Transmission-distribution-micro multi-level power grid distributed collaborative optimization scheduling method

Abstract

The invention relates to the technical field of power grid optimization scheduling, and discloses a transmission-distribution-micro multi-level power grid distributed collaborative optimization scheduling method. The method comprises the following steps: respectively establishing distributed optimization models of a micro-grid, a power distribution network and a power transmission network, obtaining a schedulable domain, and obtaining distributed optimization scheduling results of the micro-grid, the power distribution network and the power transmission network by adopting a robustness optimization method based on time sequence state deduction; and finally, based on the obtained distributed optimization scheduling results of the micro-grid, the power distribution network and the power transmission network, adopting an iterative optimization method of a transmission-distribution-micro multi-level power grid based on alternative reconstruction of the schedulable domain to obtain the transmission-distribution-micro system schedulable domain meeting the security constraint of the multi-level power grid. According to the method, the uncertainty of new energy power generation is considered, and optimal scheduling of a complex nonlinear power system with multiple levels and multiple scheduling objects is realized based on a robustness optimization method of time sequence state deduction.

Description

Technical Field

[0001] This invention relates to the field of power grid optimization scheduling technology, specifically a distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids. Background Technology

[0002] In recent years, with the rapid transformation of the energy structure and the increasing maturity of power electronic equipment power generation control technology, the penetration rate of wind and solar new energy power generation in the power system has been increasing year by year. However, while promoting the clean and low-carbon transformation of society, new energy power generation has also profoundly changed the operating characteristics of the power grid. At the distribution network level, with the access of a large number of new energy power generation equipment and energy storage devices, the distribution network has transformed from a traditional load system into a multi-source-load-storage system with distribution lines as the backbone and microgrids as nodes. The surge in the number of controllable objects has led to the curse of dimensionality in power flow optimization calculations. From the perspective of the main grid, the access of a large number of new energy power plants has also placed higher demands on the transmission system's cross-sectional control, frequency and voltage regulation, and power source ramping constraints. Therefore, traditional global optimization control methods are no longer suitable for the "high-power and high-constraint" power system with its complex controllable objects and numerous constraints. To solve the curse of dimensionality caused by massive distributed power sources and energy storage, aggregated control based on the microgrid energy management system is a feasible solution. In addition, the total adjustable capacity of the distribution network is small, and relying solely on its own level of regulation capabilities, it is difficult to optimize the allocation of source and load resources of each microgrid while simultaneously taking into account the various safety and stability control requirements of the upper-level transmission network. Therefore, designing a coordinated control and optimized scheduling method for transmission networks, distribution networks, and microgrids is the most effective solution to achieve coordinated sharing of multi-level regulation resources in the new power system, facilitate the consumption of new energy sources, and improve the stability, economy, and environmental friendliness of power system operation.

[0003] Achieving coordinated optimization of multi-level power grids requires considering the system characteristics and security constraints of each level and finding the globally optimal scheduling solution within the target optimization timescale. Traditional global optimization methods, such as the Lagrange relaxation node method, particle swarm optimization algorithm, mixed-integer linear programming, and neural network algorithms, suffer from excessive computational complexity and time consumption, making them unsuitable for today's multi-level systems with numerous control resources. To avoid the computational complexity of global solutions, distributed solution algorithms have been extensively studied. The core idea of ​​these algorithms is to partition the multi-level system into layers based on its operating characteristics, optimize sub-problems in each partition to obtain feasible solutions that satisfy the security constraints of that partition, and then perform coordinated optimization on the upper-level partitions to find the optimal scheduling solution that satisfies the global security constraints.

[0004] Existing research, based on its own research perspective, partitions multi-level power grids and employs different algorithms to achieve collaborative optimization of these grids. From an algorithm performance perspective, the alternating direction multiplier method is simple in principle, achieves global convergence, and has high accuracy, but its convergence speed is slow. The iterative solution of the objective cascade method may yield locally converged solutions, requiring reconstruction and correction of the partitioned schedulable solutions under constraint conditions, which is also time-consuming. Optimal condition decomposition can achieve global convergence calculation relatively quickly, but its solution process adjusts the partitioned optimization results to meet the safety constraints of each partition, resulting in slightly less economic efficiency in the final outcome. Furthermore, existing research, when jointly solving multi-level power grids, does not perform full-time-domain collaborative optimization and correction within the target time period, which may lead to future exceedances of limits for partitioned voltage, power flow, and other electrical quantities.

[0005] To address the issue of scheduling optimization results potentially exceeding limits in future time periods, existing research includes point-by-point traversal methods, gradient descent iterative methods, and constraint exchange iterative methods. Among these, while point-by-point traversal methods offer the highest accuracy, the sheer number of running points required for analysis results in a massive computational burden, hindering practical applications.

[0006] In summary, distributed collaborative optimization of transmission-distribution-micro multi-level power grids faces the following challenges: First, the integration of a large number of distributed new energy sources and energy storage systems exacerbates the nonlinearity of the system, and the number of control objects increases dramatically, leading to a surge in safety constraints and a massive computational burden. Second, while distributed collaborative optimization algorithms can improve computational efficiency, they cannot guarantee that each zone meets safety and stability constraints throughout the entire time period. Third, the topology of the distribution network is difficult to obtain accurately and completely, posing challenges to collaborative optimization calculations. Summary of the Invention

[0007] To address the aforementioned problems, the present invention aims to provide a distributed collaborative optimization scheduling method for multi-level power grids (transmission-distribution-microgrids). Taking into account the uncertainties of new energy generation, this method utilizes a robust optimization approach based on time-series state deduction to achieve optimized scheduling of complex nonlinear power systems with multiple levels and scheduling objects. The technical solution is as follows:

[0008] A distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids includes the following steps:

[0009] Step 1: Establish a distributed optimization model for the microgrid to obtain the dispatchable domain of the microgrid for new energy consumption. Then, use a robust optimization method based on time-series state deduction to correct the dispatchable domain of the microgrid and obtain the distributed optimization scheduling results at the microgrid level.

[0010] Step 2: Establish a distributed optimization model for the distribution network, obtain the dispatchable domain of the distribution network based on the dimension-upgrading mapping algorithm in high-dimensional linear space, and obtain the distributed optimization scheduling results at the distribution network level based on the robust optimization method of time-series state deduction.

[0011] Step 3: Establish a distributed optimization model for the power transmission network to obtain the dispatchable domain of the power transmission network. Based on the robust optimization method of time-series state deduction, the distributed optimization scheduling results at the power transmission network level can be obtained.

[0012] Step 4: Based on the distributed optimization scheduling results of microgrids, distribution networks and transmission networks obtained in Steps 1-3, an iterative optimization method based on alternating reconfiguration of schedulable domains for transmission-distribution-micro multi-level power grids is adopted to obtain the schedulable domains of the transmission-distribution-micro system that satisfy the security constraints of the multi-level power grid.

[0013] The beneficial effects of this invention are:

[0014] 1) This invention establishes a distributed solution model structure for collaborative optimization of transmission-distribution-micro multi-level power grids, and proposes a robust optimization method based on time-series state deduction. It also takes into account the uncertainty of new energy output, and realizes the objective function optimization solution that satisfies various security constraints of the transmission, distribution and micro three-level power grids within the target time period.

[0015] 2) This invention proposes a method for obtaining the equivalent power flow model of a distribution network system based on an upgraded mapping function. An equivalent linearized model of the power flow of the distribution network is established in a high-dimensional space. This solves the problems of difficulty in establishing the power flow equation and time-consuming solution caused by the strong nonlinearity of the distribution network, the large number of devices, and the difficulty in fully obtaining the parameters under the high proportion of new energy access. This effectively improves the computational efficiency of the power flow of the distribution network.

[0016] 3) This invention proposes a distributed collaborative optimization method for transmission-distribution-micro multi-level power grids. It corrects and reconstructs the distributed solution results for the three-level power grid by using an alternating reconfiguration method of schedulable domains. Based on the idea of ​​partitioned iterative solution, it effectively solves the problem of drastically increased computational load caused by global iterative solution in traditional partitioned optimization methods. Attached Figure Description

[0017] Figure 1 This is a flowchart of the distributed collaborative optimization scheduling method for multi-level power grids (transmission-distribution-micro) of the present invention.

[0018] Figure 2 This is a topology diagram of the simulation example of the transmission-distribution-micro multi-level power grid of the present invention.

[0019] Figure 3(a) shows the predicted range / actual power curve of photovoltaic power generation on a day with sufficient sunshine.

[0020] Figure 3(b) shows the predicted range / actual curve of photovoltaic power generation on cloudy days.

[0021] Figure 3(c) shows the predicted range / actual power curve of wind power on windy days.

[0022] Figure 3(d) shows the predicted range / actual curve of wind power on days with low wind speeds.

[0023] Figure 4(a) shows the optimized scheduling results of the new energy low-power daily microgrid.

[0024] Figure 4(b) shows the optimized scheduling results of the new energy high-power daily microgrid.

[0025] Figure 5(a) shows the switching power and dispatchable domain of a new energy low-power microgrid.

[0026] Figure 5(b) shows the switching power and dispatchable domain of the new energy high-power microgrid.

[0027] Figure 6(a) shows the voltage comparison of distribution network nodes—node voltage of the 30th-order upgraded model of residential load.

[0028] Figure 6(b) shows the voltage comparison of distribution network nodes—node voltage of the 54th-order upgraded model of residential load.

[0029] Figure 6(c) shows the voltage comparison of distribution network nodes—node voltage of the 30th-order upgraded model of industrial load.

[0030] Figure 6(d) shows the voltage comparison of distribution network nodes—node voltage of the 54th-order upgraded model of industrial load.

[0031] Figure 7(a) shows the results of the optimized scheduling of the distribution network.

[0032] Figure 7(b) shows the power exchange between the distribution network and the transmission network and the dispatchable domain of the distribution network.

[0033] Figure 8 To optimize the scheduling results of the power transmission network. Detailed Implementation

[0034] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0035] 1. Distributed optimization model of transmission-distribution-micro multi-level power grid:

[0036] This embodiment establishes a distributed optimization model for a multi-level power grid (transmission-distribution-microgrid) oriented towards optimal absorption of new energy. First, optimization models for each level of the microgrid, distribution network, and transmission network are established respectively. Then, a collaborative optimization method for the multi-level power grid (transmission-distribution-microgrid) based on the boundary reconfiguration method is given.

[0037] (1) Dispatchable domain of microgrids for renewable energy consumption:

[0038] First, based on power prediction methods, the power prediction results of new energy sources and electricity load in the future period are obtained. Spatiotemporal graph convolutional networks can explore the spatiotemporal correlation characteristics of power generation and consumption between various nodes, and achieve accurate prediction of the power of new energy sources and loads at each node. Fuzzy clustering algorithms can divide the sample weather according to the volatility and particularity of the sample data, and use the grouped samples to train the spatiotemporal graph convolutional network. The above algorithms have been widely used in the field of power prediction, and will not be elaborated on in this invention.

[0039] The following describes the optimized scheduling results at the microgrid level during the target time period. The controllable objects in the microgrid system include controllable power sources, energy storage, and demand response loads. The power of each controllable object is integrated as a control variable of the microgrid, as shown in Equation (1).

[0040] (1); In the formula: superscript " (i=1… This indicates that the electrical quantity belongs to the i-th microgrid. Indicates the number of microgrids within the same distribution network. This represents the control variable vector of microgrid i. , and These represent the controllable load power, controllable power supply power, and energy storage charging and discharging power within the microgrid, respectively, with t representing time.

[0041] The inherent loads that cannot be adjusted within the microgrid, the power generation of new energy sources, and the remaining energy storage capacity determined by the energy storage operation status are taken as the state variables of the microgrid and arranged into a vector form as shown in equation (2).

[0042] (2); In the formula: This represents the state variable vector of microgrid i. , and These represent the predicted values ​​of unadjustable power load, new energy power, and remaining energy storage capacity of microgrid i, respectively.

[0043] Energy storage remaining power state vector It can be obtained by time-domain integration of energy storage charging and discharging, as shown in equation (3).

[0044] (3); In the formula: This represents the initial remaining power of the energy storage device in microgrid i. Indicates the initial time. , These represent the charging and discharging efficiencies of the energy storage device, respectively. , They represent the target time periods respectively. ~ Inner The charging and discharging power of energy storage within a specific time period, and the target time period. ~ The unit duration within is T represents the target time period. ~ Number of time periods within.

[0045] The main constraints within a microgrid are equipment operating power constraints and energy storage remaining power status constraints. Among them, equipment operating power constraints include new energy power generation constraints, controllable load output constraints, controllable power supply power constraints, energy storage charging and discharging power constraints, and the exchange power constraints between the microgrid and the distribution network. The expressions for the above constraints are shown in Equation (4).

[0046] (4); In the formula: Indicates the rated power of new energy sources. and These represent the maximum and minimum adjustable power of the controllable load, respectively. and Indicates the maximum and minimum generating power of the controllable power source. and Indicates the upper and lower limits of energy storage charging and discharging power. and These represent the exchange power between microgrid i and its distribution network, and the maximum allowable power for interaction, respectively.

[0047] The constrained power of the remaining energy storage state is shown in equation (5).

[0048] (5); In the formula: , These represent the maximum and minimum remaining energy storage states within microgrid i, respectively.

[0049] The operating cost function of various resources in the microgrid is shown in equation (6).

[0050] (6); In the formula: , and These represent the controllable power generation cost, energy storage charging and discharging cost, and controllable load dispatching cost, respectively. This represents the operating cost of a controllable power supply as a function of its operating power. This indicates the basic operating cost of a controllable power supply. This represents the unit charge / discharge cost of the energy storage device after conversion. This indicates the unit compensation amount for demand response load; For the first in the microgrid indivual Controllable power supply during different time periods; For the first in the microgrid indivual Energy storage charging and discharging power during a given time period; For the first in the microgrid indivual The change in controllable load power over a given period of time.

[0051] The output of new energy sources is uncertain, and forecasting errors may lead to the curtailment of new energy power or excessive demand response load, resulting in additional operating costs. The forecasting error of new energy sources can be characterized by probability density functions, as shown in equations (7) and (8).

[0052] (7); (8); In the formula: , Let represent the probability density functions of photovoltaic and wind power, respectively. It usually approximates the Beta distribution. , It is the shape parameter of the Beta distribution, obtained from historical data of light resources. , This indicates the actual and maximum values ​​of photovoltaic power generation. Represents the gamma function; It typically approximates a Weibull distribution, where c and d are the shape parameters of the Weibull distribution, obtained from historical wind resource data. , These represent the cut-in wind speed and rated wind speed of the fan, respectively. , These represent the fan output power and the rotor rated power, respectively. Let c be the k-th power of the Weiber distribution scaling parameter. Cut off the wind speed for the fan.

[0053] Based on equations (7) and (8), the risk power of new energy curtailment can be obtained, as shown in equation (9). This part of the new energy curtailment power is the additional loss caused by prediction deviation.

[0054] (9); In the formula: Indicates the power at risk of curtailment of renewable energy. , These represent the upper limits of the predicted photovoltaic power and wind power, respectively.

[0055] The following describes a robust optimization method based on time-series state deduction. The core of this method lies in ensuring that the optimization results within the target time period satisfy system safety constraints, taking into account the impact of uncertainties. The specific steps of this method are as follows:

[0056] (a) With the goal of the most economical operation of the microgrid, while taking into account the weight of the curtailment of new energy, and using the curtailment risk power caused by the prediction deviation of new energy as a penalty term, and taking into account the power and energy storage remaining power constraints, the optimal scheduling function of microgrid i can be obtained, as shown in Equation (10).

[0057] (10); In the formula: For time-of-use electricity pricing, This represents the actual amount of renewable energy curtailed within microgrid i. The risk factor representing the power at risk of renewable energy curtailment within microgrid i is influenced by multiple factors, including equipment operational safety and green certificate indicators, and is determined by the microgrid operation and dispatching agency. , , This is the weighting factor.

[0058] (b) Simulate the operating status of the microgrid during the target time period.

[0059] based on The initial state variables of the microgrid at time t are derived from the state equations. ~ The microgrid state variables at each moment within the time period are obtained, and combined with the objective function shown in (10), the microgrid control variables that satisfy the safety constraints within each unit of time are optimized to form the microgrid dispatchable domain. The time-series state derivation formula of the microgrid is shown in equation (11).

[0060] (11); In the formula: , , These represent the state matrix, control matrix, and input matrix of the uncertain disturbance of microgrid i, respectively. This represents the disturbance input vector of microgrid i.

[0061] (c) Correct the obtained microgrid dispatchable domain.

[0062] Establish a robustness verification inequality constraint set, and perform security constraint verification on the microgrid dispatchable domain obtained by equations (10) and (11). At the same time, take into account the degree of exceeding the limit within the tolerable unit time period, calculate the time-domain cumulative robustness constraint discrimination index, and the specific calculation formula is shown in equation (12).

[0063] (12); In the formula: ν represents the solution result of the linear programming and serves as a criterion. When ν > 0, it indicates that the microgrid state variables at time t do not meet the safety constraints. Indicates the state variables of microgrid i in The sensitivity index matrix of the time period to the boundary conditions of the constraint. Denotes the relaxation vector of microgrid i. This vector represents the level of over-limit tolerance at time t, determined by the device parameters.

[0064] (d) Update ~ The modified dispatchable domain of microgrid i within the time period.

[0065] After removing the scheduling results that do not meet the safety constraints calculated based on equation (8), a new schedule is generated based on the constraint set. ~ The dispatchable boundary of microgrid i within the time period is obtained, and then the modified dispatchable domain of microgrid i that can be satisfied throughout the entire time period is obtained. The specific calculation formula is shown in Equation (13).

[0066] (13); In the formula: , , Let i represent the original feasible region, the modified feasible region, and the newly generated feasible region of microgrid i, respectively.

[0067] Then return to (b) to perform state deduction for the next moment, until completion. ~ The calculation of all time periods is performed to obtain the complete dispatchable domain of the microgrid i.

[0068] (e) Based on the obtained ~ Within the time period, the complete dispatchable domain of microgrid i is calculated. ~ The exchange power between microgrid i and its distribution network may have a solution. Based on the enumeration method, and using the power equation and the active and reactive power constraints of equipment operation, the distribution network node at that time is calculated. ~ The active and reactive power available for dispatch within a given time period are used as elements to form a set of available ranges for distribution network nodes, as shown in (14).

[0069] (14); In the formula: This represents the set of dispatchable ranges for the aforementioned distribution network nodes; To represent the reactive power exchange between microgrid i and its distribution network; the subscript "*" can be "L" (non-adjustable power load), "L_C" (controllable load), "S" (controllable power source), "E" (energy storage) or "N" (new energy). Indicates the type of equipment within the microgrid at time t. The corresponding active power; Indicates the type of equipment within the microgrid at time t. The corresponding reactive power; To represent the equipment categories within the M layer of a microgrid The reactive-active power coupling function.

[0070] (2) Dispatchable domain of distribution network based on high-dimensional linear space algorithm:

[0071] With the gradual advancement of power grid construction, the number of nodes and lines in distribution network systems has surged. The computational load for power flow calculations is exponentially related to the number of nodes, making it difficult for traditional calculation methods to achieve fast solutions. Furthermore, with the integration of a high proportion of renewable energy sources, distribution network control equipment has become more diverse, and system nonlinearity has significantly increased, which will further reduce the calculation speed.

[0072] To address this, a dimension-upgrading mapping algorithm based on Koopman theory was designed in a high-dimensional linear space to obtain the equivalent network structure of the distribution network in the high-dimensional space, which serves as the basis for fast power flow calculation. Furthermore, the safety constraints of the distribution network are taken into account to obtain the feasible region of the distribution network.

[0073] The dimension-scaling mapping algorithm in high-dimensional linear space is based on Koopman theory. By establishing a dimension-scaling mapping function and training based on historical data, the low-dimensional nonlinear equation is subjected to dimension-scaling and power-reduction processing to obtain an equivalent high-dimensional linear equation. At the same time, it makes up for the defect of local convergence of nonlinear and non-convex relations, achieves fast global convergence of solution, and improves the operation speed. The specific process is as follows.

[0074] The principle of the dimension-upgrading mapping is shown in Equation (15), that is, for any nonlinear multivariate function, a corresponding linear multivariate function can be found in the high-dimensional space by establishing a dimension-upgrading mapping relationship.

[0075] (15); In the formula: Represents a nonlinear nth-order function. (g=1…n) represents the independent variable of order g, and H(x) is N. f The dimensional mapping function can be flexibly selected according to the properties of the computational object. This is an up-dimensional mapping matrix.

[0076] The historical actual values ​​of power flow in the distribution network are used as the training set. (Upgraded dimension mapping matrix) It can be regarded as the weight of the dimensional mapping from each independent variable to the dependent variable. The weight values ​​of each variable can be solved by the least squares method based on historical data, with the minimum trend regression error as the benchmark. The specific expression is shown in Equation (16).

[0077] (16); In the formula: This represents the Moore-Penrose generalized inverse matrix.

[0078] The solution complexity of the distribution network is approximately exponential in relation to the number of nodes. Therefore, this invention selects a logarithmic series based on the Euclidean distance weights of the independent variable and the basis vector to construct the dimension-upgrading mapping function, the specific expression of which is shown in equation (17).

[0079] (17); In the formula: This represents the basis vectors for increasing the dimension.

[0080] Each microgrid node in the distribution network can be regarded as a PQ node, while the power plant nodes such as hydropower, thermal power and energy storage can be regarded as PV nodes, and the nodes connected to the transmission network are regarded as balancing nodes. Therefore, in the distribution network dimensionality upgrade mapping, the mapping relationship from independent variable x to dependent variable y is established as shown in equation (18).

[0081] (18); In the formula: , , These represent the active power, reactive power, and voltage amplitude at the PQ node, respectively. , , These represent the active power, reactive power, and voltage amplitude of the PV node, respectively.

[0082] Based on the upgraded mapping relationship established by equations (16) to (18), historical data is used for training to obtain the upgraded linearized power flow equation of the distribution network, as shown in equation (19).

[0083] (19); In the formula: superscript " (j=1…n)p This indicates that the electrical quantity belongs to the j-th distribution network. , , , Let represent the node voltage vector, node injected active power vector, node injected reactive power vector, and other high-dimensional linear variables of distribution network j, respectively. , , represents the sensitivity matrix of active power, reactive power, and high-dimensional linear variables to node voltage within distribution network j, respectively, where each element is a constant value obtained based on historical data training.

[0084] The voltage constraints of distribution network nodes can be obtained based on the power grid operation criteria and are given by the distribution network operation and dispatching agency. Based on the upper and lower bounds of the voltage of each node in distribution network j, the power relationship equations of each node in distribution network j are characterized as shown in equation (20).

[0085] (20); In the formula: and Let represent the upper and lower bounds of the voltage constraint at node p within the distribution network j, respectively. , , These represent the sensitivity coefficients of the active power, reactive power, and other high-dimensional linear variables of the m-th node in the distribution network j to the voltage amplitude of the p-th node, respectively. , and The active power, reactive power, and other high-dimensional linear variables of the m-th node in the j-th distribution network are respectively represented by the superscript "~" and the subscript "~". The superscript "~" and the subscript "~" respectively represent the node variables corresponding to the upper and lower bounds of the voltage of the p-th node in the distribution network.

[0086] Based on equation (20), the power constraint relationship of each node when the voltage upper and lower bounds of each node in the distribution network can be obtained, which can be used as the voltage constraint condition of the distribution network nodes. In addition, except for microgrid nodes, the safety constraints that the regulating equipment of other nodes in the distribution network should meet are shown in equation (21).

[0087] (twenty one); In the formula: and These represent the renewable energy generation power and rated power of the p-th node within the distribution network j, respectively. , and Let represent the actual, maximum, and minimum power of the controllable power source at node p within the distribution network j, respectively. , and Let these represent the actual, upper limit, and lower limit power of the energy storage at node p within the distribution network j, respectively. and These represent the exchange power between distribution network j and its transmission network, and the maximum allowable exchange power, respectively. , and These represent the real-time, upper limit, and lower limit of the energy storage capacity at node p within the distribution network j, respectively.

[0088] Similar to Equation (10), the goal is still to achieve the most economical operation of the distribution network, while also taking into account the weight of new energy curtailment. The power of curtailment risk caused by new energy prediction deviation is used as a penalty term to obtain the optimization function of distribution network j, as shown in Equation (22).

[0089] (twenty two); In the formula: and They represent time periods respectively. The power generation cost and energy storage charging and discharging cost of the controllable power source at node p within the distribution network j; , and These represent the actual power curtailment, curtailment risk factor, and curtailment risk power of new energy sources within distribution network j, respectively. , and As a weighting factor; For the distribution network j in time period The exchange power between the internal power supply and the local power grid; Let be the equivalent active power exchanged between distribution network node j and its connected microgrid i at time t.

[0090] Similarly, the robust optimization method based on time-series state deduction can obtain the dispatchable domain of the distribution network, as shown in Equation (23).

[0091] (twenty three); In the formula: This represents the set of dispatchable ranges for distribution network j.

[0092] (3) Distributed optimization model of power transmission network:

[0093] The power flow equations of the transmission network are shown in equation (24).

[0094] (twenty four); In the formula: , and These represent the injected active power, reactive power, and AC voltage at node s in the transmission network, respectively. This indicates that transmission network node s is adjacent to transmission network node k; This represents the AC voltage at node k in the transmission network; , , , These represent the real part of the node admittance matrix element, the node phase angle difference, the imaginary part of the node admittance matrix element, and the parallel s s of the transmission network node, respectively.

[0095] The power flow constraints of the transmission network are shown in equation (25).

[0096] (25); In the formula: and These are the power flow and thermal stability limits of the SK branch of the transmission network, respectively. This represents the power flow transfer coefficient from branch sk to branch lr, where s, k, l, r = 1…n s n represents a node within the power transmission network. s This represents the total number of nodes within the power transmission network.

[0097] The safety constraints of various operating equipment in the power transmission network are shown in equation (26).

[0098] (26); In the formula: This represents the output power of unit a within the transmission network. , Indicates the maximum and minimum output of unit a. , This indicates the standby capacity of unit a during primary frequency regulation. , These represent the real-time SOC (State of Charge) status and output power of the centralized energy storage b within the transmission network, respectively. , These represent the maximum and minimum SOC states of energy storage b, respectively. , These represent the maximum and minimum charging power of energy storage b, respectively.

[0099] The power grid regulating power supply ramping constraint is shown in equation (27).

[0100] (27); In the formula: This indicates the ramp-up rate of generating units within the power transmission network. Indicates the unit's assembly capacity. , These represent the changes in load power and renewable energy power at time t, respectively. This indicates the maximum available energy storage capacity within the transmission network. This indicates that a backup power grid safety measure has been reserved.

[0101] Similar to equations (10) and (23), the optimization function at the power transmission network level is established as shown in equation (28).

[0102] (28); In the formula: and These represent the controllable power generation cost and energy storage charging and discharging cost at node s within the transmission network, respectively. This represents the blocking price coefficient. , and These represent the actual power curtailment, curtailment risk factor, and curtailment risk power of new energy sources within the transmission network, respectively. , and As a weighting factor; , and These represent the active power output of the controllable power source at node s of the transmission network at time t, the active power output of the centralized energy storage system at node s of the transmission network at time t, and the active power output of the new energy power source at node s of the transmission network at time t, respectively.

[0103] The robust optimization method based on time-series state deduction can obtain the optimized scheduling results at the power grid level, as shown in Equation (29).

[0104] (29); In the formula: This represents the set of schedulable ranges of node s within the transmission network. and Let represent the feasible solutions for active power and reactive power of node s in the transmission network, respectively.

[0105] 2. Coordinated optimization method for transmission-distribution-micro multi-level power grids:

[0106] Based on the distributed optimization results of microgrids, distribution networks and transmission networks obtained above, this embodiment presents an iterative optimization method for a multi-level transmission-distribution-microgrid based on alternating reconfiguration of schedulable domains, to obtain a schedulable domain for the transmission-distribution-microgrid system that satisfies the security constraints of the multi-level power grid.

[0107] Within the transmission network level, the dispatchable domain of each node is shown in Equation (29), which includes distribution network nodes, adjustable power source nodes, and centralized energy storage nodes. Within the distribution network level, the dispatchable range obtained after optimization is shown in Equation (23). Therefore, by reconstructing the intersection of the two, the dispatchable domain of the distribution network that satisfies the optimized dispatching result of the transmission network can be obtained, as shown in Equation (30).

[0108] (30); In the formula: This represents the schedulable domain of the updated distribution network j.

[0109] The result of formula (30) has two possibilities, namely or .like This indicates that within the distribution network level, there exists a solution in the optimized dispatchable domain that satisfies the optimal dispatch at the transmission network level, therefore it is unnecessary to update the optimal dispatch results at the distribution network level.

[0110] However, if If the optimal scheduling solution for the distribution network does not exist within the optimized dispatchable domain at the distribution network level, then the optimal scheduling result for the distribution network needs to be readjusted. That is, after supplementing Equation (22) with the constraints shown in Equation (31), the optimal scheduling result for the distribution network j after iteration is obtained through re-optimization calculation.

[0111] (31); In the formula: This indicates the updated optimized scheduling results of the distribution network.

[0112] After updating the optimal scheduling results of all distribution networks, the optimal scheduling results based on distribution network j are then used. The dispatchable domain of each microgrid is verified. Nodes within the distribution network include microgrid nodes, controllable power source nodes, and centralized energy storage nodes, based on... Extract the schedulable domain of microgrid nodes and connect it with... Perform intersection reconstruction, as shown in equation (32).

[0113] (32); Similarly, if This indicates that within the microgrid level, there exists a solution within the optimized dispatchable domain that satisfies the optimal dispatching requirements of the distribution network level. However, if... If so, the microgrid i needs to be recalculated and optimized.

[0114] After supplementing Equation (10) with the constraints shown in Equation (33), the optimized scheduling result of microgrid i is obtained by re-optimizing the calculation.

[0115] (33); It is worth noting that, in ~ During the alternating reconstruction of schedulable domains within the domain, if If the reconfigurable domain reconfiguration results of each level of the power grid within ~t are not empty sets, and an empty set result only appears at time t, then it is only necessary to reconfigure the reconfigurable domain reconfiguration results at time t. A recalculation can then be performed to optimize the process. In summary, the complete process of the multi-level collaborative optimization scheduling method for transmission-distribution-microgrids can be obtained, such as... Figure 1 As shown.

[0116] 3. Case Analysis:

[0117] To verify the effectiveness of the proposed distributed collaborative optimization scheduling method for multi-level power grids (transmission-distribution-microgrid), based on... Figure 2 A case study analysis is conducted on the transmission-distribution-micro multi-level power grid system shown. The transmission network adopts the IEEE-14 node system, which includes 3 synchronous generators, 3 photovoltaic power stations, 3 wind farms, 2 energy storage power stations, and 3 distribution networks.

[0118] Each distribution network adopts the IEEE-33 node system. Each distribution network has one synchronous motor node, one wind farm node, one photovoltaic power station node, and one energy storage power station node. In addition, three microgrid nodes are set up, and the remaining nodes are all set up as load nodes.

[0119] The simulation examples are programmed based on the MATLAB-2022 environment and the Cplex tool is used to calculate and solve the model. The hardware environment for running the simulation examples is a computer with an Intel Core i9-14900 / F CPU, 5.80GHz, and 64GB of memory.

[0120] In the optimization scheduling problem, the unit time is taken as =1h, optimize the target time period ~ =24h, , These correspond to 00:00-24:00, which is the start and end time of a natural day.

[0121] (1) Optimized scheduling results at the microgrid level:

[0122] The training dataset consists of new energy-related data nodes with a time scale of 1 hour within one year. The data categories are classified based on the fuzzy clustering algorithm, and the training set is divided into days with sufficient sunshine, cloudy days, windy days, and light wind days. Power prediction is then performed based on the spatiotemporal graph convolutional network model.

[0123] Based on this, the risk range of wind and solar forecast deviations was characterized using probability density functions, and the results are shown in Figures 3(a)-(d). Furthermore, Table 1 provides quantitative evaluation data of the forecast results based on normalized root mean square error (NRMSE) and mean absolute percentage error (MAPE).

[0124] Figures 3(a) and (b) show the predicted risk range of photovoltaic power on sunny days and cloudy days, respectively. It can be seen that the actual photovoltaic power curve is almost entirely contained within the predicted risk range in both scenarios. This demonstrates the accuracy of the photovoltaic risk range characterization based on the probability density function. Furthermore, comparing Figures 3(a) and (b), it can be seen that the predicted risk range of photovoltaic power is smaller on sunny days, while the risk range is larger on cloudy days. Moreover, during certain periods on cloudy days, the actual photovoltaic power even approaches or slightly exceeds the predicted risk range. This is because photovoltaic output during cloudy periods exhibits greater randomness and uncertainty.

[0125] Figures 3(c) and (d) show the predicted risk range of wind power on windy and light wind days, respectively. The results also demonstrate the accuracy of the probability density function-based characterization of the wind power risk range. It is easy to see that the predicted risk range of wind power is the opposite of that of photovoltaics; that is, the risk range is larger during windy periods and smaller during light wind periods. Moreover, the actual power during windy periods is even close to or slightly exceeds the predicted risk range. This is because wind power output has stronger randomness and uncertainty on windy days.

[0126] Table 1. Prediction Accuracy Data Indicators .

[0127] The optimized scheduling results at the microgrid level are shown in Figures 4(a) and (b). Figure 4(a) shows the optimized scheduling results when the renewable energy power is relatively low. From 00:00 to 08:00, the renewable energy output in the microgrid is less than the load, and the microgrid receives power from the distribution network system. From 09:00 onwards, the photovoltaic output increases rapidly, and the renewable energy output exceeds the load. At this time, the optimized result is energy storage charging, until 13:00 when the renewable energy power in the microgrid equals the power consumption, at which point charging stops. By 17:00, renewable energy generation declines rapidly, and energy storage begins to discharge. Especially after 19:00, wind resources further decline, and the economic optimization result arranges controllable load reduction until 24:00.

[0128] Figure 4(b) shows the optimized scheduling results of microgrid 2 within distribution network 1, obtained based on the distributed collaborative optimization scheduling method of transmission-distribution-microgrid multi-level power grid, during periods of high renewable energy generation. From 00:00 to 04:00, renewable energy generation exceeds the load, and energy storage begins charging. From 04:00 to 08:00, renewable energy generation is less than the load, and the economic optimization results do not schedule energy storage discharge. From 08:00 to 15:00, photovoltaic power increases rapidly, and energy storage continues charging until 14:00 when it reaches full capacity. From 14:00 to 15:00, energy storage lacks charging regulation capabilities. From 16:00 onwards, renewable energy output is less than the load, at which point energy storage discharges until 22:00, when the load rapidly decreases, energy storage regains charging capability and continues charging, achieving renewable energy absorption within the microgrid.

[0129] Figures 5(a) and (b) illustrate the power exchange between the microgrid and the distribution network, corresponding to Figures 4(a) and (b), and the dispatchable domain of the microgrid. Taking power reception from the distribution network as the positive direction, the dispatchable range in the positive direction originates from energy storage charging, while the dispatchable range in the negative direction originates from energy storage charging and controllable load shutdowns. Combining Figures 5(a) and (b) with the operating states of energy storage and controllable loads in Figures 4(a) and (b), it can be seen that when energy storage is charging, the forward dispatchable range of the microgrid will decrease, while the reverse dispatchable range will increase. This is due to insufficient power reception capacity from the distribution network caused by the consumption of energy storage charging capacity. Similarly, when energy storage discharges or controllable loads shut down, the forward dispatchable range of the microgrid will decrease, while the reverse dispatchable range will increase. In particular, when energy storage is charging or discharging at full power or when all controllable loads are reduced, the dispatchable domain of the microgrid will be compressed into a smaller space, at which point the microgrid's controllability will decrease. Changes in the dispatchable domain can clearly characterize the power regulation capability of a microgrid at different times, providing effective support for optimizing the interaction between the microgrid and the distribution network.

[0130] (2) Verification and optimization scheduling results of the power flow model for the upgraded power distribution network:

[0131] Using one year of power flow data from the IEEE-33-node distribution network model as training data samples, an equivalent power flow model of the distribution network in a high-dimensional linear space was obtained by training based on the up-dimensional mapping function. Table 2 shows the accuracy of the node voltages calculated based on the linearized equivalent model in different up-dimensional spaces, as well as the computation time of each model.

[0132] Table 2. Node voltage error and computation rate in different dimensional spaces. .

[0133] It can be seen that as the dimension of the mapping relationship increases, the accuracy of the equivalent linearized model gradually improves, but the required computation time also gradually increases. After the 54th-order dimensionality upgrade, the voltage error of the distribution network nodes calculated by the linearized equivalent model is controlled at around 1%, and its accuracy is sufficient for power flow calculation and analysis of the distribution network. However, its computation time is significantly faster than that of the traditional model; the computation time of the 54th-order dimensionality upgrade model is only about 1 / 10 of that of the complete power flow model.

[0134] Figures 6(a)-(d) show a comparison of the node voltage results calculated based on the equivalent power flow model of the distribution network. Figures 6(a) and (b) show the voltage curves of node 4 (microgrid 1 - residential load) within distribution network 1 under the 30th and 54th order upgraded equivalent models, respectively. It can be seen that the 30th order equivalent model has some deviations during the early morning and midday periods, while the 54th order model can match the actual voltage much better. Figures 6(c) and (d) show the voltage curves of node 33 (microgrid 3 - industrial park) within distribution network 1 under the 30th and 54th order upgraded equivalent models, respectively. The voltage curve obtained by the 54th order model also shows good approximation to the actual voltage. These calculation results verify the effectiveness of the proposed method in improving the calculation rate of power flow in distribution networks.

[0135] Figures 7(a) and (b) show the optimized scheduling results and dispatchable domain of distribution network 1. As shown in Figure 7(a), during the period of abundant wind and solar resources from 00:00 to 15:00, the combined output of new energy sources and the minimum output limit of adjustable units within the network is less than the sum of the power consumption of the network load. The new energy consumption gap is relatively large from 03:00 to 09:00, and the economic optimization result calls for charging of centralized energy storage stations within the distribution network. From 11:00 to 13:00, no power curtailment occurred within the distribution network, and energy storage was also charged. This is the result of coordinated optimized scheduling with the transmission network. Figure 8 The optimization results shown can be verified and will be discussed in detail in the next subsection. After 16:00, the output of photovoltaic power decreases, the power consumption of the grid load is greater than the lower limit of the output of new energy and the output of adjustable units, and the adjustable thermal power increases the power generation. At the same time, during the period of large gap from 18:00 to 23:00, the energy storage station discharges.

[0136] Figure 7(b) shows the exchange power between the distribution network and the transmission network, and the dispatchable domain of the distribution network, corresponding to the optimized scheduling results in Figure 7(a). Taking the distribution network receiving power from the transmission network as the positive direction, when the energy storage is charging, the forward dispatchable range of the microgrid will decrease, while the reverse dispatchable range will increase. Conversely, when the energy storage discharges or the adjustable thermal power increases its power, its reverse dispatchable domain will decrease, while its forward dispatchable domain will increase. The above optimization results verify the effectiveness of the proposed method.

[0137] (3) Results of optimized dispatching of the power transmission network:

[0138] Figure 8 The optimized scheduling results of the power transmission network are presented. As the highest level of the power grid, the power transmission network needs to maintain a balance between power generation and consumption at all times. Figure 8 It can be seen that between 03:00 and 04:00, the power generation of the power grid is greater than the power consumption. At this time, the energy storage devices in the transmission network are charging. At the same time, due to the micro-coordinated optimization scheduling of transmission and distribution, the power generation and consumption guidance is transmitted to the distribution network, so the energy storage in the distribution network also begins to charge.

[0139] From 05:00 to 07:00, in order to absorb new energy, the output of adjustable power sources in the transmission network is kept at the minimum. From 08:00 to 09:00, although the load is slightly higher than the minimum generating capacity of the transmission network, its price orientation is not obvious. However, in order to absorb new energy during the noon period, energy storage is still discharging. This is because the time-series recursive optimization scheduling method can predict in advance the grid safety constraints exceeding the limit and the economic decline event in the subsequent period, and take measures to adjust in advance to meet the optimality of the optimization result throughout the entire period.

[0140] Between 11:00 and 13:00, the minimum generating capacity of the transmission network exceeds the sum of the load demand. Even with all energy storage fully charged, the demand from renewable energy sources cannot be met. Based on micro-coordinated optimization scheduling of transmission and distribution, the power generation and consumption guidance is transmitted to the distribution network, corresponding to... Figure 8 The charging behavior of distribution network 1 shown is observed from 11:00 to 13:00. From 16:00 to 19:00, renewable energy sources decrease, and energy storage systems discharge. These results verify the effectiveness of the proposed method for the coordinated optimization of transmission and distribution micro-multi-level power grids.

[0141] (4) Comparison of computational performance of distributed algorithms:

[0142] The simulation comparison was conducted based on the following four methods to optimize the scheduling of the multi-level power transmission and distribution network system within 2 hours. The calculation results and calculation rate were analyzed. The specific comparison results are shown in Table 3.

[0143] Method 1: Centralized optimization scheduling method;

[0144] Method 2: Independent Optimization Scheduling Method;

[0145] Method 3: Distributed optimization scheduling method based on alternating multipliers;

[0146] Method 4: The optimized scheduling algorithm proposed in this invention.

[0147] Table 3 Performance Comparison of Different Algorithms .

[0148] As shown in Table 3, the centralized optimization scheduling method yields the lowest renewable energy curtailment rates in both transmission and distribution networks, at 5.6891% and 6.2446% respectively, but it is extremely time-consuming. While the independent optimization scheduling method offers the fastest computation speed, its renewable energy curtailment rate is very high. Method 3 performs worse than the proposed method in both computation time and renewable energy curtailment rate. The proposed method can complete multi-level system collaborative optimization in a relatively short time, and its renewable energy curtailment rate is close to that of the centralized algorithm. It is more economical and better facilitates renewable energy absorption. These results verify the effectiveness of the proposed method in improving photovoltaic absorption in multi-level transmission and distribution networks.

[0149] 4. Conclusion:

[0150] This invention proposes a distributed collaborative optimization scheduling method for a multi-level power grid (transmission-distribution-microgrid) under high-proportion renewable energy access. This method considers the uncertainties of renewable energy generation and utilizes a robust optimization method based on time-series state deduction to achieve optimized scheduling of complex nonlinear power systems with multiple levels and scheduling objects. Based on numerical examples, the following conclusions can be drawn:

[0151] 1) The robust optimization scheduling method based on time-series state deduction proposed in this invention can ensure that the obtained optimal scheduling solution is closer to the optimal economic efficiency and new energy consumption under the premise of ensuring safety constraints throughout the entire target time period.

[0152] 2) The high-dimensional linearized equivalent power flow model of the distribution network based on the up-dimensional mapping function can be trained based on historical power flow data and accurately reproduce the linearized equation of the power flow of the distribution network. While ensuring the calculation accuracy, it significantly improves the power flow calculation speed of multi-scheduling objects and strongly nonlinear distribution networks.

[0153] 3) The established multi-level distributed collaborative optimization scheduling method for power grids can take into account the operating characteristics and security constraints of each level of the power grid, realize efficient optimization in a hierarchical and partitioned manner, and iterative correction in a collaborative manner between transmission, distribution and micro-grids, significantly improve computational efficiency, and avoid the dimensionality curse caused by global optimization.

[0154] The proposed method achieves better economic efficiency and renewable energy integration compared to the optimized scheduling results obtained from centralized global algorithms, while saving nearly 90% of computation time. This effectively addresses the problem of the explosive growth in computational load for optimized scheduling under new power systems. With the increasing proportion of renewable energy and power electronics in power systems, the challenges faced by optimized scheduling and photovoltaic energy integration will become even more severe, and the proposed strategy will play an even more significant role.

Claims

1. A distributed collaborative optimization scheduling method for a multi-level power grid (transmission-distribution-microgrid), characterized in that, Includes the following steps: Step 1: Establish a distributed optimization model for the microgrid to obtain the dispatchable domain of the microgrid for new energy consumption. Then, use a robust optimization method based on time-series state deduction to correct the dispatchable domain of the microgrid and obtain the distributed optimization scheduling results at the microgrid level. Step 2: Establish a distributed optimization model for the distribution network, obtain the dispatchable domain of the distribution network based on the dimension-upgrading mapping algorithm in high-dimensional linear space, and obtain the distributed optimization scheduling results at the distribution network level based on the robust optimization method of time-series state deduction. Step 3: Establish a distributed optimization model for the power transmission network to obtain the dispatchable domain of the power transmission network. Based on the robust optimization method of time-series state deduction, the distributed optimization scheduling results at the power transmission network level can be obtained. Step 4: Based on the distributed optimization scheduling results of microgrids, distribution networks and transmission networks obtained in Steps 1-3, an iterative optimization method based on alternating reconfiguration of schedulable domains for transmission-distribution-micro multi-level power grids is adopted to obtain the schedulable domains of the transmission-distribution-micro system that satisfy the security constraints of the multi-level power grid.

2. The distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids according to claim 1, characterized in that, In step 1, the microgrid optimization model is established as follows: In a microgrid system, controllable objects include controllable power sources, energy storage, and demand response loads. The power of each controllable object is integrated as a control variable for the microgrid, as shown below: （1）； In the formula: superscript This indicates that the current electrical quantity belongs to the i-th microgrid. Indicates the number of microgrids within the same distribution network; This represents the control variable vector of microgrid i. , and These represent the controllable load power, controllable power supply power, and energy storage charging and discharging power within the microgrid, respectively, with t representing time. The inherent loads that cannot be adjusted within the microgrid, the power generation capacity of new energy sources, and the remaining energy storage capacity determined by the energy storage operation status are considered as state variables of the microgrid, and are organized into a vector form as shown below: （2）； In the formula: This represents the state variable vector of microgrid i. , and These represent the unadjustable power load forecast, renewable energy power forecast, and energy storage remaining power state vector of microgrid i, respectively. Energy storage remaining power state vector It is obtained by time-domain integration of energy storage charging and discharging, as shown below: （3）； In the formula: This represents the initial remaining power of the energy storage device in microgrid i. Indicates the initial time; and These represent the charging and discharging efficiencies of the energy storage device, respectively. , They represent the target time periods respectively. ~ Inner indivual The charging and discharging power of energy storage during the time period, For the target time period ~ The unit duration within; T represents the target time period. ~ Number of time periods within; Constraints within a microgrid include equipment operating power constraints and energy storage remaining power state constraints; The power constraints for equipment operation include constraints on new energy power generation, controllable load output, controllable power supply, energy storage charging and discharging power, and the exchange power constraints between the microgrid and the distribution network, as shown in the following expressions: （4）； In the formula: Indicates the rated power of new energy sources. and These are the maximum and minimum adjustable power of the controllable load, respectively; and Indicates the maximum and minimum generating power of a controllable power source; and Indicates the upper and lower limits of energy storage charging and discharging power; and These represent the exchange power between microgrid i and its distribution network, and the maximum allowable power for interaction, respectively. The power constraints for the remaining energy storage state are shown below: （5）； In the formula: and These represent the maximum and minimum remaining energy storage states within microgrid i, respectively; The operating cost functions for various resources within a microgrid are shown below: （6）； In the formula: , and These represent the cost of controllable power generation, the cost of energy storage charging and discharging, and the cost of controllable load dispatching, respectively. This represents the operating cost of a controllable power supply as a function of its operating power. This indicates the basic operating cost of a controllable power supply. This represents the unit charge / discharge cost of the energy storage device after conversion. This indicates the unit compensation amount for demand response load; For the first in the microgrid indivual Controllable power supply during different time periods; For the first in the microgrid indivual Energy storage charging and discharging power during a given time period; For the first in the microgrid indivual The change in controllable load power over a given period of time; The prediction bias of new energy sources can be characterized by the probability density function as shown below: （7）； （8）； In the formula: and Let represent the probability density functions of photovoltaic and wind power, respectively; and It is the shape parameter of the beta distribution, obtained from historical data statistics of light resources; and These represent the actual and maximum values ​​of photovoltaic power generation, respectively. represents the gamma function; c and d are the shape parameters of the Weiber distribution, obtained from historical wind resource data statistics; and These represent the cut-in wind speed and the rated wind speed of the fan, respectively. and These represent the fan output power and the rotor rated power, respectively. d is the power of the Weiber distribution scaling parameter c; Cut off the wind speed for the fan; Based on equations (7) and (8), the risk power of new energy curtailment is obtained as follows: （9）； In the formula: Indicates the power at risk of curtailment of renewable energy. and These represent the upper limits of the predicted photovoltaic power and wind power, respectively.

3. The distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids according to claim 2, characterized in that, In step 1, the robust optimization method based on time-series state deduction is as follows: Step (a): Taking the most economical operation of the microgrid as the objective, while also considering the weight of renewable energy curtailment, and using the curtailment risk power caused by renewable energy prediction deviation as a penalty term, and taking into account power and remaining energy storage capacity constraints, the optimal scheduling function of microgrid i is obtained, as shown below: （10）； In the formula: For time-of-use electricity pricing, Indicates the microgrid i Actual curtailment of renewable energy during the specified time period The risk factor representing the power at risk of curtailment of renewable energy within microgrid i; , and As a weighting factor; , and They represent Costs of time-controlled power generation, energy storage charging and discharging, and controllable load dispatching; express The exchange power between microgrid i and its distribution network during the time period; This is the penalty coefficient for the risk of renewable energy curtailment at the microgrid layer; For microgrids during time periods The risk of curtailment of renewable energy power; Step (b): Simulate the operating status of the microgrid during the target time period; based on The initial state variables of the microgrid at time t are derived from the state equations. ~ The microgrid state variables at each moment within the time period are obtained, and combined with the objective function shown in Equation (10), the microgrid control variables that satisfy the safety constraints within each unit of time are optimized to form the microgrid dispatchable domain; the time-series state derivation formula of the microgrid is shown below: （11）； In the formula: , , Let represent the state matrix, control matrix, and input matrix of the uncertain disturbance of microgrid i, respectively; This represents the disturbance input vector of microgrid i; Step (c): Correct the obtained microgrid dispatchable domain; A robustness check inequality constraint set is established, and the safety constraint check is performed on the dispatchable domain of the microgrid obtained by equations (10) and (11). At the same time, the degree of limit violation within a tolerable unit time period is taken into account, and the time-domain cumulative robustness constraint discrimination index is calculated. The specific calculation formula is as follows: （12）； In the formula: ν represents the solution result of the linear programming and serves as a criterion. When ν > 0, it indicates that the microgrid state variables at time t do not meet the safety constraints. Indicates the state variables of microgrid i in The sensitivity index matrix of the time period to the boundary conditions of the constraint; Denotes the relaxation vector of microgrid i. The vector representing the tolerable level of exceedance at time t is determined by the equipment parameters; Step (d): Update ~ The modified dispatchable domain of microgrid i within the time period; After removing the scheduling results that do not meet the safety constraints calculated based on equation (8), a new schedule is generated based on the constraint set. ~ The dispatchable boundary of microgrid i within a time period is determined, and then the modified dispatchable domain that satisfies the requirements of microgrid i throughout the entire time period is obtained. The specific calculation formula is as follows: （13）； In the formula: , and Let i represent the original feasible region, the modified feasible region, and the newly generated feasible region, respectively. Vector of degree of exceeding limits transpose; This is the constraint threshold; Then return to step (b) to perform state deduction for the next moment until completion. ~ The calculation of all time periods is performed to obtain the complete dispatchable domain of the microgrid i. Step (e): Based on the obtained ~ Within the time period, the complete dispatchable domain of microgrid i is calculated. ~ The exchange power between microgrid i and its distribution network may have a solution. Based on the enumeration method, and using the power equation and the active and reactive power constraints of the equipment operation, the corresponding distribution network node is calculated. ~ The available range of active and reactive power within a given time period is used as an element to form a set of available ranges for the distribution network nodes, as shown in the following expression: （14）； In the formula: This represents the set of dispatchable ranges for the aforementioned distribution network nodes; To represent the reactive power exchange between microgrid i and its distribution network; subscript for Or N, the corresponding equipment categories are non-adjustable power load, controllable load, controllable power supply, energy storage and new energy; Indicates the type of equipment within the microgrid at time t. The corresponding active power; Indicates the type of equipment within the microgrid at time t. The corresponding reactive power; To represent the equipment categories within the M layer of a microgrid The reactive-active power coupling function.

4. The distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids according to claim 3, characterized in that, Step 2 is as follows: Step 2.1: Based on the principle of dimension-upgrading mapping, for any nonlinear multivariate function, by establishing a dimension-upgrading mapping relationship, we can find the corresponding linear multivariate function in the high-dimensional space, as shown below: （15）； In the formula: Represents a nonlinear nth-order function; H(x) represents the g-order independent variable, where g = 1…n; H(x) is N. f The dimension mapping function should be flexibly selected based on the properties of the computational object. For the Nth f Dimensional mapping function; It is a higher-dimensional mapping matrix; The dependent variable; and As the independent variable; For the set of independent variables in N f Linearized representation of vectors in an upgraded dimensional space; Step 2.2: Use the historical actual values ​​of power flow in the distribution network as the training set; upgrade the mapping matrix. The weights are considered as the weights of the dimensionality-up mapping from each independent variable to the dependent variable. The weight values ​​of each item are solved using the least squares method based on historical data, with the minimum power flow regression error as the benchmark. The specific expressions are as follows: （16）； In the formula: Represents the Moore-Penrose generalized inverse matrix; As the independent variable; Selection based on independent variables with basis vectors The logarithmic series of the Euclidean distance weights is used to construct the dimension-upgrading mapping function, the specific expression of which is shown below: （17）； In the formula: Represents the basis vectors for increasing dimensionality; Step 2.3: In the distribution network dimensionality upgrade mapping, establish the independent variable To dependent variable The mapping relationship is as follows: （18）； In the formula: , and These represent the active power, reactive power, and voltage amplitude of a microgrid node, respectively. , , These represent the active power, reactive power, and voltage amplitude at the power plant nodes, respectively. Based on the upgraded mapping relationship established by equations (16) to (18), historical data is used for training to obtain the upgraded linearized power flow equations of the distribution network, as shown below: （19）； In the formula: superscript This indicates that the current electrical quantity belongs to the j-th distribution network, where j=1…n p n p The number of distribution networks; , , and Let represent the node voltage vector, node injected active power vector, node injected reactive power vector, and other high-dimensional linear variables of distribution network j, respectively. , and These represent the sensitivity matrices of active power, reactive power, and high-dimensional linear variables to node voltage within distribution network j, respectively. Step 2.4: Based on the upper and lower bounds of the voltage at each node in distribution network j, the power relationship equations for each node in distribution network j are characterized as follows: （20）； In the formula: and Let represent the upper and lower bounds of the voltage constraint at node p within the distribution network j, respectively. , , These represent the sensitivity coefficients of the active power, reactive power, and other high-dimensional linear variables of the m-th node in the distribution network j to the voltage amplitude of the p-th node, respectively. , and The active power, reactive power, and other high-dimensional linear variables of the m-th node in the distribution network j are respectively represented by the superscript "~" and the subscript "~". The superscript "~" and the subscript "~" respectively represent the node variables corresponding to the voltage upper and lower bounds of the p-th node in the distribution network. The safety constraints that the remaining node regulating equipment in the distribution network should meet are as follows: （21）； In the formula: and These represent the renewable energy generation power and rated power of the p-th node within the distribution network j, respectively. , and Let represent the actual, maximum, and minimum power of the controllable power source at node p within the distribution network j, respectively. , and Let these represent the actual, upper limit, and lower limit power of the energy storage at node p within the distribution network j, respectively. and These represent the exchange power between distribution network j and its transmission network, and the maximum allowable exchange power, respectively. , and These represent the real-time, upper limit, and lower limit energy storage capacity of the p-th node within the j-th distribution network, respectively. Step 2.5: Taking the most economical operation of the distribution network as the objective, while also considering the weight of renewable energy curtailment, and using the curtailment risk power caused by renewable energy prediction deviation as a penalty term, the optimization function of distribution network j is obtained, as shown below: （22）； In the formula: and They represent time periods respectively. The power generation cost and energy storage charging and discharging cost of the controllable power source at node p within the distribution network j; , and These represent the actual power curtailment, curtailment risk factor, and curtailment risk power of new energy sources within distribution network j, respectively. , and As a weighting factor; For the distribution network j in time period The exchange power between the internal power supply and the local power grid; The equivalent active power exchanged between distribution network node j and its connected microgrid i at time t; A robust optimization method based on time-series state deduction obtains the dispatchable domain of the distribution network, as shown in the following equation: （23）； In the formula: Represents the set of dispatchable ranges for distribution network j; Let be the equivalent reactive power exchange between distribution network node j and the upstream transmission network at time t.

5. The distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids according to claim 4, characterized in that, Step 3 is as follows: Step 3.1: The power flow equations of the transmission network are expressed as follows: （24）； In the formula: , and These represent the injected active power, reactive power, and AC voltage at node s in the transmission network, respectively. This indicates that transmission network node s is adjacent to transmission network node k; This represents the AC voltage at node k in the transmission network; , , , These represent the real part of the nodal admittance matrix element, the nodal phase angle difference, the imaginary part of the nodal admittance matrix element, and the parallel s s of the transmission network node, respectively. Step 3.2: The power flow constraints of the transmission network are expressed as follows: （25）； In the formula: and These are the power flow and thermal stability limits of the SK branch of the transmission network, respectively. This represents the power flow transfer coefficient from branch sk to branch lr, where s, k, l, r = 1…n s n represents a node within the power transmission network. s This represents the total number of nodes within the transmission network. The safety constraints of various operating equipment within the power transmission network are expressed as follows: （26）； In the formula: This represents the output power of unit a within the transmission network. and Indicates the maximum and minimum output of unit a; and This indicates the standby capacity of unit a during primary frequency regulation; and These represent the real-time SOC status and output power of centralized energy storage b within the transmission network, respectively. and These represent the maximum and minimum SOC states of energy storage b, respectively. and These are the maximum and minimum charging power of energy storage b, respectively; The power grid regulating power source ramping constraint is expressed as: （27）； In the formula: This indicates the ramp-up rate of generating units within the power transmission network. Indicates the unit's assembly capacity; and These represent the changes in load power and renewable energy power at time t, respectively. This indicates the maximum available energy storage power within the transmission network; This indicates that a backup power grid safety measure has been reserved. Step 3.3: Establish the optimization function at the transmission network level as shown in the following equation: （28）； In the formula: and These represent the controllable power generation cost and energy storage charging and discharging cost at node s within the transmission network, respectively. This represents the blocking price coefficient. , and These represent the actual power curtailment, curtailment risk factor, and curtailment risk power of new energy sources within the transmission network, respectively. , and As a weighting factor; , and These represent the active power output of the controllable power source at node s of the transmission network at time t, the active power output of the centralized energy storage system at node s of the transmission network at time t, and the active power output of the new energy power source at node s of the transmission network at time t, respectively. The robust optimization method based on time-series state deduction can obtain the optimized scheduling results at the transmission network level, as shown in equation (29): （29）； In the formula: This represents the set of schedulable ranges of node s within the transmission network. and Let represent the feasible solutions for active power and reactive power of node s in the transmission network, respectively.

6. The distributed collaborative optimization scheduling method for transmission-distribution-micro multi-level power grids according to claim 5, characterized in that, Step 4 specifically includes: Step 4.1: Reconstruct the intersection of the dispatchable domains of each node in the transmission network level shown in equation (29) and the optimized dispatchable range in the distribution network level shown in equation (23) to obtain the dispatchable domain of the distribution network that satisfies the optimized dispatching results of the transmission network, as shown below: （30）； In the formula: This represents the schedulable domain of the updated distribution network j; The result of formula (30) has two possibilities, namely or ; An empty set indicates that at this moment, after reconstructing the intersection of the schedulable domains, there are no feasible solutions that satisfy the constraints. like There is no need to update the optimized scheduling results at the distribution network level; like Then, the results of the distribution network optimization scheduling are readjusted; that is, after supplementing Equation (22) with the constraints shown in Equation (31), the optimized scheduling results of distribution network j after iteration are obtained by re-optimizing the calculation: （31）； In the formula: This indicates the updated optimized dispatching results for the distribution network; After updating the optimal scheduling results of all distribution networks, the optimal scheduling results based on distribution network j are then used. Verify the dispatchable domain of each microgrid; The nodes within the distribution network include microgrid nodes, controllable power source nodes, and centralized energy storage nodes, based on Extract the dispatchable domain of microgrid nodes and combine it with the dispatchable range set of distribution network nodes. The intersection is reconstructed as shown in the following equation: （32）； Similarly, if This indicates that within the microgrid level, there exists a solution within the optimized dispatchable domain that satisfies the optimal dispatching requirements of the distribution network level. However, if... If so, the microgrid i needs to be recalculated and optimized. This indicates the optimized scheduling result of the updated distribution network nodes; After supplementing equation (10) with the constraints shown in equation (33), the optimized scheduling result for microgrid i is obtained by recalculating the optimization: （33）； exist ~ During the alternating reconstruction of schedulable domains within the domain, if If the reconfigurable domain reconfiguration results of each level of the power grid within ~t are not empty sets, and an empty set result only appears at time t, then it is only necessary to reconfigure the reconfigurable domain reconfiguration results at time t. A recalculation can be performed to optimize the calculation.

Images