phase imbalance, without considering the impact of phase sequence adjustment on the distribution of transformer phase-to-phase capacity margins, thus failing to simultaneously reduce imbalance and increase photovoltaic power consumption capacity.

[0004] For transformer capacity assessment, existing technologies typically use the static load factor, which is the ratio of current load power to rated capacity. However, transformers have thermal inertia, and short-term overloads do not immediately lead to insulation damage. The static load factor assessment method is too conservative and can easily lead to unnecessary curtailment of photovoltaic power in high photovoltaic penetration scenarios.

[0005] To address reverse power flow suppression, existing technologies mostly employ a single approach of energy storage system absorption or photovoltaic power limiting, without considering a tiered coordination strategy when energy storage capacity is insufficient, nor establishing a causal relationship between phase sequence reconfiguration and reverse power flow suppression. Summary of the Invention

[0006] In summary, the existing technology has the following shortcomings: there is a lack of technical correlation between phase sequence reconfiguration and transformer capacity margin; the static method used for transformer capacity assessment is too conservative; the reverse power flow suppression lacks a hierarchical coordination mechanism; and the independent operation of each functional module has not formed a closed-loop optimization.

[0007] To solve the above problems, the present invention adopts the following technical solution:

[0008] Specifically, a distribution substation microgrid operation optimization system is proposed, including:

[0009] The topology identification module is used to generate a topology phase sequence correlation matrix based on the voltage amplitude and power injection data of each node on the low-voltage side of the distribution substation.

[0010] The phase sequence optimization module is connected to the topology identification module and is used to receive the topology phase sequence correlation matrix and solve the phase sequence adjustment scheme with the goal of reducing the three-phase imbalance and equalizing the transformer phase capacity margin.

[0011] The margin assessment module, connected to the phase sequence optimization module, is used to calculate the dynamic absorption space based on the load distribution of each phase and the transformer thermal circuit model corresponding to the phase sequence adjustment scheme.

[0012] The power flow suppression module, connected to the margin assessment module, is used to generate graded suppression instructions in the order of priority absorption by energy storage, time-series adjustment of transferable load, and smooth derating of photovoltaic output when the photovoltaic output exceeds the dynamic absorption space.

[0013] The reactive power coordination module, connected to the phase sequence optimization module, is used to calculate the reactive power adjustment margin between phases and allocate reactive power compensation based on the load balance of each phase corresponding to the phase sequence adjustment scheme.

[0014] The loss verification module is connected to the phase sequence optimization module and is used to calculate the additional loss of three-phase imbalance according to the phase sequence adjustment scheme. When the additional loss of three-phase imbalance increases, it sends a constraint correction signal to the phase sequence optimization module.

[0015] Furthermore, the topology identification module includes:

[0016] The time-scale alignment unit is used to align the voltage amplitude sampling data and power injection sampling data of each node according to the same absolute time scale, eliminate abnormal sampling points whose time scale deviation exceeds the clock synchronization accuracy of the sampling device, and form the voltage amplitude sequence and power injection sequence within the same time scale window;

[0017] The correlation coefficient calculation unit, connected to the time scale alignment unit, is used to calculate the Pearson correlation coefficient between the power injection change of the j-th node and the voltage amplitude change of the i-th node, and uses the Pearson correlation coefficient as the time-varying correlation coefficient.

[0018] The matrix generation unit, connected to the correlation coefficient calculation unit, is used to identify the phase sequence of each node based on the matrix characteristics of the time-varying correlation coefficient. When the time-varying correlation coefficient between the power injection change of the j-th node and the voltage amplitude change of phase A of the i-th node is greater than the time-varying correlation coefficient between the voltage amplitude changes of phase B and phase C, the j-th node is determined to be a phase A load node.

[0019] Furthermore, the phase sequence optimization module includes a load clustering unit, which is connected to the matrix generation unit;

[0020] The load clustering unit is used to classify load nodes into constant base loads, random fluctuating loads, and photovoltaic coupled loads based on the power curve characteristics of each node extracted from the topological phase sequence correlation matrix and the K-means clustering algorithm based on Euclidean distance.

[0021] The constant base load class includes load nodes whose daily power curve peak-to-mean ratio is less than a first ratio threshold and whose power volatility is less than a first volatility threshold.

[0022] The random fluctuation load class includes load nodes whose daily power curve peak-to-mean ratio is greater than the first ratio threshold and whose power fluctuation rate is greater than the first fluctuation threshold.

[0023] The photovoltaic coupled load class includes load nodes whose daily power curve and the photovoltaic power generation curve of the distribution area have a Pearson correlation coefficient greater than the correlation coefficient threshold.

[0024] When a load node simultaneously meets the classification criteria for both the random fluctuation load category and the photovoltaic coupled load category, it is preferentially classified into the photovoltaic coupled load category.

[0025] Furthermore, the phase sequence optimization module also includes a model building unit and a solution unit, wherein the model building unit is connected to the load clustering unit;

[0026] The model building unit is used to fix the phase sequence label of the constant base load class to the current phase sequence, and use the phase sequence labels of the random fluctuation load class and the photovoltaic coupled load class as 0-1 integer decision variables to build a phase sequence optimization model with the optimization objective of minimizing the weighted sum of the reduction of three-phase current imbalance and the equalization of inter-phase capacity margin.

[0027] The weighting coefficient for phase-to-phase capacity margin equalization is dynamically set using a piecewise linear function based on the current load rate of the transformer. When the load rate is less than 0.5, the weighting coefficient is 0.3. When the load rate is between 0.5 and 0.8, the weighting coefficient is 0.3 plus 0.7 multiplied by the difference between the load rate and 0.5 and then divided by 0.3. When the load rate is greater than 0.8, the weighting coefficient is 1.0 plus 0.67 multiplied by the difference between the load rate and 0.8 and then divided by 0.2.

[0028] The solution unit is connected to the model building unit and is used to solve the phase sequence optimization model using the branch and bound method to obtain the optimal values ​​of each integer decision variable, and generate the phase sequence adjustment scheme for each load node based on the optimal values.

[0029] Furthermore, the margin assessment module includes:

[0030] The load prediction unit is used to allocate each load node to the corresponding phase sequence according to the phase sequence adjustment scheme, and then calculate the load power prediction curves of phase A, phase B and phase C respectively.

[0031] The temperature calculation unit, connected to the load prediction unit, is used to calculate the hourly change value of the hot spot temperature of each phase winding in the next 24 hours based on the load power prediction curve of each phase and the thermal time constant of the transformer winding thermal circuit model, using the recursive thermal balance equation. The recursive thermal balance equation is the exact analytical solution of the first-order differential equation of the transformer thermal circuit model.

[0032] The absorption extraction unit, connected to the temperature calculation unit, is used to compare the maximum value of the hourly change value of the hot spot temperature of each phase winding with the transformer insulation heat resistance temperature. The difference between the insulation heat resistance temperature and the maximum value is the temperature margin. The increase in load power of each phase is determined according to the temperature margin and the conversion factor. The minimum increase in load power of each phase is the dynamic absorption space.

[0033] Furthermore, the power flow suppression module includes:

[0034] An overflow calculation unit is used to calculate the difference between the reverse active power value on the high-voltage side of the distribution transformer and the dynamic absorption space, wherein the difference is the power overflow amount.

[0035] A strategy matching unit, connected to the overflow calculation unit, is used to compare the power overflow amount with the current available charging power of the energy storage system. When the power overflow amount is less than or equal to the current available charging power, a first-level strategy is matched. When the power overflow amount is greater than the current available charging power, the excess portion is compared with the transferable capacity of the transferable load. When the excess portion is less than or equal to the transferable capacity, a second-level strategy is matched; otherwise, a third-level strategy is matched.

[0036] The instruction synthesis unit, connected to the strategy matching unit, is used to synthesize a suppression instruction that only increases the charging power of the energy storage system under the first-level strategy, to synthesize a suppression instruction that increases the charging power of the energy storage system and adjusts the power consumption timing of the transferable load under the second-level strategy, and to synthesize a suppression instruction that increases the charging power of the energy storage system, adjusts the power consumption timing of the transferable load, and limits the photovoltaic output value according to the set power reduction rate under the third-level strategy.

[0037] Furthermore, the reactive power coordination module includes:

[0038] The margin calculation unit is used to calculate the reactive power regulation margin of each phase photovoltaic power generation unit and the reactive power regulation margin of each phase energy storage system based on the load balance degree of each phase corresponding to the phase sequence adjustment scheme.

[0039] The deviation grading unit, connected to the margin calculation unit, is used to classify nodes into voltage over-limit nodes and voltage qualified nodes according to the ratio of the voltage deviation value of each node in the distribution area to the allowable deviation range, and only performs reactive power compensation allocation for voltage over-limit nodes.

[0040] The output allocation unit, connected to the deviation classification unit, is used to allocate the reactive power deficit of the voltage over-limit node into the reactive power output adjustment amount of the photovoltaic power generation unit and the reactive power compensation amount of the energy storage system in the order of priority consumption of the reactive power adjustment margin of the phase where the voltage over-limit node is located, supplementary consumption of the reactive power adjustment margin of the same phase energy storage, and cross-phase support of the reactive power adjustment margin of the adjacent phase.

[0041] Furthermore, the loss verification module includes:

[0042] The unbalance calculation unit is used to calculate the A-phase current, B-phase current and C-phase current of each branch according to the phase sequence assignment of each load node and the load power value of each node in the phase sequence adjustment scheme, and to calculate the negative sequence current component and zero sequence current component of each branch according to the A-phase current, the B-phase current and the C-phase current.

[0043] The loss comparison unit is connected to the unbalance calculation unit and is used to calculate the three-phase unbalance additional loss of each branch after phase sequence reconstruction based on the negative sequence current component and the zero sequence current component, and to compare the three-phase unbalance additional loss with the three-phase unbalance additional loss of the corresponding branch before reconstruction at the branch level.

[0044] The weight correction unit, connected to the loss comparison unit, is used to generate a constraint correction signal when the additional loss of the three-phase imbalance after phase sequence reconstruction of a certain branch is greater than the additional loss of the three-phase imbalance before reconstruction. The constraint correction signal is the result of multiplying the imbalance penalty term coefficient of the branch in the phase sequence optimization model by a correction coefficient. The correction coefficient is determined according to the increase ratio of the additional loss of the branch and has a value range of 1.0 to 2.0. When the correction coefficient is greater than 2.0, the phase sequence labels of all adjustable load nodes on the branch are forcibly fixed to the current phase sequence.

[0045] Furthermore, the phase sequence optimization module generates an initial phase sequence adjustment scheme on a day-ahead time scale, and the loss verification module verifies the initial phase sequence adjustment scheme on a day-ahead time scale.

[0046] The margin assessment module updates the dynamic absorption capacity based on the deviation between the actual load and the predicted load on an hourly time scale.

[0047] The power flow suppression module updates the graded suppression command on a minute-level time scale based on the deviation between the actual and predicted photovoltaic output values.

[0048] When the deviation rate between the actual load on the hourly time scale and the day-ahead predicted load exceeds a set ratio, the phase sequence optimization module is triggered to regenerate the phase sequence adjustment scheme, thereby realizing a rolling closed loop of day-ahead phase sequence optimization and hourly absorption assessment.

[0049] Furthermore, the duration of the time-scale window in the time-scale alignment unit is dynamically adjusted according to the day-ahead fluctuation rate of the distribution area load;

[0050] When the day-ahead volatility is greater than the high volatility threshold, the duration of the time-scale window is shortened to a first duration.

[0051] When the day-ahead volatility is less than the low volatility threshold, the duration of the time-scaled window is extended to a second duration.

[0052] The day-ahead volatility is the ratio of the standard deviation to the mean of the load power values ​​at each hour within the previous 24 hours.

[0053] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects:

[0054] The thermal dynamic margin assessment module calculates the absorption space based on the dynamic curve of the transformer winding hot spot temperature. It takes into account the thermal inertia effect of the transformer. As long as the hot spot temperature does not exceed the insulation heat resistance temperature, a certain absorption space is still reserved, avoiding the conservative curtailment caused by static load rate assessment.

[0055] The phase-to-phase reactive power margin coordination module utilizes the reactive power adjustment margin released by the load balancing of each phase after phase sequence reconstruction, and allocates reactive power compensation in a three-level order of priority for photovoltaic surplus reactive power, energy storage supplementation, and cross-phase support, thereby reducing the transmission loss of reactive power on low-voltage lines. Attached Figure Description

[0056] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0057] 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.

[0058] like Figure 1 This illustrates a distribution substation microgrid operation optimization system, comprising:

[0059] The topology identification module is used to generate a topology phase sequence correlation matrix based on the voltage amplitude and power injection data of each node on the low-voltage side of the distribution substation.

[0060] The phase sequence optimization module is connected to the topology identification module and is used to receive the topology phase sequence correlation matrix and solve the phase sequence adjustment scheme with the goal of reducing the three-phase imbalance and equalizing the transformer phase capacity margin.

[0061] The margin assessment module, connected to the phase sequence optimization module, is used to calculate the dynamic absorption space based on the load distribution of each phase and the transformer thermal circuit model corresponding to the phase sequence adjustment scheme.

[0062] The power flow suppression module, connected to the margin assessment module, is used to generate graded suppression instructions in the order of priority absorption by energy storage, time-series adjustment of transferable load, and smooth derating of photovoltaic output when the photovoltaic output exceeds the dynamic absorption space.

[0063] The reactive power coordination module, connected to the phase sequence optimization module, is used to calculate the reactive power adjustment margin between phases and allocate reactive power compensation based on the load balance of each phase corresponding to the phase sequence adjustment scheme.

[0064] The loss verification module is connected to the phase sequence optimization module and is used to calculate the additional loss of three-phase imbalance according to the phase sequence adjustment scheme. When the additional loss of three-phase imbalance increases, it sends a constraint correction signal to the phase sequence optimization module.

[0065] It should be noted that this system is deployed on the edge computing nodes of the distribution substation or on the distribution IoT cloud platform, and data transmission between modules uses MQTT message queues or Kafka message buses. The detailed implementation steps, calculation formulas, and technical effects of each module are explained below.

[0066] Step S11: Using standard smart meters deployed at the low-voltage side outgoing terminals of the transformer in the distribution substation and at the ends of each branch line, the three-phase voltage amplitude, three-phase active power, and reactive power of each node are collected at a sampling period of 1 minute. The time stamp accuracy of the collected data is not less than 1 millisecond, and the local clock of each smart meter is synchronized and calibrated using the IEEE 1588 precision time protocol or GPS second pulse signal.

[0067] Step S12: Align the voltage amplitude sampling data and power injection sampling data of each node with the same absolute time scale, remove abnormal sampling points whose time scale deviation exceeds the clock synchronization accuracy of the sampling device, and form a voltage amplitude sequence and power injection sequence within the same time scale window.

[0068] Step S13: Calculate the time-varying correlation coefficient. For each pair of nodes (i,j), calculate the Pearson correlation coefficient between the change in power injection and the change in voltage amplitude, and use the Pearson correlation coefficient as the time-varying correlation coefficient.

[0069] Step S14: Generate a topological phase sequence correlation matrix based on the time-varying correlation coefficient. The topological phase sequence correlation matrix reflects the changing pattern of load type of each node over time.

[0070] Step S15: Identify the phase sequence assignment of each node based on matrix characteristics. When the time-varying correlation coefficient between the power injection change of node j and the A-phase voltage amplitude change of node i is greater than the time-varying correlation coefficients with phases B and C, and is greater than the threshold, node j is determined to be an A-phase load node. The threshold is determined based on the resistance-reactance ratio of the low-voltage lines in the distribution substation area, and its value ranges from 0.65 to 0.85.

[0071] Step S21: Receive the topological phase sequence correlation matrix and extract the power curve features of each node, including daily peak power, daily valley power, power fluctuation rate and power factor.

[0072] Step S22: The load nodes are classified into three categories using a K-means clustering algorithm based on Euclidean distance: constant base load, random fluctuation load, and photovoltaic coupled load. The number of clusters K=3, and the Euclidean distance is calculated based on the daily average power, power standard deviation, and mean power factor. The constant base load category includes loads such as lighting and refrigerators that operate continuously for 24 hours with small power fluctuations; the random fluctuation load category includes loads such as air conditioners and electric water heaters that start and stop randomly and have large power fluctuations; and the photovoltaic coupled load category includes loads such as electric vehicle charging piles whose power curves are correlated with photovoltaic output. When a load node meets the classification criteria for both the random fluctuation load and photovoltaic coupled load categories, it is preferentially classified into the photovoltaic coupled load category.

[0073] Step S23: Fix the phase sequence label of the constant base load class to the current phase sequence, and use the phase sequence labels of the random fluctuating load class and the photovoltaic coupled load class as 0-1 integer decision variables to construct a phase sequence optimization model with the optimization objective of minimizing the weighted sum of the reduction in three-phase current imbalance and the equalization of inter-phase capacity margin. The weight coefficient of the equalization of inter-phase capacity margin is dynamically set according to the current load rate of the transformer; the higher the current load rate of the transformer, the larger the weight coefficient of the equalization of inter-phase capacity margin.

[0074] Step S24: Solve the above mixed integer programming model using the branch and bound method. The branching strategy branches the integer decision variables in descending order of load power. The bounding strategy uses the upper bound as the objective function value of the current optimal integer solution and the lower bound as the objective function value of the linear relaxation solution. The iteration terminates when the difference between the upper and lower bounds is less than the convergence accuracy.

[0075] Step S25: Generate the phase sequence adjustment scheme for each load node based on the optimal values ​​of the integer decision variables obtained from the solution.

[0076] Step S31: After allocating each load node to the corresponding phase sequence according to the phase sequence adjustment scheme, the load power prediction curves for phases A, B, and C are calculated respectively. The load power prediction uses the support vector regression algorithm, and the input features include historical load power, daily type identifier, and meteorological temperature.

[0077] Step S32: Based on the load power prediction curves of each phase and the thermal time constant of the transformer winding thermal circuit model, the hourly variation of the hot spot temperature of each phase winding over the next 24 hours is calculated using a recursive heat balance equation. The recursive heat balance equation is the exact analytical solution of the first-order differential equation of the transformer thermal circuit model, with the initial condition being the current measured hot spot temperature or an estimate based on the current load rate.

[0078] Step S33: Compare the maximum value of the hourly variation of the hot spot temperature of each phase winding with the transformer insulation heat resistance temperature. The difference between the insulation heat resistance temperature and the maximum value is the temperature margin. The insulation heat resistance temperature is determined according to the insulation material grade: 105 degrees Celsius for Class A insulation, 130 degrees Celsius for Class B insulation, and 155 degrees Celsius for Class F insulation.

[0079] Step S34: Calculate the increase in load power for each phase based on the temperature margin. The minimum increase in load power for each phase is the dynamic absorption capacity. The conversion factor is determined by fitting the load loss test results from the transformer's factory test data.

[0080] Step S41: Real-time acquisition of the three-phase active power on the high-voltage side of the transformer. When the active power on the high-voltage side is negative, it indicates that there is reverse power flow. Calculate the reverse active power value.

[0081] Step S42: Calculate the power overflow amount, which is the difference between the reverse active power value and the dynamic absorption space.

[0082] Step S43: Compare the power overflow with the current available charging power of the energy storage system. If the power overflow is less than or equal to the current available charging power, the first-level strategy is matched to increase the charging power of the energy storage system only. If the power overflow is greater than the current available charging power, the remaining overflow is calculated and compared with the transferable capacity of the transferable load. If the remaining overflow is less than or equal to the transferable capacity, the second-level strategy is matched to increase the charging power of the energy storage system and adjust the power consumption sequence of the transferable load. Otherwise, the third-level strategy is matched to increase the charging power of the energy storage system, adjust the power consumption sequence of the transferable load, and limit the photovoltaic output value at a rate not exceeding 5% of the rated photovoltaic power per minute.

[0083] Step S51: Based on the load balance of each phase corresponding to the phase sequence adjustment scheme, calculate the reactive power regulation margin of each phase photovoltaic power generation unit and the reactive power regulation margin of each phase energy storage system.

[0084] Step S52: Based on the ratio of the voltage deviation value of each node in the distribution transformer area to the allowable deviation range, the nodes are divided into voltage over-limit nodes and voltage qualified nodes. Reactive power compensation is only allocated to the voltage over-limit nodes. The allowable deviation range is determined according to GB / T12325-2008.

[0085] Step S53: According to the order of priority consumption of photovoltaic reactive power regulation margin in the phase where the voltage over-limit node is located, supplementary consumption of reactive power regulation margin of in-phase energy storage, and cross-phase support of reactive power regulation margin of adjacent phases, the reactive power deficit of the voltage over-limit node is allocated as reactive power output regulation amount of photovoltaic power generation unit and reactive power compensation amount of energy storage system.

[0086] Step S61: Calculate the A-phase current, B-phase current and C-phase current of each branch according to the phase sequence assignment of each load node and the load power value of each node in the phase sequence adjustment scheme.

[0087] Step S62: Calculate the negative sequence current component and zero sequence current component of each branch according to the symmetrical component method.

[0088] Step S63: Calculate the three-phase unbalanced additional losses of each branch after phase sequence reconstruction based on the negative sequence current component and the zero sequence current component, and compare the three-phase unbalanced additional losses with the three-phase unbalanced additional losses of the corresponding branch before reconstruction at the branch level.

[0089] Step S64: When the additional three-phase imbalance loss of a branch after phase sequence reconstruction is greater than the additional three-phase imbalance loss before reconstruction, a constraint correction signal is generated. The constraint correction signal is the unbalance penalty term coefficient of the branch in the phase sequence optimization model multiplied by a correction coefficient. The correction coefficient is determined according to the increase ratio of the additional loss of the branch, and its value ranges from 1.0 to 2.0. When the correction coefficient is greater than 2.0, the phase sequence labels of all adjustable load nodes on the branch are forcibly fixed to the current phase sequence.

[0090] Through the coordinated work of the above modules, this system achieves the following technical effects: Phase sequence reconfiguration not only reduces the three-phase imbalance, but also improves the photovoltaic absorption space by balancing the inter-phase capacity margin of the transformer, making the phase sequence reconfiguration and reverse power flow suppression form a non-obvious causal relationship; the dynamic absorption space calculated based on the dynamic curve of winding hot spot temperature reflects the actual capacity margin of the transformer more accurately than the traditional static load rate assessment; the additional loss iterative verification module realizes the closed loop of optimization verification and correction through constraint weight correction, avoiding the problem of increased local branch losses after phase sequence reconfiguration.

[0091] As an optional embodiment, the topology identification module includes:

[0092] The time-scale alignment unit is used to align the voltage amplitude sampling data and power injection sampling data of each node according to the same absolute time scale, eliminate abnormal sampling points whose time scale deviation exceeds the clock synchronization accuracy of the sampling device, and form the voltage amplitude sequence and power injection sequence within the same time scale window;

[0093] The correlation coefficient calculation unit, connected to the time scale alignment unit, is used to calculate the Pearson correlation coefficient between the power injection change of the j-th node and the voltage amplitude change of the i-th node, and uses the Pearson correlation coefficient as the time-varying correlation coefficient.

[0094] The matrix generation unit, connected to the correlation coefficient calculation unit, is used to identify the phase sequence of each node based on the matrix characteristics of the time-varying correlation coefficient. When the time-varying correlation coefficient between the power injection change of the j-th node and the voltage amplitude change of phase A of the i-th node is greater than the time-varying correlation coefficient between the voltage amplitude changes of phase B and phase C, the j-th node is determined to be a phase A load node.

[0095] It should be noted that the implementation steps of the time-stamp alignment unit are as follows: Step T11, each smart meter records the time stamp of the sampled data based on its local clock. The time stamp format is a Unix timestamp with millisecond precision. Step T12, using the system clock of the distribution area master station server as the reference clock, calculate the time-stamp deviation between the local clock of each smart meter and the reference clock. Step T13, for each sampled data point, calculate the corrected time stamp. Step T14, discard sampled points whose time-stamp deviation exceeds the clock synchronization precision of the sampling device. The clock synchronization precision is determined based on the crystal oscillator precision of the smart meter, with a typical value of ±10 milliseconds.

[0096] The formula for calculating the Pearson correlation coefficient is:

[0097]

[0098] in and Let be the sample means of ΔPj and ΔVi, respectively, and sum them over all sampling time pairs within the same time window. The Pearson correlation coefficient ranges from -1 to 1, with an absolute value greater than 0.7 indicating a strong correlation, 0.4 to 0.7 indicating a moderate correlation, and less than or equal to 0.4 indicating a weak correlation.

[0099] The phase sequence assignment logic of the matrix generation unit is as follows: For each load node j, calculate its time-varying correlation coefficients ρ{j,A}, ρ{j,B}, and ρ{j,C} with the A-phase voltage, B-phase voltage, and C-phase voltage, respectively. If ρ{j,A} is greater than ρ{j,B}, ρ{j,A} is greater than ρ{j,C}, and ρ{j,A} is greater than the threshold, then node j is determined to be an A-phase load node; if ρ{j,B} is greater than ρ{j,A}, ρ{j,B} is greater than ρ{j,C}, and ρ{j,B} is greater than the threshold, then it is determined to be a B-phase load node; if ρ{j,C} is greater than ρ{j,A}, ρ{j,C} is greater than ρ{j,B}, and ρ{j,C} is greater than the threshold, then it is determined to be a C-phase load node. If the time-varying correlation coefficients of all three phases are less than the threshold, they are marked as uncertain nodes and will not be adjusted in phase sequence optimization. The threshold is determined based on the resistance-to-reactance ratio R / X of the low-voltage line in the distribution substation. The calculation formula is 0.5 + 0.2 multiplied by arctan(R / X) and then divided by π / 2. When R / X = 0.5, the threshold is 0.63; when R / X = 1.0, the threshold is 0.70; and when R / X = 2.0, the threshold is 0.77.

[0100] Technical benefits: Time-stamp alignment ensures the synchronization of multi-node data, avoiding topology identification errors caused by asynchronous data; Pearson correlation coefficient quantifies the linear correlation between power and voltage between nodes, making it more suitable for distribution substation scenarios with random load fluctuations than the traditional sensitivity matrix method; and independent three-phase determination of phase sequence assignment solves the problem of unclear phase sequence caused by missing historical construction records.

[0101] As an optional embodiment, the phase sequence optimization module includes a load clustering unit, which is connected to the matrix generation unit;

[0102] The load clustering unit is used to classify load nodes into constant base loads, random fluctuating loads, and photovoltaic coupled loads based on the power curve characteristics of each node extracted from the topological phase sequence correlation matrix and the K-means clustering algorithm based on Euclidean distance.

[0103] The constant base load class includes load nodes whose daily power curve peak-to-mean ratio is less than a first ratio threshold and whose power volatility is less than a first volatility threshold.

[0104] The random fluctuation load class includes load nodes whose daily power curve peak-to-mean ratio is greater than the first ratio threshold and whose power fluctuation rate is greater than the first fluctuation threshold.

[0105] The photovoltaic coupled load class includes load nodes whose daily power curve and the photovoltaic power generation curve of the distribution area have a Pearson correlation coefficient greater than the correlation coefficient threshold.

[0106] When a load node simultaneously meets the classification criteria for both the random fluctuation load category and the photovoltaic coupled load category, it is preferentially classified into the photovoltaic coupled load category.

[0107] It should be noted that the steps for extracting power curve features are as follows: extract the 24-hour power curves of each node from the topological phase sequence correlation matrix, and calculate the following features: daily average power is the average power over 24 hours, daily peak power is the maximum power over 24 hours, peak-to-mean ratio is the ratio of daily peak power to daily average power, power fluctuation rate is the ratio of power standard deviation to daily average power, and power factor mean is the average power factor over 24 hours.

[0108] The implementation steps of K-means clustering are as follows: Step C11, determine the number of clusters K=3. Step C12, select initial cluster centers using the K-means++ initialization strategy: first, randomly select a node as the first cluster center, and then for each subsequent cluster center, select a node with a probability proportional to the squared distance to each existing cluster center. Step C13, calculate the Euclidean distance between the power feature vector of each node and the three cluster centers, and assign each node to the cluster with the closest distance. Step C14, update the cluster centers to the mean of the feature vectors of all nodes in the current cluster. Step C15, repeat steps C13 and C14 until the change in the number of cluster centers is less than the convergence threshold.

[0109] The specific values ​​for the classification thresholds are as follows: the first ratio threshold is determined based on the statistical characteristics of the distribution area load, with a typical value of 2.0; the first volatility threshold is determined based on the degree of random load fluctuation, with a typical value of 0.3; and the correlation coefficient threshold is determined based on the photovoltaic coupling strength, with a typical value of 0.5. When a load node simultaneously meets the conditions for both random fluctuation load and photovoltaic coupled load, it is preferentially classified into the photovoltaic coupled load category because the power curve of photovoltaic coupled load has deterministic characteristics related to photovoltaic output, making it easier to model and predict than simple random fluctuations.

[0110] Technical effects: By classifying loads according to their time-varying characteristics through K-means clustering, a differentiated processing strategy is provided for phase sequence optimization: constant base loads are used as anchor nodes to keep the phase sequence fixed, ensuring the stability of phase sequence adjustment; random fluctuating loads and photovoltaic coupled loads are used as adjustable variables, which reduces the solution dimension while retaining the main optimization space; and the ambiguity of load classification is avoided through classification cross-processing and priority inclusion rules.

[0111] As an optional embodiment, the phase sequence optimization module further includes a model building unit and a solution unit, wherein the model building unit is connected to the load clustering unit;

[0112] The model building unit is used to fix the phase sequence label of the constant base load class to the current phase sequence, and use the phase sequence labels of the random fluctuation load class and the photovoltaic coupled load class as 0-1 integer decision variables to build a phase sequence optimization model with the optimization objective of minimizing the weighted sum of the reduction of three-phase current imbalance and the equalization of inter-phase capacity margin.

[0113] The weighting coefficient for phase-to-phase capacity margin equalization is dynamically set using a piecewise linear function based on the current load rate of the transformer. When the load rate is less than 0.5, the weighting coefficient is 0.3. When the load rate is between 0.5 and 0.8, the weighting coefficient is 0.3 plus 0.7 multiplied by the difference between the load rate and 0.5 and then divided by 0.3. When the load rate is greater than 0.8, the weighting coefficient is 1.0 plus 0.67 multiplied by the difference between the load rate and 0.8 and then divided by 0.2.

[0114] The solution unit is connected to the model building unit and is used to solve the phase sequence optimization model using the branch and bound method to obtain the optimal values ​​of each integer decision variable, and generate the phase sequence adjustment scheme for each load node based on the optimal values.

[0115] It should be noted that the mathematical form of the phase sequence optimization model is: the optimization objective is α multiplied by the unbalance reduction amount plus β(μ) multiplied by the standard deviation of the phase capacity margin, where α is the unbalance reduction weight coefficient with a fixed value of 1.0, β(μ) is the phase capacity margin equalization weight coefficient that changes dynamically with the load rate, and μ is the current load rate of the transformer, i.e., the ratio of the apparent power on the high-voltage side to the rated capacity.

[0116] The precise values ​​of the piecewise linear function are: β=0.30 when μ=0.5, β=0.65 when μ=0.65, β=1.00 when μ=0.8, and β=1.67 when μ=1.0. When the transformer load rate is low, the primary objective is to reduce imbalance; when the load rate is high, the primary objective is to enhance the equalization of phase-to-phase capacity margin.

[0117] The constraints include: each adjustable load node must be assigned to one of phases A, B, or C; the total power of each phase does not exceed the rated capacity of the phase winding; and the number of load nodes requiring phase sequence adjustment does not exceed 30% of the total number of adjustable load nodes. This threshold is determined based on the workload of phase switching operations in the distribution substation and the power supply reliability requirements.

[0118] The steps of the branch and bound method are as follows: Select branch variables in descending order of load power, prioritizing the phase sequence assignment of high-power loads; solve the linear relaxation problem for each branch node to obtain the lower bound; prune branches if the lower bound is greater than the objective function value of the current optimal integer solution; terminate the method when all branches have been pruned or an integer solution has been found. The convergence accuracy is set to be less than 0.1% between the optimal integer solution and the optimal lower bound.

[0119] Technical effects: By classifying and differentiating loads, the number of integer decision variables in the phase sequence optimization model is reduced by 40% to 60%; the piecewise linear dynamic weight coefficient enables the optimization target to be adaptively adjusted with the load rate, allowing phase sequence reconfiguration to prioritize the equalization of inter-phase capacity margin in high load rate scenarios, thereby improving the photovoltaic absorption capacity; the 30% phase commutation ratio constraint ensures the feasibility of phase sequence adjustment and the reliability of power supply.

[0120] As an optional embodiment, the margin assessment module includes:

[0121] The load prediction unit is used to allocate each load node to the corresponding phase sequence according to the phase sequence adjustment scheme, and then calculate the load power prediction curves of phase A, phase B and phase C respectively.

[0122] The temperature calculation unit, connected to the load prediction unit, is used to calculate the hourly change value of the hot spot temperature of each phase winding in the next 24 hours based on the load power prediction curve of each phase and the thermal time constant of the transformer winding thermal circuit model, using the recursive thermal balance equation. The recursive thermal balance equation is the exact analytical solution of the first-order differential equation of the transformer thermal circuit model.

[0123] The absorption extraction unit, connected to the temperature calculation unit, is used to compare the maximum value of the hourly change value of the hot spot temperature of each phase winding with the transformer insulation heat resistance temperature. The difference between the insulation heat resistance temperature and the maximum value is the temperature margin. The increase in load power of each phase is determined according to the temperature margin and the conversion factor. The minimum increase in load power of each phase is the dynamic absorption space.

[0124] It should be noted that the implementation steps of the load forecasting unit are as follows: According to the phase sequence adjustment scheme, each load node is assigned to phase A, phase B, or phase C according to the phase sequence label; the load power of each phase is predicted on a given day. The prediction model adopts the support vector regression algorithm, the kernel function is the radial basis function, and the input feature vector includes the load power of the same period of the previous 7 days, the day type identifier, and the predicted day temperature; the output is the load power prediction curve of each phase for the next 24 hours, with a time resolution of 1 hour.

[0125] The detailed form of the recursive heat balance equation is: the current hot spot temperature equals the previous hot spot temperature multiplied by e raised to the power of negative Δt / τ, plus the sum of the ambient temperature and the loss temperature rise multiplied by 1 minus e raised to the power of negative Δt / τ. Here, Δt is the calculation step size, taken as 15 minutes, and τ is the transformer winding thermal time constant, typically 2.5 hours for oil-immersed distribution transformers and 1.0 hour for dry-type distribution transformers. The ambient temperature is determined based on weather forecast data, and the loss temperature rise is calculated based on the load loss and no-load loss in the transformer's factory test report.

[0126] The conversion method for dynamic load absorption capacity is as follows: The temperature margin is the insulation heat resistance temperature minus the maximum predicted hot spot temperature of each phase winding within the next 24 hours. The insulation heat resistance temperature is determined according to the insulation material grade: 105 degrees Celsius for Class A insulation, 130 degrees Celsius for Class B insulation, and 155 degrees Celsius for Class F insulation. The increase in load power for each phase is the temperature margin divided by a conversion factor. The conversion factor is determined by fitting the load loss test results in the transformer's factory test data, with a typical range of 1.2 to 2.0 degrees Celsius per kilowatt. The dynamic load absorption capacity is the minimum value for all three phases.

[0127] Technical benefits: Compared with the traditional static load rate assessment method, this module assesses the absorption space based on the dynamic curve of winding hot spot temperature, taking into account the thermal inertia effect of the transformer. Even if the short-term load rate exceeds 100% during the high photovoltaic output period, as long as the hot spot temperature does not exceed the insulation heat resistance temperature, a certain absorption space is still reserved, thereby avoiding photovoltaic curtailment caused by conservative estimation.

[0128] As an optional embodiment, the power flow suppression module includes:

[0129] An overflow calculation unit is used to calculate the difference between the reverse active power value on the high-voltage side of the distribution transformer and the dynamic absorption space, wherein the difference is the power overflow amount.

[0130] A strategy matching unit, connected to the overflow calculation unit, is used to compare the power overflow amount with the current available charging power of the energy storage system. When the power overflow amount is less than or equal to the current available charging power, a first-level strategy is matched. When the power overflow amount is greater than the current available charging power, the excess portion is compared with the transferable capacity of the transferable load. When the excess portion is less than or equal to the transferable capacity, a second-level strategy is matched; otherwise, a third-level strategy is matched.

[0131] The instruction synthesis unit, connected to the strategy matching unit, is used to synthesize a suppression instruction that only increases the charging power of the energy storage system under the first-level strategy, to synthesize a suppression instruction that increases the charging power of the energy storage system and adjusts the power consumption timing of the transferable load under the second-level strategy, and to synthesize a suppression instruction that increases the charging power of the energy storage system, adjusts the power consumption timing of the transferable load, and limits the photovoltaic output value according to the set power reduction rate under the third-level strategy.

[0132] It should be noted that the implementation steps of the overflow calculation unit are as follows: real-time acquisition of the three-phase active power on the high-voltage side of the transformer; when the active power on the high-voltage side is negative, the reverse active power value is calculated; the dynamic absorption space output by the margin assessment module is read; the power overflow amount is calculated as the reverse active power value minus the dynamic absorption space; when the power overflow amount is less than or equal to zero, it indicates that the reverse power flow has not exceeded the absorption space, and there is no need to generate a suppression command.

[0133] The strategy matching implementation steps are as follows: First, read the current available charging power of the energy storage system, which is the difference between the maximum charging power of the energy storage system and the current charging power. Second, compare the power overflow with the current available charging power. If the power overflow is less than or equal to the current available charging power, match the first-level strategy; otherwise, calculate the remaining overflow. Third, compare the remaining overflow with the transferable capacity of the transferable load. If the remaining overflow is less than or equal to the transferable capacity, match the second-level strategy; otherwise, match the third-level strategy. The transferable capacity is the total power of the transferable load that can be postponed to other time periods in the current time period.

[0134] The specific details of the suppression command synthesis are as follows: Under the first-level strategy, the energy storage system charging power command is the current charging power plus the power overflow. Under the second-level strategy, the energy storage charging power command is the current charging power plus the currently available charging power, and the transferable load timing adjustment command transfers the remaining overflow load power from the current time period to the adjacent low-load time period. Under the third-level strategy, the energy storage charging power command is the current charging power plus the currently available charging power, the transferable load timing adjustment command transfers the load power of the transferable capacity, and the photovoltaic output limiting command limits the photovoltaic output value at a rate not exceeding 5% of the rated photovoltaic power per minute.

[0135] Technical effect: Through the progressive matching of the three-level strategy, the energy storage system's rapid response capability is used first to absorb the overflow power, then the power gap is reduced by the timing adjustment of the transferable load, and finally the photovoltaic output is limited as a backup measure, thus achieving a balance between economy and absorption capacity.

[0136] As an optional embodiment, the reactive power coordination module includes:

[0137] The margin calculation unit is used to calculate the reactive power regulation margin of each phase photovoltaic power generation unit and the reactive power regulation margin of each phase energy storage system based on the load balance degree of each phase corresponding to the phase sequence adjustment scheme.

[0138] The deviation grading unit, connected to the margin calculation unit, is used to classify nodes into voltage over-limit nodes and voltage qualified nodes according to the ratio of the voltage deviation value of each node in the distribution area to the allowable deviation range, and only performs reactive power compensation allocation for voltage over-limit nodes.

[0139] The output allocation unit, connected to the deviation classification unit, is used to allocate the reactive power deficit of the voltage over-limit node into the reactive power output adjustment amount of the photovoltaic power generation unit and the reactive power compensation amount of the energy storage system in the order of priority consumption of the reactive power adjustment margin of the phase where the voltage over-limit node is located, supplementary consumption of the reactive power adjustment margin of the same phase energy storage, and cross-phase support of the reactive power adjustment margin of the adjacent phase.

[0140] It should be noted that the implementation steps of the margin calculation unit are as follows: determine the load distribution and load balance of each phase according to the phase sequence adjustment scheme; calculate the photovoltaic reactive power adjustment margin of each phase as the maximum reactive power capacity of the photovoltaic inverter minus the current reactive power output value, the maximum reactive power capacity of the photovoltaic inverter is determined according to the inverter power factor operating range; calculate the energy storage reactive power adjustment margin of each phase as the maximum reactive power compensation value of the energy storage converter minus the current reactive power output value.

[0141] The implementation steps for deviation grading are as follows: Calculate the voltage deviation rate of each node by dividing the difference between the measured voltage and the nominal voltage by the nominal voltage and then multiplying by 100%; determine the allowable deviation range according to GB / T12325-2008, which is ±7% and ±10% of the nominal voltage for 220V single-phase power supply, and ±7% of the nominal voltage for 380V three-phase power supply; when the absolute value of the voltage deviation rate exceeds the allowable deviation range, it is judged as a voltage over-limit node, otherwise it is a voltage qualified node.

[0142] The priority order for reactive power output allocation is as follows: First priority is the reactive power regulation margin of the photovoltaic system in the phase where the voltage over-limit node is located. When the reactive power margin of the phase is sufficient, the photovoltaic system will fully bear the reactive power compensation. Second priority is the reactive power regulation margin of the same phase energy storage system. When the reactive power margin of the photovoltaic system is insufficient, the energy storage system will supplement it. Third priority is cross-phase support of the reactive power regulation margin of adjacent phases. When the reactive power margin of the current phase is exhausted, the reactive power compensation capacity will be borrowed from the adjacent phases.

[0143] Technical effects: By assessing the load balance of each phase after phase sequence reconfiguration, the reactive power adjustment margin released by load balancing of each phase is accurately calculated; through a three-level priority allocation strategy, the remaining reactive power capacity of the photovoltaic inverter is used first for local reactive power compensation, reducing the transmission loss of reactive power on low-voltage lines; the cross-phase support mechanism realizes the optimized allocation of reactive power resources between phases when the three-phase reactive power margin is unbalanced.

[0144] As an optional embodiment, the loss verification module includes:

[0145] The unbalance calculation unit is used to calculate the A-phase current, B-phase current and C-phase current of each branch according to the phase sequence assignment of each load node and the load power value of each node in the phase sequence adjustment scheme, and to calculate the negative sequence current component and zero sequence current component of each branch according to the A-phase current, the B-phase current and the C-phase current.

[0146] The loss comparison unit is connected to the unbalance calculation unit and is used to calculate the three-phase unbalance additional loss of each branch after phase sequence reconstruction based on the negative sequence current component and the zero sequence current component, and to compare the three-phase unbalance additional loss with the three-phase unbalance additional loss of the corresponding branch before reconstruction at the branch level.

[0147] The weight correction unit, connected to the loss comparison unit, is used to generate a constraint correction signal when the additional loss of the three-phase imbalance after phase sequence reconstruction of a certain branch is greater than the additional loss of the three-phase imbalance before reconstruction. The constraint correction signal is the result of multiplying the imbalance penalty term coefficient of the branch in the phase sequence optimization model by a correction coefficient. The correction coefficient is determined according to the increase ratio of the additional loss of the branch and has a value range of 1.0 to 2.0. When the correction coefficient is greater than 2.0, the phase sequence labels of all adjustable load nodes on the branch are forcibly fixed to the current phase sequence.

[0148] It should be noted that the implementation steps of the unbalance calculation unit are as follows: determine the phase sequence assignment of each load node and the corresponding load power value of the phase sequence according to the phase sequence adjustment scheme; for each branch, calculate the A-phase current of the branch as the sum of the A-phase load power flowing through the branch divided by the A-phase voltage, and the B-phase and C-phase are calculated in the same way; calculate the negative sequence current according to the symmetrical component method as the sum of the A-phase current plus a squared multiplied by the B-phase current plus a multiplied by the C-phase current divided by 3, and the zero sequence current as the sum of the three-phase current divided by 3, where a is e raised to the power of j120 degrees.

[0149] The formula for calculating the additional losses due to three-phase imbalance is as follows: After phase sequence reconstruction, the additional losses due to three-phase imbalance in each branch are calculated as 3 times the square of the zero-sequence current multiplied by the branch resistance plus 3 times the square of the negative-sequence current multiplied by the branch resistance. The branch resistance is obtained from the distribution network GIS system based on the line type and length.

[0150] The implementation steps of constraint weight correction are as follows: For each branch, compare the additional loss of three-phase imbalance after phase sequence reconstruction with the additional loss of three-phase imbalance before reconstruction; if the loss after reconstruction is greater than the loss before reconstruction, calculate the correction coefficient as 1 and add the difference between the loss after reconstruction and the loss before reconstruction divided by the loss before reconstruction; generate a constraint correction signal and multiply the imbalance penalty term coefficient of the branch in the phase sequence optimization model by the correction coefficient; if the correction coefficient is greater than 2.0, force the phase sequence label of all adjustable load nodes on the branch to be fixed to the current phase sequence.

[0151] Technical effects: By verifying the unbalanced additional loss at the branch level, the problem of increased local branch loss that may occur under the global optimal solution of phase sequence optimization is avoided; by dynamically correcting the constraint weight coefficients, the loss verification results are fed back to the phase sequence optimization model, forming an iterative closed loop of optimization verification and correction; by forcibly fixing the phase sequence label of high-loss branches, the executability and security of phase sequence reconstruction are guaranteed.

[0152] As an optional embodiment, the phase sequence optimization module generates an initial phase sequence adjustment scheme on a day-ahead time scale, and the loss verification module verifies the initial phase sequence adjustment scheme on a day-ahead time scale.

[0153] The margin assessment module updates the dynamic absorption capacity based on the deviation between the actual load and the predicted load on an hourly time scale.

[0154] The power flow suppression module updates the graded suppression command on a minute-level time scale based on the deviation between the actual and predicted photovoltaic output values.

[0155] When the deviation rate between the actual load on the hourly time scale and the day-ahead predicted load exceeds a set ratio, the phase sequence optimization module is triggered to regenerate the phase sequence adjustment scheme, thereby realizing a rolling closed loop of day-ahead phase sequence optimization and hourly absorption assessment.

[0156] It should be noted that the implementation steps of multi-timescale coordination are as follows: At the day-ahead timescale, the process is executed daily at 00:00. The phase sequence optimization module generates an initial phase sequence adjustment scheme based on the day-ahead load forecast and photovoltaic forecast. The loss verification module verifies and corrects the initial scheme, outputting the final day-ahead phase sequence adjustment scheme. At the hourly timescale, the process is executed every hour on the hour. The margin assessment module updates the dynamic absorption capacity based on the deviation between the actual load and the predicted load. The deviation rate is the absolute value of the difference between the actual load and the predicted load divided by the predicted load and then multiplied by 100%. At the minute-level timescale, the process is executed every 15 minutes. The power flow suppression module updates the graded suppression instructions based on the deviation between the actual photovoltaic output value and the predicted value.

[0157] Rolling closed-loop trigger condition: When the deviation rate on the hourly time scale exceeds a set percentage, the phase sequence optimization module is triggered to re-execute phase sequence optimization. The set percentage is determined based on the load forecast accuracy of the distribution substation area, with a typical value of 20%. After triggering, the day-ahead forecast data is replaced with the latest actual load data, keeping the already executed adjustments in the phase sequence adjustment scheme unchanged, and only re-optimizing the unexecuted parts.

[0158] Technical effects: Through multi-timescale hierarchical coordination at the day-ahead, hourly, and minute levels, a combination of coarse-grained global optimization of phase sequence optimization and fine-grained real-time response of power flow suppression is achieved; the rolling closed-loop mechanism automatically triggers re-optimization when there is a significant deviation between the actual load value and the predicted value, ensuring the system's adaptability and robustness.

[0159] As an optional embodiment, the duration of the time scale window in the time scale alignment unit is dynamically adjusted according to the day-ahead fluctuation rate of the distribution area load;

[0160] When the day-ahead volatility is greater than the high volatility threshold, the duration of the time-scale window is shortened to a first duration.

[0161] When the day-ahead volatility is less than the low volatility threshold, the duration of the time-scaled window is extended to a second duration.

[0162] The day-ahead volatility is the ratio of the standard deviation to the mean of the load power values ​​at each hour within the previous 24 hours.

[0163] It should be noted that the implementation steps for dynamically adjusting the time-scaled window duration are as follows: Calculate the day-ahead volatility as the standard deviation of the load power values ​​at each hour within the previous 24 hours divided by the mean; determine the relationship between the day-ahead volatility and the high and low volatility thresholds. The high volatility threshold is typically 0.35 based on the random fluctuation characteristics of the distribution area load, and the low volatility threshold is typically 0.15; when the day-ahead volatility is greater than the high volatility threshold, the time-scaled window duration is shortened to a first duration of 15 minutes to improve the ability of the time-varying correlation coefficient to capture rapid load changes; when the day-ahead volatility is less than the low volatility threshold, the time-scaled window duration is extended to a second duration of 60 minutes to smooth random noise and improve the stability of correlation identification; when the day-ahead volatility is between the high and low volatility thresholds, the time-scaled window duration remains at the default value of 30 minutes.

[0164] Technical effect: By dynamically adjusting the duration of the time-stamped window, the topology identification module can adapt to changes in the load fluctuation characteristics of the distribution area, improve the temporal resolution of identification in high-fluctuation scenarios, and improve the statistical stability of identification in low-fluctuation scenarios, thereby improving the topology identification accuracy of the entire system under different operating conditions.

[0165] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A distribution substation microgrid operation optimization system, characterized in that, include: The topology identification module is used to generate a topology phase sequence correlation matrix based on the voltage amplitude and power injection data of each node on the low-voltage side of the distribution substation. The phase sequence optimization module is connected to the topology identification module and is used to receive the topology phase sequence correlation matrix and solve the phase sequence adjustment scheme with the goal of reducing the three-phase imbalance and equalizing the transformer phase capacity margin. The margin assessment module, connected to the phase sequence optimization module, is used to calculate the dynamic absorption space based on the load distribution of each phase and the transformer thermal circuit model corresponding to the phase sequence adjustment scheme. The power flow suppression module, connected to the margin assessment module, is used to generate graded suppression instructions in the order of priority absorption by energy storage, time-series adjustment of transferable load, and smooth derating of photovoltaic output when the photovoltaic output exceeds the dynamic absorption space. The reactive power coordination module, connected to the phase sequence optimization module, is used to calculate the reactive power adjustment margin between phases and allocate reactive power compensation based on the load balance of each phase corresponding to the phase sequence adjustment scheme. The loss verification module is connected to the phase sequence optimization module and is used to calculate the additional loss of three-phase imbalance according to the phase sequence adjustment scheme. When the additional loss of three-phase imbalance increases, it sends a constraint correction signal to the phase sequence optimization module.

2. The distribution substation microgrid operation optimization system according to claim 1, characterized in that, The topology identification module includes: The time-scale alignment unit is used to align the voltage amplitude sampling data and power injection sampling data of each node according to the same absolute time scale, eliminate abnormal sampling points whose time scale deviation exceeds the clock synchronization accuracy of the sampling device, and form the voltage amplitude sequence and power injection sequence within the same time scale window; The correlation coefficient calculation unit, connected to the time scale alignment unit, is used to calculate the Pearson correlation coefficient between the power injection change of the j-th node and the voltage amplitude change of the i-th node, and uses the Pearson correlation coefficient as the time-varying correlation coefficient. The matrix generation unit, connected to the correlation coefficient calculation unit, is used to identify the phase sequence of each node based on the matrix characteristics of the time-varying correlation coefficient. When the time-varying correlation coefficient between the power injection change of the j-th node and the voltage amplitude change of phase A of the i-th node is greater than the time-varying correlation coefficient between the voltage amplitude changes of phase B and phase C, the j-th node is determined to be a phase A load node.

3. The distribution substation microgrid operation optimization system according to claim 2, characterized in that, The phase sequence optimization module includes a load clustering unit, which is connected to the matrix generation unit. The load clustering unit is used to classify load nodes into constant base loads, random fluctuating loads, and photovoltaic coupled loads based on the power curve characteristics of each node extracted from the topological phase sequence correlation matrix and the K-means clustering algorithm based on Euclidean distance. The constant base load class includes load nodes whose daily power curve peak-to-mean ratio is less than a first ratio threshold and whose power volatility is less than a first volatility threshold. The random fluctuation load class includes load nodes whose daily power curve peak-to-mean ratio is greater than the first ratio threshold and whose power fluctuation rate is greater than the first fluctuation threshold. The photovoltaic coupled load class includes load nodes whose daily power curve and the photovoltaic power generation curve of the distribution area have a Pearson correlation coefficient greater than the correlation coefficient threshold. When a load node simultaneously meets the classification criteria for both the random fluctuation load category and the photovoltaic coupled load category, it is preferentially classified into the photovoltaic coupled load category.

4. The distribution substation microgrid operation optimization system according to claim 3, characterized in that, The phase sequence optimization module further includes a model building unit and a solution unit, wherein the model building unit is connected to the load clustering unit; The model building unit is used to fix the phase sequence label of the constant base load class to the current phase sequence, and use the phase sequence labels of the random fluctuation load class and the photovoltaic coupled load class as 0-1 integer decision variables to build a phase sequence optimization model with the optimization objective of minimizing the weighted sum of the reduction of three-phase current imbalance and the equalization of inter-phase capacity margin. The weighting coefficient for phase-to-phase capacity margin equalization is dynamically set using a piecewise linear function based on the current load rate of the transformer. When the load rate is less than 0.5, the weighting coefficient is 0.

3. When the load rate is between 0.5 and 0.8, the weighting coefficient is 0.3 plus 0.7 multiplied by the difference between the load rate and 0.5 and then divided by 0.

3. When the load rate is greater than 0.8, the weighting coefficient is 1.0 plus 0.67 multiplied by the difference between the load rate and 0.8 and then divided by 0.

2. The solution unit is connected to the model building unit and is used to solve the phase sequence optimization model using the branch and bound method to obtain the optimal values ​​of each integer decision variable, and generate the phase sequence adjustment scheme for each load node based on the optimal values.

5. The distribution substation microgrid operation optimization system according to claim 4, characterized in that, The margin assessment module includes: The load prediction unit is used to allocate each load node to the corresponding phase sequence according to the phase sequence adjustment scheme, and then calculate the load power prediction curves of phase A, phase B and phase C respectively. The temperature calculation unit, connected to the load prediction unit, is used to calculate the hourly change value of the hot spot temperature of each phase winding in the next 24 hours based on the load power prediction curve of each phase and the thermal time constant of the transformer winding thermal circuit model, using the recursive thermal balance equation. The recursive thermal balance equation is the exact analytical solution of the first-order differential equation of the transformer thermal circuit model. The absorption extraction unit, connected to the temperature calculation unit, is used to compare the maximum value of the hourly change value of the hot spot temperature of each phase winding with the transformer insulation heat resistance temperature. The difference between the insulation heat resistance temperature and the maximum value is the temperature margin. The increase in load power of each phase is determined according to the temperature margin and the conversion factor. The minimum increase in load power of each phase is the dynamic absorption space.

6. The distribution substation microgrid operation optimization system according to claim 5, characterized in that, The power flow suppression module includes: An overflow calculation unit is used to calculate the difference between the reverse active power value on the high-voltage side of the distribution transformer and the dynamic absorption space, wherein the difference is the power overflow amount. A strategy matching unit, connected to the overflow calculation unit, is used to compare the power overflow amount with the current available charging power of the energy storage system. When the power overflow amount is less than or equal to the current available charging power, a first-level strategy is matched. When the power overflow amount is greater than the current available charging power, the excess portion is compared with the transferable capacity of the transferable load. When the excess portion is less than or equal to the transferable capacity, a second-level strategy is matched; otherwise, a third-level strategy is matched. The instruction synthesis unit, connected to the strategy matching unit, is used to synthesize a suppression instruction that only increases the charging power of the energy storage system under the first-level strategy, to synthesize a suppression instruction that increases the charging power of the energy storage system and adjusts the power consumption timing of the transferable load under the second-level strategy, and to synthesize a suppression instruction that increases the charging power of the energy storage system, adjusts the power consumption timing of the transferable load, and limits the photovoltaic output value according to the set power reduction rate under the third-level strategy.

7. The distribution substation microgrid operation optimization system according to claim 4, characterized in that, The reactive power coordination module includes: The margin calculation unit is used to calculate the reactive power regulation margin of each phase photovoltaic power generation unit and the reactive power regulation margin of each phase energy storage system based on the load balance degree of each phase corresponding to the phase sequence adjustment scheme. The deviation grading unit, connected to the margin calculation unit, is used to classify nodes into voltage over-limit nodes and voltage qualified nodes according to the ratio of the voltage deviation value of each node in the distribution area to the allowable deviation range, and only performs reactive power compensation allocation for voltage over-limit nodes. The output allocation unit, connected to the deviation classification unit, is used to allocate the reactive power deficit of the voltage over-limit node into the reactive power output adjustment amount of the photovoltaic power generation unit and the reactive power compensation amount of the energy storage system in the order of priority consumption of the reactive power adjustment margin of the phase where the voltage over-limit node is located, supplementary consumption of the reactive power adjustment margin of the same phase energy storage, and cross-phase support of the reactive power adjustment margin of the adjacent phase.

8. The distribution substation microgrid operation optimization system according to claim 7, characterized in that, The loss verification module includes: The unbalance calculation unit is used to calculate the A-phase current, B-phase current and C-phase current of each branch according to the phase sequence assignment of each load node and the load power value of each node in the phase sequence adjustment scheme, and to calculate the negative sequence current component and zero sequence current component of each branch according to the A-phase current, the B-phase current and the C-phase current. The loss comparison unit is connected to the unbalance calculation unit and is used to calculate the three-phase unbalance additional loss of each branch after phase sequence reconstruction based on the negative sequence current component and the zero sequence current component, and to compare the three-phase unbalance additional loss with the three-phase unbalance additional loss of the corresponding branch before reconstruction at the branch level. The weight correction unit, connected to the loss comparison unit, is used to generate a constraint correction signal when the additional loss of the three-phase imbalance after phase sequence reconstruction of a certain branch is greater than the additional loss of the three-phase imbalance before reconstruction. The constraint correction signal is the result of multiplying the imbalance penalty term coefficient of the branch in the phase sequence optimization model by a correction coefficient. The correction coefficient is determined according to the increase ratio of the additional loss of the branch and has a value range of 1.0 to 2.

0. When the correction coefficient is greater than 2.0, the phase sequence labels of all adjustable load nodes on the branch are forcibly fixed to the current phase sequence.

9. The distribution substation microgrid operation optimization system according to claim 8, characterized in that, The phase sequence optimization module generates an initial phase sequence adjustment scheme on a day-ahead time scale, and the loss verification module verifies the initial phase sequence adjustment scheme on a day-ahead time scale. The margin assessment module updates the dynamic absorption capacity based on the deviation between the actual load and the predicted load on an hourly time scale. The power flow suppression module updates the graded suppression command on a minute-level time scale based on the deviation between the actual and predicted photovoltaic output values. When the deviation rate between the actual load on the hourly time scale and the day-ahead predicted load exceeds a set ratio, the phase sequence optimization module is triggered to regenerate the phase sequence adjustment scheme, thereby realizing a rolling closed loop of day-ahead phase sequence optimization and hourly absorption assessment.

10. A distribution substation microgrid operation optimization system according to claim 2, characterized in that, The duration of the time-scale window in the time-scale alignment unit is dynamically adjusted according to the day-ahead fluctuation rate of the distribution area load; When the day-ahead volatility is greater than the high volatility threshold, the duration of the time-scale window is shortened to a first duration. When the day-ahead volatility is less than the low volatility threshold, the duration of the time-scaled window is extended to a second duration. The day-ahead volatility is the ratio of the standard deviation to the mean of the load power values ​​at each hour within the previous 24 hours.

Images