Transmission and distribution micro-grid hierarchical scheduling method, device and equipment and storage medium

Abstract

The invention belongs to the electromechanical field, and discloses a transmission and distribution micro-grid hierarchical scheduling method, device and equipment and a storage medium, and the method comprises the steps: obtaining power grid data of a transmission and distribution micro-grid collaborative scheduling system, establishing a first optimal power flow problem between the power transmission network and the power distribution network and a second optimal power flow problem between the power distribution network and the micro-grid based on the acquired data; decomposing the second optimal power flow problem into a plurality of sub-problems; solving the first optimal power flow problem to obtain a first result, taking the first result as the input of a second optimal power flow problem, and solving the plurality of sub-problems according to a double-loop learning assisted asynchronous alternating direction multiplier method to obtain a second result; and scheduling the transmission and distribution micro-grid collaborative scheduling system based on the second result. A three-layer collaborative architecture comprising a power transmission network, a power distribution network and a micro-grid is designed, the optimal power flow problem among different power grids is solved, and the overall architecture of the power system is optimized.

Description

Technical Field

[0001] The present invention belongs to the field of electric power, relates to a dispatching method, and particularly relates to a hierarchical dispatching method, device, equipment and storage medium for a transmission and distribution microgrid. Background Art

[0002] With the continuous increase of energy consumption and environmental pressure, controlling power costs, promoting the transformation of the energy structure, and improving the utilization rate of renewable resources have become the consensus of the international community. Driven by green and low-carbon policies, a large number of distributed resources are connected to the distribution network. However, the large-scale access of intermittent and uncertain renewable energy sources such as wind energy and solar energy not only exacerbates the difficulty of power grid dispatching but also increases the safety and stability risks of the system. How to effectively utilize the collaborative potential of distributed resources and incorporate them effectively into the power grid operation process has become an important issue that has attracted much attention.

[0003] In this context, a microgrid integrating energy storage, new energy, and intelligent control strategies has emerged. A microgrid is a means capable of effectively managing distributed resources. From the perspective of the power grid management and operation architecture, a microgrid can be regarded as a group of units within the distribution network that have self-balancing capabilities and can interact with other parts of the distribution network in a two-way power mode. Currently, there have been many studies on the optimal operation of the microgrid itself, and at the same time, some studies have focused on the collaborative optimization of the distribution network and the microgrid.

[0004] With the gradual increase in the number of distributed resources and microgrids, distributed resources not only affect the distribution network but also begin to have an impact on the collaborative operation mode of the transmission and distribution network. Currently, most regions face the problem of insufficient regulating resources at the transmission network level, and the distributed resources located in the distribution network have the potential to provide support to the transmission network and can be regarded as a potential regulating resource. However, since the distributed resources are located in the microgrid, when the transmission network invokes its regulating potential, it must also ensure the operation requirements of the distribution network and the microgrid. Therefore, when formulating a dispatching plan, it is necessary to consider the safe operation requirements of the transmission network, the distribution network, and the microgrid simultaneously.

[0005] In this regard, a direct solution is to combine the models of the transmission grid, distribution grid, and microgrid, for example, upload them all to the transmission grid control center, and then the latter solves them to obtain the required scheduling plan. However, in many scenarios, the management entities of the transmission grid, distribution grid, and microgrid are different. Transmission system operators and distribution system operators are independent control entities with their own rules, strategies, and goals. Due to internal model data privacy considerations, they do not allow sharing internal information other than boundary node voltages and powers with each other. Therefore, a central scheduling framework that requires each system management entity to share all its information with the central control agency may not be suitable for the operation of the entire power system. Because of the non-interoperability of system information and the difference in decision-making entities, traditional centralized algorithms cannot be used to solve large-scale regional power grid problems. Therefore, many distributed algorithms have been widely applied to the collaborative optimization of regional power grids such as transmission-connected power grids and distribution grids with multiple microgrids. The management entities of the transmission grid, distribution grid, and microgrid only interact boundary information with each other using a distributed iterative method, continuously update the boundary information, and re-interact, so as to finally converge to a coordinated solution. Although this solution makes sense technically, at present, most distributed algorithms mainly target the two-layer collaborative architecture, while the transmission-distribution-microgrid collaboration is a three-layer structure, and many algorithms are difficult to directly apply. In addition, most distributed algorithms usually only have guaranteed convergence and optimality for convex optimization problems. How to handle the non-convex constraints related to AC power flow in transmission-distribution-microgrid collaboration is also a key factor in ensuring the calculation effect of distributed algorithms. It can be seen that for the optimal scheduling problem considering transmission-distribution-microgrid collaboration, while satisfying the information protection of each system itself, a reasonable distributed parallel algorithm needs to be sought to meet the calculation requirements for the jointly optimized scheduling model.

[0006] Compared with the distribution-microgrid and transmission-distribution collaboration problems, the transmission-distribution-microgrid collaborative optimization problem is more difficult. First, the interconnection structure is a three-layer structure, involving double collaborative mechanisms between transmission and distribution, and between distribution and microgrid, and it is necessary to ensure that non-boundary information of different power grids is confidential to each other. Second, since voltage constraints in the distribution grid and microgrid are currently the main limiting factors for their safe operation, it is necessary to use the ACOPF model to describe the safe operation constraints of the transmission-distribution-microgrid. However, a large number of non-convex ACOPF constraints make the design of distributed algorithms more difficult. Third, classic distributed algorithms such as ADMM usually only guarantee their convergence for convex optimization problems and are not directly applicable to non-convex complex three-layer optimization problems. Summary of the Invention

[0007] In view of this, the present invention discloses a hierarchical scheduling method, device, equipment, and storage medium for a transmission-distribution-microgrid, which can solve the deficiencies in the related art.

[0008] To achieve the above object, the technical solutions disclosed by the present invention are as follows: According to the first aspect of the present invention, a hierarchical scheduling method for a transmission and distribution microgrid is proposed and applied to a transmission and distribution microgrid collaborative scheduling system. The system includes a transmission grid, a distribution grid, and a microgrid. The method includes: Obtain the grid data of the transmission and distribution microgrid collaborative scheduling system, and establish a first optimal power flow problem between the transmission grid and the distribution grid and a second optimal power flow problem between the distribution grid and the microgrid based on the obtained data; Decompose the second optimal power flow problem into multiple sub-problems, and each sub-problem in the multiple sub-problems corresponds to a microgrid; Solve the first optimal power flow problem to obtain a first result, and use the first result as the input of the second optimal power flow problem. Solve the multiple sub-problems according to the asynchronous alternating direction multiplier method assisted by double-loop learning to obtain a second result; When the second result meets the convergence conditions of the first optimal power flow problem and the second optimal power flow problem, schedule the transmission grid, the distribution grid, and the microgrid of the transmission and distribution microgrid collaborative scheduling system based on the second result.

[0009] As a preferred solution, the step of solving the first optimal power flow problem to obtain a first result, using the first result as the input of the second optimal power flow problem, and solving the multiple sub-problems according to the asynchronous alternating direction multiplier method assisted by double-loop learning to obtain a second result includes: Solve the optimal power flow problem based on the polynomial semi-definite programming cutting and chordal sparse relaxation model.

[0010] As a preferred solution, the polynomial semi-definite programming cutting and chordal sparse relaxation model is expressed as:

[0011]

[0012] As a preferred solution, the step of solving the multiple sub-problems according to the asynchronous alternating direction multiplier method assisted by double-loop learning to obtain a second result includes: For each sub-problem, when the shared variable value of the adjacent sub-problem is not received, predict the shared variable value based on the regression technique, and perform iterative solution by penalizing the consistency constraint in the objective function based on the augmented Lagrangian relaxation method; Design the abnormal switch controller to read the local objective function value of the sub-problem and perform abnormal detection during the iterative process; When the result of the iterative solution passes the abnormal detection, determine the second result.

[0013] As a preferred solution, the prediction of the shared variable value based on the regression technique includes: Obtain the missing shared variable values in the previous two iterations and form a linear extrapolation term; Assume that the missing shared variable values jump in the extrapolation direction, define the momentum term as the inertia of the iteration trend of the shared variable value, and the momentum term is a correction term added to the extrapolation to correct the prediction direction and make it bounded.

[0014] As a preferred solution, the anomaly detection during the iteration process includes: Perform anomaly detection based on an online flow unsupervised ground anomaly learner.

[0015] As a preferred solution, the power grid data includes node data, line data, and network topology data; the node data includes voltage amplitude, injected active power, active power demand, reactive power demand, and reactive power provided by the generator, the line data includes active power flow, reactive power flow, voltage phase difference angle, resistance, reactance, and the network topology data is the connection relationship between each node and line.

[0016] According to the second aspect of the present invention, a hierarchical scheduling device for a transmission and distribution microgrid is proposed, which is applied to a transmission and distribution microgrid collaborative scheduling system. The system includes a transmission grid, a distribution grid, and a microgrid. The device includes: An acquisition unit: acquire the node data, line data, and network topology data of the transmission and distribution microgrid collaborative scheduling system, and establish a first optimal power flow problem between the transmission grid and the distribution grid and a second optimal power flow problem between the distribution grid and the microgrid based on the acquired data; A decomposition unit: decompose the optimal power flow problem between the distribution grid and the microgrid into multiple sub-problems, and each sub-problem in the multiple sub-problems corresponds to a microgrid; A solution unit: solve the first optimal power flow problem to obtain a first result, and use the first result as the input of the second optimal power flow problem, and solve the multiple sub-problems according to the asynchronous alternating direction multiplier method assisted by double-loop learning to obtain a second result; A scheduling unit: when the second result meets the convergence conditions of the first optimal power flow problem and the second optimal power flow problem, schedule the transmission grid, distribution grid, and microgrid of the transmission and distribution microgrid collaborative scheduling system based on the second result.

[0017] According to the third aspect of the present invention, an electronic device is proposed, including: A processor; A memory for storing processor-executable instructions; Among them, the processor realizes the steps of the method described in the first aspect by running the executable instructions.

[0018] According to the fourth aspect of the present invention, a computer-readable storage medium is provided, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method described in the first aspect are realized.

[0019] As can be seen from the above technical solutions, for the hierarchical scheduling method of the transmission and distribution microgrid disclosed in the present invention, on the one hand, a three-layer collaborative architecture including the transmission grid, the distribution grid, and the microgrid is designed, and the optimal power flow problem between different power grids is solved, optimizing the overall architecture of the power system; on the other hand, the optimal power flow problem between the distribution grid and the microgrid is solved in parallel by the asynchronous alternating direction multiplier method assisted by double-loop learning, which not only improves the calculation efficiency, but also increases the accuracy of the calculation results, thereby enhancing the accuracy of power system scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is an architecture diagram of a transmission and distribution microgrid collaborative scheduling system provided by an exemplary embodiment; Figure 2 is a flowchart of a hierarchical scheduling method for a transmission and distribution microgrid provided by an exemplary embodiment; Figure 3 is a schematic diagram of a 4-node loop network using chord sparsity provided by an exemplary embodiment; Figure 4 is a schematic diagram of transmission and distribution grid collaboration and distribution and microgrid collaboration provided by an exemplary embodiment; Figure 5 is a schematic diagram of an asynchronous alternating direction multiplier method assisted by double-loop learning provided by an exemplary embodiment; Figure 6 is a schematic structural diagram of a device provided by an exemplary embodiment; Figure 7 is a block diagram of a hierarchical scheduling device for a transmission and distribution microgrid provided by an exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of the present invention as detailed in the appended claims.

[0022] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily executed in the order shown and described in the present invention. In some other embodiments, the steps included in the method may be more or less than those described in the present invention. In addition, a single step described in the present invention may be decomposed into multiple steps for description in other embodiments; and multiple steps described in the present invention may also be combined into a single step for description in other embodiments.

[0023] To further illustrate the present invention, the following embodiments are provided: In the related art, there is little research considering the coordination of transmission, distribution, and microgrids. Compared with the problems of distribution-microgrid and transmission-distribution coordination, the problem of transmission-distribution-microgrid coordinated optimization is more difficult. First, the interconnected structure is a three-layer structure, involving dual coordination mechanisms between transmission and distribution and between distribution and microgrids, and it is necessary to ensure the confidentiality of non-boundary information of different power grids from each other. Second, since voltage constraints in the distribution network and microgrid are currently the main limiting factors for their safe operation, it is necessary to use the ACOPF model to describe the safe operation constraints of the transmission-distribution-microgrid. However, a large number of non-convex ACOPF constraints make the design of distributed algorithms more difficult. Third, classic distributed algorithms such as ADMM usually only guarantee their convergence for convex optimization problems and are not directly applicable to non-convex complex three-layer optimization problems.

[0024] To solve the deficiencies in the related art, the present invention proposes a hierarchical scheduling method for a transmission-distribution-microgrid.

[0025] Figure 1 It is an architecture diagram of a coordinated scheduling system for a transmission-distribution-microgrid provided by an exemplary embodiment. As Figure 1 shown, the system includes a transmission network 11, a distribution network 12, and a microgrid cluster 13.

[0026] The transmission network 11 is used for a high-voltage power transmission network, responsible for transmitting electric energy from power plants (such as thermal power, hydropower, nuclear power, etc.) over a long distance to the load center or distribution nodes.

[0027] The distribution network 12 is used to receive electric energy from medium-voltage / low-voltage substations and distribute it to end-users (households, industries and businesses, etc.).

[0028] The microgrid cluster 13 includes multiple microgrids (microgrid 131, microgrid 132, microgrid 133). Each microgrid is a small-scale autonomous power system and can operate independently (island mode) or in parallel with the main grid. In the grid-connected mode, the microgrid exchanges power with the main grid. When the main grid fails, it switches to the island mode for autonomous power supply.

[0029] The three cooperate with each other to jointly build a modern power system.

[0030] Figure 2It is a flowchart of a hierarchical scheduling method for a transmission and distribution microgrid provided by an exemplary embodiment. This method is applied to a coordinated scheduling system for a transmission and distribution microgrid, and the system includes a transmission grid, a distribution grid, and a microgrid. As Figure 2 shown, this method may include the following steps: Step 201, obtain the grid data of the coordinated scheduling system for the transmission and distribution microgrid, and based on the obtained data, establish a first optimal power flow problem between the transmission grid and the distribution grid and a second optimal power flow problem between the distribution grid and the microgrid.

[0031] Optimal power flow (OPF) is a core optimization problem in the operation and planning of power systems. On the basis of traditional power flow calculation, it introduces optimization objectives such as economy, security, or environmental protection, and by adjusting control variables such as generator output, transformer tap positions, and reactive power compensation equipment, the system reaches an optimal operating state under the condition of satisfying all physical constraints.

[0032] In the present invention, the objective of OPF is to minimize the generation cost or the total network loss while ensuring that the voltage, current, and power do not exceed the limits.

[0033] Step 202, decompose the second optimal power flow problem into multiple sub-problems, and each sub-problem in the multiple sub-problems corresponds to a microgrid.

[0034] Step 203, solve the first optimal power flow problem to obtain a first result, and use the first result as the input of the second optimal power flow problem, and solve the multiple sub-problems according to the asynchronous alternating direction multiplier method assisted by double-loop learning to obtain a second result.

[0035] In one embodiment, the optimal power flow problem is solved based on a polynomial semi-definite programming cutting and chordal sparse relaxation model.

[0036] Here, the semi-definite programming (SDP) optimal power flow model is introduced first.

[0037] For each power grid, there are power flow constraint equations in the same form, which can be expressed as:

[0038]

[0039] The power balance constraint is expressed as:

[0040]

[0041] The power boundary constraint can be expressed as:

[0042]

[0043] The voltage and phase angle constraints can be expressed as:

[0044]

[0045] Let the complex power be , and the power equation can be rewritten as:

[0046]

[0047] Define the N-dimensional matrix W, which considers the product of complex variables. Let:

[0048] The above equation satisfies: .

[0049] By substituting the given variables, the power flow equation is converted into a linear equation:

[0050] And the matrix W satisfies the following constraints: ; ; Therefore, the voltage and phase angle constraints can also be expressed in the W space as:

[0051] By discarding the non-convex rank constraint in the SDP model, the SDP relaxation model is:

[0052] Where , , are the active powers of the generators in the transmission, distribution, and microgrids respectively; is the discharging active power in the energy storage system, and the objective functions , and are the generation costs in the transmission grid, distribution grid, and microgrid respectively. That is, they can be defined in the form of quadratic functions respectively:

[0053]

[0054]

[0055]

[0056]

[0057] A necessary and sufficient condition for a symmetric n×n matrix to be positive semi - definite is that all its principal sub - matrices are non - negative. The principal minors are the determinants of sub - matrices, which are formed by deleting some rows and the corresponding columns of the matrix. With this feature, the SDP constraints can be equivalently represented as a set of polynomial constraints (i.e., its principal minor constraints are positive), and this method is called SDP determinant cutting.

[0058] According to the characterization of the principal co - minors, the SDP constraints are equivalent to the set of determinant constraints derived from principal co - minors:

[0059] where is the th sub - matrix.

[0060] For large - scale power grids, the size of the positive semi - definite matrix is too large, significantly increasing the computational complexity and thus prolonging the solution time of the optimal power flow problem. To balance computational accuracy and efficiency, it is necessary to further relax the positive semi - definite constraint.

[0061] The present invention proposes a method for decomposing the positive semi - definite constraint of a matrix using chord sparsity. The theories of maximum clique, chord sparsity, matrix completion, decomposition, etc. in graph theory are applied to power grids with sparse characteristics, and the positive semi - definite constraint of the full matrix variable in the SDP problem is equivalently decomposed into the positive semi - definite constraints of multiple sub - matrices composed of non - sparse elements, thus achieving fast optimization.

[0062] The computational challenge of SDP relaxation stems from the positive semi - definite constraint of the matrix. For large grids with many nodes, directly enforcing the positive semi - definite constraint on the matrix usually leads to computational difficulties for grids with hundreds of buses. Therefore, the positive semi - definite constraint of the matrix can be replaced by the positive semi - definite constraints of multiple smaller matrices to achieve matrix dimensionality reduction and simplify the calculation. In large grids with many nodes, the matrix satisfies the conditions of a sparse matrix, so the present invention can utilize the chord sparsity of the network to address the computational challenge.

[0063] Since the is positive semi - definite if and only if all sub - matrices associated with the maximum cliques of the chordal extension are positive semi - definite. Therefore, using the sparse pattern of the Cholesky decomposition of the network adjacency matrix, the chordal extension of the graph can be constructed. Figure 3 The application of chordal sparsity in a 4 - node circular network in a grid is illustrated with a small test case.

[0064] As Figure 3 shown, in a ring network, a chord is an edge connecting two non - adjacent nodes. Therefore, Figure 3 the network represented by the black line in corresponds to the original circular network and contains no chords. By adding a chord (represented by the dashed line) in the circular network, nodes 2 and 3 are connected, which are in the cycle but not adjacent. Adding the red dashed line will create a chordal extension of the network topology. The chordal extension of the graph has two maximum cliques, denoted as , .

[0065] Using chordal sparsity to factorize the matrix will result in more positive semi - definite matrix constraints, but the factored matrix is much smaller than the original matrix, which has a great advantage in terms of computational time.

[0066] For a cyclic network, consider using chordal sparsity to relax the original positive semi - definite constraint.

[0067] For the original positive semi - definite constraint, first construct the maximum cliques using the chordal extension, and then reduce its dimension using the sparsity of the matrix, that is ; where, is the number of maximum cliques, is the positive semi - definite matrix corresponding to the -th maximum clique.

[0068] According to the SDP relaxation model, for each in the system:

[0069] where, represents the dimension of the matrix .

[0070] For a large - scale system, the number of principal minors becomes too large to significantly reduce the computational time for solving the optimal power flow problem. Therefore, further relaxation is required.

[0071] According to the requirements of computational accuracy and time, the principal minor constraint can be relaxed to a determinant constraint set consisting of principal minors: ; Therefore, considering the general form of all systems, the SDP relaxation model can be further relaxed into a polynomial semidefinite programming cuts and chordal sparsity relaxation (PCCSR) model.

[0072] The PCCSR model is as follows:

[0073] As Figure 4 shown, the first optimal power flow problem is solved normally, and the second optimal power flow problem is solved based on the learning-aided asynchronous alternating direction method of multipliers (LA-ADMM).

[0074] The learning-aided asynchronous alternating direction method of multipliers (LA-ADMM) is used to solve the synchronization bottleneck and model dependence in distributed optimization.

[0075] Consider decomposing the entire system into multiple regions, with each subsystem as a region. An OPF subproblem is established for each region to minimize its local operating cost while satisfying the local equality and inequality constraints of the network and components. The generator power generation and bus voltage angle within the region are the control variables of the corresponding subproblem. The regions are physically connected through tie lines. The voltage angles of the tie line terminals , , , and (i.e., ) are the shared variables between the subproblems.

[0076] Taking the ADMM between two regions as an example, assume that the DC-OPF problem is decomposed into two subproblems and . The local variables of subproblems and (e.g., the power generated by the generator and the bus voltage angle) are represented by and respectively. The voltage angle of the boundary bus is the shared variable represented by , which is replicated to create and . The consistency constraint is . The OPF subproblem is as follows:

[0077] Among them, is the cost function in the region , and represent the local equality and inequality constraints of the sub-problem . is the decision variable, is the known parameter received from the sub-problem .

[0078] The OPF sub-problem is as follows:

[0079] Among them, is the variable, is the known value received from the sub-problem .

[0080] The augmented Lagrangian relaxation method is used to penalize the consistency constraint in the objective function. The OPF sub-problem is iteratively solved as follows:

[0081] Among them, , and represent the iteration number, Lagrange multiplier, and penalty parameter, respectively. At each iteration , the above formula is executed in sequence. If the original residual ( ) and the dual residual ( ) become smaller than the expected criterion , then ADMM converges.

[0082]

[0083] Step 204: When the second result satisfies the convergence conditions of the first optimal power flow problem and the second optimal power flow problem, the transmission network, distribution network, and microgrid of the transmission and distribution microgrid coordinated scheduling system are scheduled based on the second result.

[0084] In this embodiment, on the one hand, a three-layer coordinated architecture including a transmission network, a distribution network, and a microgrid is designed, and the optimal power flow problem between different power grids is solved, optimizing the overall architecture of the power system; on the other hand, the optimal power flow problem between the distribution network and the microgrid is solved in parallel by the asynchronous alternating direction multiplier method assisted by double-loop learning, which not only improves the calculation efficiency but also increases the accuracy of the calculation results, thereby enhancing the accuracy of power system scheduling.

[0085] In one embodiment, the multiple sub-problems are solved according to the asynchronous alternating direction multiplier method assisted by dual-loop learning to obtain a second result, including: for each sub-problem, in the case where the shared variable value of the adjacent sub-problem is not received, the shared variable value is predicted based on the regression technology, and the penalty consistency constraint is penalized in the objective function based on the augmented Lagrangian relaxation method for iterative solution; the abnormal switch controller is designed to read the local objective function value of the sub-problem, and anomaly detection is performed in the iterative process; when the result of the iterative solution passes the anomaly detection, the second result is determined.

[0086] Furthermore, the prediction of shared variable values ​​based on regression technology includes: obtaining the missing shared variable values ​​in the first two iterations and forming a linear extrapolation term; assuming that the missing shared variable values ​​jump in the extrapolation direction, defining the momentum term as the inertia of the iterative trend of the shared variable value, and the momentum term is a correction term added to the extrapolation to correct the predicted direction and make it bounded.

[0087] It has the ability to predict information and can handle a considerable degree of asynchrony between sub-problems. For the specific LA-ADMM algorithm flow chart, see Figure 5 . The number of iterations of a subproblem is defined as .

[0088] In each iteration At the beginning, each OPF subproblem (assuming that the subproblem ) Checks whether it has been removed from its adjacent subproblem Receive shared variable value If not, regression-based techniques are used to predict the shared variable values ​​and allow the corresponding OPF subproblems to be solved. The abnormal switch controller is designed to read the subproblem The local objective function value of , which is used for anomaly detection. If the number of consecutive iterations of lost information exceeds a few iterations and prediction oscillations are observed, the switching controller bypasses the error propagation effect on the Lagrange multipliers; otherwise, the inevitable prediction errors may accumulate and degrade the convergence performance or even diverge.

[0089] A procedure is needed to determine which OPF subproblem, when and how to perform the prediction step. exist Request when unavailable , will be assumed to be the missing value that should be predicted. Inspired by the Nesterov technique for accelerated gradient descent, this paper proposes a momentum-based extrapolation method. At that time, the values of the missing shared variables obtained in the first two iterations are used to form a linear extrapolation term. Assume that the values of the missing shared variables jump in the extrapolation direction. The momentum term is defined as the inertia of the trend of the shared variable values from iteration 1 to The inertia of the trend of the shared variable values. Momentum is a correction term added to the extrapolation to correct the direction of the prediction and make it bounded, even in the case of missing information in several consecutive iterations of ADMM. If missing information is observed in multiple iterations, the momentum term also prevents the prediction from moving far from the vicinity of the latest known value of the shared variable. As the prediction step is iteratively executed again and again, the predicted value will become saturated to prevent large prediction errors while allowing the subproblems to be continuously solved.

[0090] Due to the heterogeneity of the calculations, different iteration indices are assigned to different OPF subproblems. Accordingly, the subproblem can be formulated as follows:

[0091]

[0092]

[0093]

[0094] Obviously, the momentum extrapolation prediction correction method is very effective for predicting the missing values of shared variables, requiring at most only a few iterations. Assume that due to a significant degree of subproblem asynchrony or communication failure / delay, no update of the actual shared variable is received after several iterations. In this case, even the correction step cannot prevent ADMM from moving towards a low-quality result point or a suboptimal / infeasible point.

[0095] This is mainly because the propagation of prediction errors on is updated after each iteration. If the degree of asynchrony between OPF subproblems is large and missing information is observed in many consecutive iterations, or the predicted values of the missing information are not accurate enough, the present invention adds some steps to the original algorithm to avoid obtaining low-quality results.

[0096] The process of performing distributed optimization is similar to an online stream, where new data is obtained through iteration. Therefore, an anomaly detection technique is needed that can be used for online stream data even when there is not enough data available in the first few iterations. This requirement makes most machine learning techniques inapplicable to anomaly detection when running distributed optimization iteratively. It is updated once every iteration , and the number of iterations for LA-ADMM to converge depends on the problem.

[0097] It should be noted that at the end of each iteration, the sub-problems exchange the updated values of the shared variables and the Lagrange multipliers. Since there is no coordinator, each sub-problem uses its shared variable values and the actual or predicted shared variable values of its neighbors to update its . During the iteration process, the Lagrange multipliers in adjacent sub-problems may be slightly different; however, as the number of iterations increases, they become closer and closer and reach the same value at convergence. The prediction is performed on the missing shared variable values, rather than the local decision variables. The goal is to determine the optimal solution of the entire problem. It is assumed that the sub-problems do not participate in a strategic game to maximize their profits.

[0098] Furthermore, anomaly detection is performed during the iteration process, including: performing anomaly detection based on an online streaming unsupervised ground anomaly learner.

[0099] To perform anomaly detection, the present invention proposes an online streaming unsupervised ground anomaly learner suitable for distributed optimization applications and performs anomaly detection based on the online streaming unsupervised ground anomaly learner: Step 301. According to the curves obtained during the 1st to the iteration, predict the power generation cost function at iteration , and then detect the severity of the anomaly of the current iteration compared with the previous iterations. For this purpose, the following steps are implemented in the present invention.

[0100] Step 3011. Wavelet denoising: Denoising is widely used in image processing to reduce the noise in images. Wavelet denoising classifies data into low-frequency and high-frequency categories. Most of the useful information exists in the low-frequency. The data in the high-frequency category is filtered out to reduce noise and smooth the data. The present invention uses wavelets to detect noise or anomalies in the power generation cost curve during the distributed optimization iteration process.

[0101] Step 3012. Moving average: The moving average function, usually called the smoothing function, slides a window of size 5 over the entire selected data and replaces each point with the average value of the last 5 iterations. This is similar to a low-pass filter. Using the moving average can reduce large fluctuations without deteriorating the trend of the curve. Moving average is a smoothing technique in the time domain, while wavelet is a smoothing technique in the frequency domain.

[0102] Step 3013. Energy difference evaluation: Wavelet denoising removes high frequencies, the moving average retains low frequencies, and smooths large jumps. Perform wavelet denoising and moving average several times to further reduce the noise until the energy difference between the two signals becomes less than a threshold (e.g., 0.1%). After completing this step, an ideal curve without anomalies and noise will be obtained .

[0103] Calculate the denoised one using the following formula and the original The relative energy difference between them is as follows:

[0104] Step 302, Detect the anomaly severity. In each iteration, use as the anomaly metric, and select the most recent 20 iterations to define the anomaly severity. The anomaly severity is a relative metric, and using a single threshold to classify severe anomalies will reduce the anomaly detection performance. Use k-means to classify and sort the anomalies into groups. Define the degree of severe anomalies according to the category to which the last iteration belongs. For example, if , the class greater than a certain threshold is regarded as a severe anomaly, and any score greater than 7 means that the inner loop should be activated. Since the Lagrange multipliers have the most critical impact on the ADMM convergence behavior, they are not updated in the inner loop. This will prevent the propagation of the shared variable prediction error on the Lagrange multipliers.

[0105] In one embodiment, the grid data includes node data, line data, and network topology data; the node data includes voltage amplitude, injected active power, active power demand, reactive power demand, and reactive power provided by the generator, the line data includes active power flow, reactive power flow, voltage phase difference angle, resistance, and reactance, and the network topology data is the connection relationship between each node and line.

[0106] Figure 6 is a schematic structural diagram of a device provided by an exemplary embodiment. Please refer to Figure 6 , at the hardware level, the device includes a processor 602, an internal bus 604, a network interface 606, a memory 608, and a non-volatile memory 610. Of course, there may also be other hardware required for other functions. One or more embodiments of the present invention can be implemented in a software manner. For example, the processor 602 reads the corresponding computer program from the non-volatile memory 610 into the memory 608 and then runs it. Of course, in addition to the software implementation manner, one or more embodiments of the present invention do not exclude other implementation manners, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, and can also be hardware or logic devices.

[0107] Please refer to Figure 7 , a hierarchical dispatching device for a transmission and distribution microgrid can be applied to the device as shown in Figure 7 to implement the technical solution of the present invention. The device may include: An acquisition unit 701, configured to acquire grid data of a transmission and distribution microgrid collaborative scheduling system, and establish a first optimal power flow problem between a transmission grid and a distribution grid and a second optimal power flow problem between the distribution grid and the microgrid based on the acquired data; A decomposition unit 702, configured to decompose the optimal power flow problem between the distribution grid and the microgrid into multiple sub-problems, where each sub-problem in the multiple sub-problems corresponds to a microgrid; A solving unit 703, configured to solve the first optimal power flow problem to obtain a first result, and use the first result as an input of the second optimal power flow problem, and solve the multiple sub-problems according to the asynchronous alternating direction multiplier method assisted by double-loop learning to obtain a second result; A scheduling unit 704, configured to schedule the transmission grid, the distribution grid, and the microgrid of the transmission and distribution microgrid collaborative scheduling system based on the second result when the second result meets the convergence conditions of the first optimal power flow problem and the second optimal power flow problem.

[0108] Optionally, the solving unit 703 is specifically configured to: Solve the optimal power flow problem based on a polynomial semi-definite programming cutting and chordal sparse relaxation model.

[0109] Optionally, the polynomial semi-definite programming cutting and chordal sparse relaxation model is expressed as:

[0110]

[0111] Optionally, the solving unit 703 is specifically configured to: For each sub-problem, when the shared variable value of an adjacent sub-problem is not received, predict the shared variable value based on a regression technique, and perform iterative solution by penalizing the consistency constraint in the objective function based on the augmented Lagrangian relaxation method; Design the abnormal switch controller to read the local objective function value of the sub-problem and perform abnormal detection during the iterative process; Determine the second result when the result of the iterative solution passes the abnormal detection.

[0112] Optionally, the solving unit 703 is specifically configured to: Acquire the missing shared variable values in the previous two iterations and form a linear extrapolation term; Assume that the missing shared variable value jumps in the extrapolation direction, and define the momentum term as the inertia of the iterative trend of the shared variable value. The momentum term is a correction term added to the extrapolation, used to correct the prediction direction and make it bounded.

[0113] Optionally, the solving unit 703 is specifically configured to: Anomaly detection is performed based on an online flow unsupervised ground anomaly learner.

[0114] Optionally, the power grid data includes node data, line data, and network topology data; the node data includes voltage amplitude, injected active power, active power demand, reactive power demand, and reactive power provided by a generator, and the line data includes active power flow, reactive power flow, voltage phase difference angle, resistance, and reactance, and the network topology data is the connection relationship between each node and line.

[0115] The systems, devices, modules, or units illustrated in the above embodiments may be specifically implemented by a computer chip or entity, or by a product with certain functions. A typical implementation device is a computer, and the specific form of the computer may be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver device, a game console, a tablet computer, a wearable device, or a combination of any several of these devices.

[0116] In a typical configuration, a computer includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0117] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM), and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0118] The computer-readable medium includes permanent and non-permanent, removable and non-removable media and can store information by any method or technology. The information may be computer-readable instructions, data structures, program modules, or other data. Examples of the computer's storage medium include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassette tapes, disk storage, quantum memory, graphene-based storage media, or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0119] For a computer-readable medium as described above or in any other form (or, a computer-readable storage medium), computer instructions can be stored thereon, and when the instructions are executed by a processor, one or more of the above-described embodiments are implemented, thereby implementing the technical solution of the present invention.

[0120] The present invention also provides a computer program, which when executed by a processor, implements one or more of the above-described embodiments, thereby implementing the technical solution of the present invention. Among them, the computer program can be specifically recorded on a computer-readable medium as described above or in any other form, and the present invention does not limit this.

[0121] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, commodity or device. Without more limitations, the element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the said element.

[0122] The above specifically describes certain embodiments of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] The terms used in one or more embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of the present invention. The singular forms "a", "the" and "said" used in one or more embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0124] It should be understood that although the terms first, second, third, etc. may be used in one or more embodiments of the present invention to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of one or more embodiments of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0125] The above are only the preferred embodiments of one or more embodiments of the present invention, and are not intended to limit one or more embodiments of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of one or more embodiments of the present invention shall be included within the scope of protection of one or more embodiments of the present invention.

Claims

1. A hierarchical dispatching method for transmission and distribution microgrids, characterized in that: include: Acquire grid data of the transmission and distribution microgrid coordinated dispatching system, and establish a first optimal power flow problem between the transmission network and the distribution network and a second optimal power flow problem between the distribution network and the microgrid based on the acquired data; Decomposing the second optimal power flow problem into a plurality of sub-problems, each of the plurality of sub-problems corresponding to a microgrid; Solving the first optimal power flow problem to obtain a first result, using the first result as an input of the second optimal power flow problem, and solving the multiple sub-problems according to a dual-loop learning-assisted asynchronous alternating direction multiplier method to obtain a second result; When the second result satisfies the convergence conditions of the first optimal power flow problem and the second optimal power flow problem, the transmission network, distribution network, and microgrid of the transmission and distribution microgrid coordinated dispatching system are dispatched based on the second result.

2. The method according to claim 1, characterized in that: The first optimal power flow problem is solved to obtain a first result, and the first result is used as an input of the second optimal power flow problem, and the plurality of sub-problems are solved according to the dual-loop learning-assisted asynchronous alternating direction multiplier method to obtain a second result, including: The optimal power flow problem is solved based on the polynomial semidefinite programming cutting and chord sparse relaxation model.

3. The method according to claim 2, characterized in that The polynomial semi-positive definite programming cutting and chord sparse relaxation model is expressed as:

4. The method according to claim 1, characterized in that: The step of solving the plurality of sub-problems according to the dual-loop learning assisted asynchronous alternating direction multiplier method to obtain a second result includes: For each sub-problem, without receiving the shared variable values ​​of the adjacent sub-problems, the shared variable values ​​are predicted based on regression technology, and the penalty consistency constraints in the objective function are iteratively solved based on the augmented Lagrangian relaxation method; The abnormal switch controller is designed to read the local objective function value of the sub-problem and perform anomaly detection in the iterative process; When the result of the iterative solution passes the anomaly detection, the second result is determined.

5. The method according to claim 4, characterized in that The prediction of the shared variable value based on the regression technology includes: Get the missing shared variable values ​​in the first two iterations and form the linear extrapolation terms; Assuming that the missing shared variable value jumps in the direction of the extrapolation, the momentum term is defined as the inertia of the iterative trend of the shared variable value. The momentum term is a correction term added to the extrapolation to correct the direction of the prediction and make it bounded.

6. The method according to claim 4, characterized in that The abnormality detection during the iteration process includes: Anomaly detection based on online stream unsupervised ground truth anomaly learner.

7. The method according to claim 1, characterized in that The power grid data includes node data, line data and network topology data; the node data includes voltage amplitude, injected active power, active power demand, reactive power demand, and reactive power provided by the generator; the line data includes active power flow, reactive power flow, voltage phase difference angle, resistance, and reactance; and the network topology data is the connection relationship between each node and line.

8. A transmission and distribution microgrid hierarchical dispatching device, characterized in that: The device comprises: Acquisition unit: acquires grid data of the transmission and distribution microgrid coordinated dispatching system, and establishes a first optimal power flow problem between the transmission network and the distribution network and a second optimal power flow problem between the distribution network and the microgrid based on the acquired data; Decomposition unit: decomposing the optimal power flow problem between the distribution network and the microgrid into multiple sub-problems, each of the multiple sub-problems corresponds to a microgrid; A solving unit: solving the first optimal power flow problem to obtain a first result, using the first result as an input of the second optimal power flow problem, and solving the multiple sub-problems according to the dual-loop learning-assisted asynchronous alternating direction multiplier method to obtain a second result; Dispatching unit: When the second result satisfies the convergence conditions of the first optimal power flow problem and the second optimal power flow problem, the transmission network, distribution network, and microgrid of the transmission and distribution microgrid coordinated dispatching system are dispatched based on the second result.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor implements the steps of the method according to any one of claims 1 to 7 by running the executable instructions.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images