Distributed-local two-stage voltage control method, system, equipment and medium

Abstract

The invention relates to the technical field of power distribution reactive voltage control, and discloses a distributed-in-place two-stage voltage control method, system, device and medium, and the method comprises the steps: solving the reactive power output of a photovoltaic inverter in a distributed photovoltaic cluster with the minimum voltage deviation in the cluster as a target, and calculating the reactive power output of the photovoltaic inverter in the cluster; and iteratively interacting cluster boundary variables through an alternating direction multiplier algorithm so as to realize cooperation of reactive power control strategies among different photovoltaic clusters. In the on-site control stage, it is considered that the reactive power control strategy obtained in the distributed cooperation stage is not suitable any more due to the fact that the strong volatility of distributed photovoltaic is likely to cause. A Q-P affine control strategy is formulated by photovoltaic power in each edge cluster by taking a reactive power strategy of a distributed coordination stage as a basic value, reactive power output of the photovoltaic power is adjusted in real time according to active power fluctuation of the photovoltaic power, and a control curve is segmented through a spatial dimension raising method, so that the flexibility of a local control strategy is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution reactive voltage control, and particularly relates to a distributed-on-site two-stage voltage control method, system, device and medium. BACKGROUND

[0002] At present, with high proportion of distributed power (DG) access to distribution system, the operation control of distribution system has many new problems, for example, because of the strong volatility of distributed photovoltaic, the voltage and voltage fluctuation problem. In the traditional voltage regulation mode, it mainly depends on on-load tap changer (OLTC) and capacitor bank (CB) and other devices to adjust, but the adjustment speed of these devices is relatively slow, and it is difficult to adapt to the fast fluctuation of photovoltaic power.

[0003] By contrast, the distributed photovoltaic connected to the grid through the inverter can directly use its remaining capacity to provide real-time reactive voltage support, which provides an effective solution for real-time voltage control of distribution network.

[0004] At present, in the field of distribution network, the method of reactive voltage control of distributed photovoltaic can be roughly divided into centralized control, distributed control and on-site control. Among them, the centralized control optimizes the whole system as a target to regulate controllable resources, and the theoretical control effect is good, but it depends too much on the central controller, whole network measurement data and reliable communication network. With the access of flexible controllable distributed resources to the grid, this way is not so sufficient in complex scene adaptability. The distributed control divides the distribution network into sub-regions, and realizes voltage control through regional coordination, which to some extent gets rid of the demand for central controller, whole network measurement data and reliable communication network. However, this control method needs regional information interaction iteration and periodic optimization scheduling, and the change of photovoltaic active power may make the control strategy and actual working condition mismatch, which is difficult to control voltage in real time.

[0005] The on-site control is based on local information to develop strategies, which can quickly respond to local power and voltage fluctuation, has small calculation amount and strong robustness. Common strategies include Q-V droop control and Q-P control. Q-V control may have stability problem, and the strategy needs to consider system stability and control convergence. Q-P control is relatively simple in design and engineering implementation. However, the lack of system-level coordination of each distributed photovoltaic unit needs to be improved.

[0006] In summary, in order to solve the voltage reactive control problem of large-scale access of distributed photovoltaic, this paper proposes a distributed-on-site two-stage collaborative voltage control strategy for the edge side distributed photovoltaic cluster of distribution network. SUMMARY

[0007] In view of the above existing problems, the present application is proposed.

[0008] Therefore, the present application provides a distributed-on-site two-stage voltage control method, system, device and medium, which can solve the voltage and reactive power control problem caused by large-scale access of distributed photovoltaic to distribution network.

[0009] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a distributed-on-site two-stage voltage control method, comprising: According to the network topology of the distribution network, the line parameters, the access position and capacity of each distributed photovoltaic cluster and the load data, the distribution network is divided into a plurality of photovoltaic cluster areas; For each cluster, a local optimization model containing power flow constraints, voltage safety constraints and inverter capacity constraints is established with the sum of the absolute values of the voltage deviations of the nodes in the cluster as the target; The coordinated variables of the boundary nodes are iteratively exchanged between the clusters by the alternating direction multiplier algorithm, and the collaborative optimization of the multi-cluster reactive power output strategy is completed, and the initial reactive power output setting value of each photovoltaic node inverter is obtained; Based on the initial reactive power output setting value, an on-site reactive power adjustment rule is configured for each photovoltaic node, which adjusts the reactive power output in real time according to the change of the actual active power output of the photovoltaic relative to the predicted value in an affine relationship.

[0010] As a preferred scheme of the distributed-on-site two-stage voltage control method of the present application, wherein: further comprising: Considering the difference in the feasible operating region of the inverter when the photovoltaic active power rises and falls, the affine relationship is divided into two independent intervals: When the active power increases, a first slope is used, and when the active power decreases, a second slope is used; By introducing two non-negative auxiliary variables to upgrade the original control variables, and performing convex hull relaxation on the upgraded feasible region, a robust optimization model containing piecewise control characteristics is constructed.

[0011] As a preferred scheme of the distributed-on-site two-stage voltage control method of the present application, wherein: further comprising: The voltage deviation uncertainty constraint in the robust optimization model is transformed into a linear constraint form by dual transformation, and the corresponding piecewise affine control parameters of each photovoltaic node are obtained by solving; During system operation, the reactive power output is dynamically adjusted using the piecewise affine control parameters according to the real-time measured active power change to suppress voltage out-of-limit.

[0012] As a preferred embodiment of the distributed-local two-stage voltage control method described in this invention, the coordination variable of the boundary node includes the injected power of the boundary node or the transmission power of the connected branch. Each cluster only needs to exchange the coordination variable with its neighboring clusters, without relying on a global central controller.

[0013] This preferred solution reduces the dependence of the control process on the global central controller, reduces system communication pressure and single point of failure risk, and improves system reliability and stability.

[0014] As a preferred embodiment of the distributed-local two-stage voltage control method of the present invention, the reference point of the affine relationship is determined by the initial reactive power output setting value, and its slope reflects the reactive power adjustment caused by the unit active power change.

[0015] As a preferred embodiment of the distributed-local two-stage voltage control method described in this invention, the convex hull relaxation is achieved by retaining all vertices of the original piecewise feasible region and constructing its minimum convex set, ensuring that the relaxed model can still accurately approximate the original non-convex control characteristics.

[0016] As a preferred embodiment of the distributed-local two-stage voltage control method described in this invention, the dual transformation is for the worst-case power fluctuation scenario. By using dual variables, the robust constraints containing maximum value operations are equivalently transformed into a set of linear inequality constraints to obtain analytical control parameters deployed on the local controller.

[0017] Secondly, the present invention provides a distributed-local two-stage voltage control system, comprising: The cluster partitioning module is used to divide the distribution network into multiple photovoltaic cluster areas based on the network topology, line parameters, access location and capacity of each distributed photovoltaic cluster, and load data. The first optimization model building module is used to build a local optimization model for each cluster, with the goal of minimizing the sum of the absolute values ​​of the voltage deviations of the nodes within the cluster, and includes power flow constraints, voltage safety constraints, and inverter capacity constraints. The numerical calculation module is used to iteratively exchange the coordination variables of the boundary nodes between the clusters through the alternating direction multiplier algorithm, complete the collaborative optimization of the reactive power output strategy of the multi-cluster, and obtain the initial reactive power output setpoint of the inverter of each photovoltaic node. The correction module is used to configure local reactive power adjustment rules for each photovoltaic node based on the initial reactive power output setting value. The rules correct the reactive power output in real time according to the change of the actual active power output of the photovoltaic relative to the predicted value, based on an affine relationship. The second optimization model building module is used to divide the affine relationship into two independent intervals, taking into account the difference in the feasible operating region of the inverter when the photovoltaic active power output increases and decreases: The first slope is used when the active power output increases, and the second slope is used when the active power output decreases. By introducing two non-negative auxiliary variables to increase the dimensionality of the original control variables and performing convex hull relaxation on the feasible region after the dimensionality increase, a robust optimization model containing piecewise control characteristics is constructed. The solution module is used to perform dual transformation on the voltage deviation uncertainty constraint in the robust optimization model, transform it into a linear constraint form, and solve for the piecewise affine control parameters corresponding to each photovoltaic node. The adjustment module is used to dynamically adjust the reactive power output based on the real-time measured changes in active power during system operation, using the segmented affine control parameters, in order to suppress voltage over-limit.

[0018] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0019] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0020] Compared with existing technologies, the beneficial effect of this invention is that it proposes a distributed-local two-stage voltage control method. The distributed photovoltaic (PV) cluster aims to minimize the voltage deviation within the cluster, solving for the reactive power output of the PV inverters within the cluster. It then uses an alternating direction multiplier algorithm to iteratively interact with the cluster boundary variables, thereby achieving coordination of reactive power control strategies among different PV clusters. In the local control stage, considering the strong volatility of distributed PV, the reactive power control strategy obtained in the distributed coordination stage may no longer be applicable. Each peripheral cluster uses the reactive power strategy from the distributed coordination stage as a base value to formulate a QP affine control strategy. This strategy adjusts the reactive power output in real time according to the active power fluctuations of the PV, and further improves the flexibility of the local control strategy by segmenting the control curve using a spatial dimensionality increase method. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1This is a flowchart of a distributed-local two-stage voltage control method provided in one embodiment of the present invention.

[0023] Figure 2 Another flowchart of a distributed-local two-stage voltage control method provided in one embodiment of the present invention.

[0024] Figure 3 The diagram shows an improved IEEE 33-node computational structure for a distributed-local two-stage voltage control method provided in one embodiment of the present invention.

[0025] Figure 4 This is a predicted distributed photovoltaic and load curve diagram of a distributed-local two-stage voltage control method provided in one embodiment of the present invention.

[0026] Figure 5 The diagram shows the voltage control effect of Scheme I of a distributed-local two-stage voltage control method provided in an embodiment of the present invention.

[0027] Figure 6 The diagram shows the voltage control effect of Scheme II of a distributed-local two-stage voltage control method provided in an embodiment of the present invention.

[0028] Figure 7 The diagram shows the voltage control effect of Scheme III of a distributed-local two-stage voltage control method provided in an embodiment of the present invention.

[0029] Figure 8 The voltage control effect diagram is shown in Scheme IV of a distributed-local two-stage voltage control method provided in an embodiment of the present invention.

[0030] Figure 9 This is an internal structural diagram of an electronic device for a distributed-local two-stage voltage control method provided in one embodiment of the present invention. Detailed Implementation

[0031] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0032] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a distributed-local two-stage voltage control method, including: This invention provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to implement this distributed-local two-stage voltage control method with reference to several embodiments. Figure 1 A flowchart of a distributed-local two-stage voltage control method is shown, including: S101, based on the network topology, line parameters, access location and capacity of each distributed photovoltaic cluster, and load data of the distribution network, the distribution network is divided into multiple photovoltaic cluster areas; In some embodiments, when dividing photovoltaic cluster areas, the key nodes and branch connections in the radial feeders are first identified based on the distribution network topology. For example, in the improved IEEE 33-node system, the main feeder extends sequentially from node 1 to node 33, forming a typical tree structure. The impedance parameters of each branch are known and can be used for electrical distance calculation. Based on this, the electrical coupling strength between nodes is quantified by combining the resistance and reactance values ​​in the line parameters, thereby determining which nodes have a high correlation in voltage response. For example, nodes 16, 17, and 18 are located at the end of the same branch, have low line impedances between them, and exhibit synchronous voltage fluctuations, thus they can be classified into the same cluster.

[0033] Simultaneously, considering the access location and capacity information of each distributed photovoltaic (PV) cluster, PV units with geographically adjacent locations and similar total installed capacity are grouped into the same region. For example, nodes 21 and 22 are connected to 200kW and 600kW PV respectively, and are connected to nodes 24 and 25 with 200kW and 600kW PV respectively via a common upstream branch on the electrical path. If their load characteristics are similar, they can be merged into one PV cluster region. Furthermore, load data is used to assess the active and reactive power demand levels of each node under typical operating scenarios. For example, during peak load periods, the area between nodes 30 and 33 experiences heavy load and low PV output, resulting in significant voltage support requirements. This area needs to be separately grouped to achieve refined reactive power control. The cluster division process integrates the above multi-dimensional information to ensure that each cluster has strong voltage coupling, similar source-load characteristics, and a controllable scale, facilitating the subsequent construction of local optimization models and the effective exchange of boundary coordination variables.

[0034] The photovoltaic cluster area here refers to a subsystem consisting of several adjacent or electrically closely related photovoltaic access nodes and their connected load nodes. This subsystem has relatively independent optimization capabilities and coordination boundaries at the voltage control level.

[0035] It should be noted that step S101 divides the distribution network into multiple photovoltaic cluster areas, which reduces the scale of the problem, enhances local voltage coupling, and provides a structural basis for subsequent distributed modeling and coordinated control.

[0036] S102. For each cluster, a local optimization model is established with the goal of minimizing the sum of the absolute values ​​of the voltage deviations of the nodes within the cluster. This model includes power flow constraints, voltage safety constraints, and inverter capacity constraints. In some embodiments, when constructing the local optimization model, the absolute values ​​of the deviations between the voltage amplitudes of all nodes in the cluster and their nominal values ​​(usually 1.0 pu) are first accumulated. This summation serves as the objective function to achieve balanced control of the overall voltage level. For example, in a photovoltaic cluster containing nodes 16 to 18, if node 17 experiences a sudden increase in photovoltaic output, causing its voltage to rise to 1.045 pu, while node 18 experiences a drop to 0.975 pu due to a heavier load, the objective function will penalize both deviations simultaneously, driving reactive power resources to adjust in the direction of reducing the overall deviation. This optimization model strictly embeds the power flow constraints of the distribution network, using branch power flow equations in the form of DistFlow to describe the nonlinear relationship between active and reactive power and the square of the node voltage. After linearization, these equations are used for iterative solution. For example, for the branch connecting nodes 17 and 18, its active power loss is determined by the line resistance and the square of the branch current, while the reactive power voltage drop is affected by the reactance. These physical relationships are accurately modeled to ensure the feasibility of the operating state.

[0037] Meanwhile, the model introduces voltage safety constraints, forcing the voltage amplitude of all nodes in the cluster to remain within a preset safety range. For example, the lower limit is set at 0.90 pu and the upper limit at 1.10 pu, and the optimization objective further tightens this to 0.98–1.02 pu to improve power quality and prevent the risk of exceeding limits. In addition, the model includes inverter capacity constraints, limiting the reactive power capacity of each photovoltaic node's inverter based on its rated apparent power. For example, node 22, connected to 600kW photovoltaic power, has an inverter capacity of 630kVA, and its adjustable reactive power range is limited by… Where P represents the real-time active power output, ensuring that reactive power regulation does not exceed the physical limits of the equipment. These three types of constraints together constitute a locally optimized problem with a clear structure and well-defined physical meaning, providing a reliable foundation for subsequent distributed collaboration.

[0038] The local optimization model here refers to a mathematical programming problem that can be solved by relying only on internal cluster measurements and parameters without requiring global information. Its output is the initial reactive power output setting value of each photovoltaic inverter in the cluster.

[0039] It should be noted that step S102 establishes a local optimization model, enabling each cluster to independently solve its own optimal reactive power strategy, taking into account voltage safety, power flow physics, and equipment capacity limitations, thereby improving control feasibility and targeting.

[0040] S103, through the alternating direction multiplier algorithm, iteratively exchanges the coordination variables of the boundary nodes among the clusters to complete the collaborative optimization of the reactive power output strategy of the multi-cluster, and obtains the initial reactive power output setting value of the inverter of each photovoltaic node. In this embodiment of the invention, the coordination variable of the boundary node includes the injected power of the boundary node or the transmission power of the connected branch. Each cluster only needs to exchange the coordination variable with its neighboring clusters, without relying on the global central controller.

[0041] In some embodiments, after constructing the local optimization model for each photovoltaic cluster, the Alternating Directional Multiplier Method (ADMM) is used to achieve collaborative optimization among multiple clusters. This method decomposes the originally coupled global optimization problem into several subproblems that can be solved in parallel by defining coordination variables on the boundary nodes shared by adjacent clusters. For example, in the improved IEEE 33-node system, if cluster A contains nodes 16–18 and cluster B contains nodes 19–21, and the two are connected by the branch (18, 19), then nodes 18 and 19 constitute a boundary node pair, and their voltage magnitude and branch power flow are selected as coordination variables. During the ADMM iteration process, each cluster independently solves its own local optimization problem and outputs the voltage estimate at the boundary node and the reactive power of the associated branch. Subsequently, each cluster sends this boundary information to the coordination center or directly exchanges it with neighboring clusters. The coordination center updates the Lagrange multipliers and penalty parameters based on the consistency constraints. For example, if cluster A calculates the voltage of node 18 as 1.015 pu, while cluster B calculates the corresponding voltage value of node 19 as 1.008 pu, the difference between the two will be introduced into the objective function of the next round through the multiplier term, forcing subsequent iterations to gradually reduce the difference.

[0042] This process is repeated until the voltage and power coordination variables of all boundary nodes meet the preset convergence accuracy. For example, the boundary voltage difference is less than 0.001 pu and the branch reactive power imbalance is less than 1 kVar in three consecutive iterations. Finally, each cluster determines the initial reactive power output setting value of its internal photovoltaic inverters based on the converged solution. For example, the inverter at node 17 outputs 120 kVar of capacitive reactive power, and the inverter at node 20 outputs -80 kVar of inductive reactive power to collaboratively support the voltage stability of the entire grid.

[0043] The initial reactive power output setting value here refers to the reactive power dispatch benchmark value of each photovoltaic node inverter that is agreed upon through a distributed coordination mechanism after considering the electrical coupling relationship of multiple clusters, and serves as the starting point for subsequent real-time control or secondary adjustment.

[0044] It should be noted that step S103 coordinates each cluster through the alternating direction multiplier algorithm, and global consistency can be achieved by exchanging only boundary variables, avoiding reliance on the central controller. This reduces communication burden and single point of failure risk while ensuring coordination effect.

[0045] S104, based on the initial reactive power output setting value, configures a local reactive power adjustment rule for each photovoltaic node. This rule corrects the reactive power output in real time according to the change in the actual active power output of the photovoltaic relative to the predicted value, based on an affine relationship.

[0046] In some embodiments, after obtaining the initial reactive power output setpoints for each photovoltaic node inverter, a local reactive power adjustment rule is configured for each node. This rule uses the deviation between the locally measurable actual photovoltaic active power output and the day-ahead or intraday predicted value as the input variable, and corrects the current reactive power output in real time through a preset affine function, thereby achieving a rapid response to voltage fluctuations. For example, the initial reactive power setpoint for node 22 is 150 kVar, and its predicted photovoltaic active power output is 500 kW, but a sudden increase in actual irradiance causes the measured active power to reach 620 kW, with a deviation ΔP of +120 kW. If the affine coefficient configured for this node is -0.3, then the reactive power correction ΔQ = -0.3 × 120 = -36 kVar, and the final real-time reactive power output adjustment is 150 - 36 = 114 kVar. The slope coefficient in this affine relationship is pre-tuned based on the sensitivity analysis of the cluster's local optimization model to ensure that voltage exceedances are effectively suppressed within the typical operating range, without exceeding the inverter capacity boundary. For example, if node 25 connects to 600 kW photovoltaic power and has an inverter capacity of 630 kVA, under the condition of initial active power of 550 kW and initial reactive power of 180 kVar, its maximum allowable reactive power reduction is limited by… The affine coefficient is approximately 112 kVar, therefore it is strictly limited to the range of [-0.25, 0] to avoid overshooting. The adjustment rules are executed entirely locally without any communication dependencies. The calculation can be completed simply by collecting the active power of the local node in real time and comparing it with the predicted value. This is suitable for scenarios with second-level power fluctuations caused by rapid cloud movement.

[0047] Here, the affine relationship refers to the linear proportional relationship between the reactive power correction and the active power deviation, in the form of ΔQ=k·ΔP, where k is a negative real number, reflecting the physical logic of reducing reactive power injection to suppress voltage rise when active power increases.

[0048] In this embodiment of the invention, the reference point of the affine relationship is determined by the initial reactive power output setting value, and its slope reflects the reactive power adjustment caused by the unit change in active power.

[0049] It should be noted that step S104 configures local reactive power adjustment rules, enabling the system to respond to photovoltaic active power fluctuations in real time under conditions of no communication, quickly correct reactive power output, and make up for the lag of offline collaborative strategies in dynamic scenarios.

[0050] In embodiments of the present invention, it further includes: Considering the difference in the feasible operating range of the inverter when photovoltaic active power output increases and decreases, the affine relationship is divided into two independent intervals: The first slope is used when the active power output increases, and the second slope is used when the active power output decreases. By introducing two non-negative auxiliary variables to increase the dimensionality of the original control variables and performing convex hull relaxation on the feasible region after the dimensionality increase, a robust optimization model containing piecewise control characteristics is constructed.

[0051] In some embodiments, when constructing local reactive power adjustment rules, the asymmetric characteristics of the photovoltaic inverter in the QP feasible region are considered. There are significant differences in its remaining reactive power capacity when active power output increases and decreases. Therefore, the original single affine relation is split into two independent intervals for separate modeling. For example, at node 30, when the actual active power of the photovoltaic system increases from the predicted value of 400kW to 520kW, the inverter's apparent power constraint tightens, and the adjustable reactive power upper limit decreases rapidly. If a fixed slope is still used at this point, the reactive power command may exceed the physical limit. However, when the active power decreases from 400kW to 280kW, the remaining capacity increases, providing stronger reactive power support and allowing for a larger reactive power adjustment range.

[0052] Therefore, a first slope k1 is used when the active power change ΔP ≥ 0, and a second slope k2 is used when ΔP < 0. Typically, |k2| > |k1| is satisfied to fully utilize the unused capacity under low active power conditions. To accurately express this piecewise characteristic in the optimization model, two non-negative auxiliary variables δ are introduced. + and δ - The original one-dimensional control variable ΔP is upgraded to a two-dimensional space (δ). + ,δ - And let ΔP = δ + -δ - Simultaneously apply δ + ·δ - The complementary constraint = 0 ensures that neither is non-zero at the same time.

[0053] Since the complementary constraint makes the problem non-convex, a convex hull relaxation is further applied to the upgraded feasible region. This involves retaining all vertices of the original piecewise feasible region (e.g., (ΔP_max, Q_min), (0, Q0), (ΔP_min, Q_max), etc.) and constructing a minimal convex polyhedron containing these vertices as an alternative feasible set. For example, under the condition of 600kW photovoltaic power at node 33 and an inverter capacity of 630kVA, the original feasible region is asymmetric and concave in the range ΔP∈[-200,+150]kW. After convex hull relaxation, it is transformed into a convex polygon bounded by five linear inequalities, preserving the key features of piecewise control while enabling efficient solution of the model. Finally, a robust optimization model is constructed based on this upgrade and relaxation results, ensuring that safe and feasible piecewise affine control parameters can still be generated under uncertain active power fluctuation scenarios.

[0054] Here, convex hull relaxation refers to using geometric methods to expand the non-convex original feasible region into a minimal convex set containing all its extreme points, thereby achieving convexization of the optimization problem while maintaining the flexibility of the control strategy.

[0055] In this embodiment of the invention, convex hull relaxation is achieved by preserving all vertices of the original segmented feasible region and constructing its minimum convex set, ensuring that the relaxed model can still accurately approximate the original non-convex control characteristics.

[0056] In some embodiments, when implementing convex hull relaxation, all extreme points of the original piecewise feasible region in the upgraded space are first identified. These extreme points correspond to the maximum or minimum reactive power output capability of the inverter under different active power output boundaries. For example, at node 28, when the active power output drops from the predicted value of 300kW to 150kW, the inverter can provide the maximum inductive reactive power +210kVar; while when the active power rises to 420kW, it can only provide capacitive reactive power -90kVar. These two operating conditions, together with the zero deviation point (ΔP=0, Q=Q0), constitute the three key vertices of the original feasible region.

[0057] Subsequently, by calculating the convex combinations of these vertices, a minimal convex polyhedron containing them is constructed. This polyhedron is precisely described by several linear inequalities. For example, three boundary hyperplanes are generated using vertices (-150, 210), (0, 150), and (+120, -90), forming a triangular convex hull. This convex hull completely covers the original non-convex region and does not introduce redundant space far from the physically feasible boundary. This mathematically ensures that the relaxed model can be solved efficiently without causing control commands to exceed limits due to excessive relaxation.

[0058] During the optimization process, all decision variables are constrained within this convex hull to ensure that the final piecewise affine control parameters can be mapped back to the actual operable QP region of the inverter under any active power fluctuation scenario.

[0059] Here, the minimum convex set refers to the convex hull spanned by all vertices of the original piecewise feasible region. Its boundary is composed of line segments or planes connecting adjacent vertices, and it is the most compact convex relaxation form that approximates the original non-convex property.

[0060] In embodiments of the present invention, it further includes: Dual transformation is performed on the voltage deviation uncertainty constraint in the robust optimization model to transform it into a linear constraint form, and the piecewise affine control parameters corresponding to each photovoltaic node are obtained by solving the solution. During system operation, reactive power is dynamically adjusted using segmented affine control parameters based on real-time measured changes in active power to suppress voltage overshoot.

[0061] In this embodiment of the invention, the dual transformation is used for the worst-case power fluctuation scenario. By using dual variables, the robust constraints containing maximum value operations are equivalently transformed into a set of linear inequality constraints, thereby obtaining analytical control parameters deployed on the local controller.

[0062] In some embodiments, when solving the robust optimization model, a dual transformation technique is used to convert the worst-case power fluctuation scenario included in the voltage deviation uncertainty constraint into a computable linear form. For example, in the cluster where node 24 is located, there is a ±10% prediction error range between the load and photovoltaic output, resulting in the node voltage deviation expression containing the max{·} operation. Direct solution would lead to nonconvexity. To address this, a dual variable is introduced corresponding to the extreme point of the uncertain set, and the original robust constraint containing the maximum value is equivalently transformed into a set of linear inequalities about the control variable and the dual variable.

[0063] This process, based on the duality theory of linear programming, explicitly expresses the support function of the original uncertainty set (such as a hypercube or ellipsoid set) and eliminates the inner-layer maximization operation through strong duality conditions, ultimately obtaining a linear constraint system containing only piecewise affine control parameters. For example, the original constraint form "for all ΔP∈[-ΔP_max,ΔP_max], |V_i-1.0|≤ε" is transformed into three linear constraints: one corresponding to ΔP=+ΔP_max, one corresponding to ΔP=-ΔP_max, and another corresponding to ΔP=0. Each constraint is embedded into the objective function or constraint system after being weighted by the corresponding dual variable weights.

[0064] The control parameters k1 and k2 obtained from this solution are the slopes in the two active power change directions, which can be directly deployed in the local controller.

[0065] During actual system operation, each photovoltaic node collects local active power in real time and calculates its change ΔP relative to the predicted value. If ΔP ≥ 0, reactive power is output according to Q = Q0 + k1·ΔP; if ΔP < 0, reactive power is output according to Q = Q0 + k2·ΔP, thereby dynamically responding to voltage fluctuations caused by sudden changes in irradiance or cloud shadows. For example, when the measured active power of node 31 drops sharply from 350kW to 220kW, ΔP = -130kW. The controller immediately calls k2 = -0.35 to generate an inductive reactive power injection of ΔQ = +45.5kVar, effectively preventing the voltage from dropping below 0.97pu.

[0066] Here, the analytical control parameters refer to the explicit piecewise linear function coefficients of the reactive power command that can be directly calculated based on local measurements without online optimization. Their values ​​are uniquely determined by offline robust optimization combined with dual transformation, ensuring that the voltage safety boundary is still met even in the worst-case fluctuation scenario.

[0067] In summary, the beneficial effects of this invention are that it proposes a distributed-local two-stage voltage control method. The distributed photovoltaic (PV) cluster aims to minimize the voltage deviation within the cluster by solving for the reactive power output of the PV inverters within the cluster. Furthermore, it uses an alternating direction multiplier algorithm to iteratively interact with the cluster boundary variables, thereby achieving synergy in reactive power control strategies among different PV clusters. In the local control stage, considering the strong volatility of distributed PV, the reactive power control strategy obtained in the distributed coordination stage may become inapplicable. Each peripheral cluster uses the reactive power strategy from the distributed coordination stage as a base value to formulate a QP affine control strategy. This strategy adjusts the reactive power output in real time according to the active power fluctuations of the PV, and further improves the flexibility of the local control strategy by segmenting the control curve using a spatial dimensionality increase method.

[0068] Example 2, refer to Figures 2-8 Based on the above embodiments, a specific implementation of a distributed-local two-stage voltage control method can be designed as follows: As attached Figure 2 As shown, the distribution network topology, line parameters, distributed photovoltaic clusters and load information are input, and the alternating direction multiplier algorithm is used to achieve the coordinated optimization of reactive power strategies of different distributed photovoltaic clusters. The voltage control model for photovoltaic clusters at the edge of the distribution network can be represented as follows: (1) (2) In the formula, Represents the network trend vector; This represents the vector of distributed photovoltaic control parameters to be optimized. Represents photovoltaic clusters Objective function; equation and inequalities Represents a cluster System security operation constraints within the system.

[0069] Assumption These are global state variables related to the boundary nodes. Specifically, the power injection variables of the boundary nodes and the power of the connecting lines are considered as coordination variables for inter-cluster interaction and iteration.

[0070] To further accelerate the convergence speed and eliminate global variables Further introduce auxiliary variables This achieves a fully distributed solution. Specifically, the global variable coordination and update steps for each photovoltaic cluster are as follows: (3) In the formula, and For cluster The feasible set of internal variables.

[0071] Criterion for iterative convergence Through the original residual and dual residual To determine, see below.

[0072] (4) (5) (6) In the formula, The iteration convergence limit is set to 0.001.

[0073] Furthermore, based on the reactive power strategy obtained from the distributed solution, a QP local control strategy is adopted to achieve real-time control of the reactive power output of the distributed photovoltaic inverter. The specific control strategy is as follows: (7) (8) In the formula, For nodes The change in actual photovoltaic output relative to the predicted value; The initial reactive power strategy for the photovoltaic inverter obtained by distributed solution; For nodes The QP affine coefficient for reactive power adjustment after changes in photovoltaic active power; For nodes Reactive power adjustment of photovoltaic inverter; To adjust the nodes in real time The reactive power output of the photovoltaic inverter.

[0074] Meanwhile, by utilizing the distribution network operating point determined during the distributed collaborative control phase, the voltage sensitivity related to photovoltaic power changes can be obtained, as shown below: (9) In the formula, and These represent the active and reactive power sensitivities of the node voltage, respectively. This refers to the number of distributed photovoltaic systems.

[0075] Each photovoltaic cluster, with the objective of minimizing system voltage deviation, can obtain the following robust optimization model: (10) st (11) (12) (13) In the formula, the objective function (4) is the sum of the absolute values ​​of the system voltage deviations; the constraints (5) and (6) give the photovoltaic power fluctuations. Node in case Robust constraint on voltage deviation; constraint (7) is an inverter capacity constraint; This represents the number of nodes within the cluster. For nodes Forecasted photovoltaic power output; For nodes Capacity of photovoltaic inverter.

[0076] Furthermore, based on the local QP control strategy, the photovoltaic output is segmented according to different situations of increase or decrease, and the optimization model is transformed into a dual model for easy solution. Considering the significant differences in the feasible region of the inverter's QP plane when photovoltaic active power output increases or decreases, to improve the flexibility of the control strategy, the control curve is divided into two segments, each using QP affine control with different slopes. Let the segmentation point be... This allows us to extend the control variables from a one-dimensional space to a two-dimensional space: (14) Variables after dimensionality upgrade With the original variable The relationship between them is as follows: (15) Then, controlling photovoltaic power through the QP strategy can be expressed as: (16) To simplify the subsequent optimization problem, the original domain (8) is subjected to convex relaxation, which expands it into a convex hull containing the original domain. This convex hull is a minimal convex set consisting of all vertices of the original domain, and the expression for the corresponding convex hull is: (17) Equation (11) can be further transformed into: (18) (19) (20) Therefore, the above robust local control model can be transformed into (twenty one) st (twenty two) (twenty three) (twenty four) (25) (26) In the formula, the objective function (21) is the sum of the absolute values ​​of the system voltage deviations; constraints (22) and (23) give the power fluctuations. Node in case Robust constraint for voltage deviation; constraint (24) is the inverter capacity constraint; constraints (25) and (26) give the variables of higher dimension. The relevant constraints.

[0077] Due to the existence of robust constraints, the above optimization problem is difficult to solve directly and requires further processing. For constraint (22), it can be equivalent to... (27) To ensure robustness, the right side of equation (21) is transformed into the following optimization problem: (28) st (29) The dual problem of the above optimization problem is: (30) st (31) (32) In the formula, As dual variables; Let be a vector in which all elements are 1. Similarly, constraint (23) can be transformed into a dual problem. By solving the dual problem of the original problem, the nonlinear part of the optimization problem is eliminated. .

[0078] Furthermore, the local QP control strategy obtained from the optimized solution is used to adjust the reactive power output in real time according to the changes in active power, thereby realizing real-time voltage reactive power control.

[0079] A test case was constructed based on the improved IEEE 33-node architecture to verify the voltage control effectiveness of the proposed control method. The test case structure is attached. Figure 3 As shown, the system voltage level is 12.66 kV, comprising a total of 32 branches operating radially. The total active power of the load is 3.715 MW, and the total reactive power is 2.3 Mvar. The predicted distributed photovoltaic and load output curves are attached. Figure 4 As shown. The safe operating upper and lower limits for the voltage amplitude of each node are set to 1.10 and 0.90, respectively; the optimized upper and lower limits for the voltage amplitude of each node are set to 1.02 and 0.98, respectively.

[0080] To fully consider the impact of high-penetration distributed photovoltaic (PV) grid connection on system operation, 12 distributed PV groups were connected to the IEEE 33-node system. The specific connection locations and capacities are shown in Table 1.

[0081] Table 1 Distributed PV Capacity and Grid Connection Location

[0082] The test program was developed using Matlab software and solved the optimization problem using the Gurobi optimization algorithm package. The computer hardware environment for performing the optimization calculations was an Intel(R) Core(TM) i7-9750H CPU with a clock speed of 2.60GHz and 16GB of memory; the software environment was a Windows 10 operating system.

[0083] This section selects the following four schemes to verify the effectiveness of the proposed distributed collaborative voltage control method for photovoltaic clusters on the edge of the distribution network.

[0084] Option I: Do not optimize the reactive power strategy of the photovoltaic cluster to obtain the initial operating state of the distribution network; Option II: Use the ADMM algorithm to achieve collaborative optimization of different distributed photovoltaic clusters; Option III: The ADMM algorithm is used to achieve collaborative optimization of different photovoltaic clusters, and reactive power output is adjusted in real time based on the linear affine rule; Option IV: A centralized optimization method is used to solve the reactive power output of the photovoltaic cluster in real time, so as to achieve the theoretically optimal control effect.

[0085] Table 2 shows a comparison of the optimization results for schemes II, III, and IV. The voltage distribution under different control schemes is as follows: Figures 5-8 As shown.

[0086] To intuitively reflect the optimization effect of voltage in each time period, the average voltage deviation (AVD) index is introduced to calculate the average voltage deviation of the whole system, as shown in Equation (33).

[0087] (33) The effectiveness of the ADMM-based distributed collaborative optimization method was verified by comparing schemes I and II. Without control, the integration of distributed photovoltaic (PV) systems would cause severe voltage fluctuations and even exceed limits. By employing a distributed optimization algorithm to collaboratively optimize the reactive power output strategies of PV inverters in different edge clusters: when the system voltage is low, the PV inverters in different edge clusters effectively support the node voltage by emitting reactive power; when the system voltage is high, the inverters absorb reactive power to reduce the voltage, maintaining the system voltage at a safe operating level.

[0088] Table 2 Comparison of the operational optimization effects of different control strategies

[0089] The effectiveness of the local affine control strategy based on linear upsegmentation was verified by comparing Schemes II and III. Since photovoltaic power output always fluctuates in real time during actual operation, it is necessary to adjust the reactive power output of the inverters according to these fluctuations. Therefore, Scheme III, based on the coordinated reactive power output strategies of each edge cluster, adjusts the reactive power output of the photovoltaic inverters through affine control. The comparison shows that, compared to Scheme II which only uses ADMM to coordinate reactive power output between different clusters, Scheme III's real-time adjustment of inverter reactive power results in a smaller voltage fluctuation range in the distribution network under its control strategy. Furthermore, Table 2 shows that, compared to Scheme II, the AVD index in Scheme III decreases from 0.5683 to 0.4417, indicating that Scheme III has better voltage control performance.

[0090] The comparison between Schemes III and IV fully demonstrates the optimization effect of the proposed distributed cooperative voltage control method. Scheme IV, based on a centralized method, optimizes the reactive power output of the inverter, achieving a globally optimal operational optimization effect, which is the theoretical optimal solution but difficult to achieve in actual operation. With the cooperation of ADMM cooperative optimization and local affine control strategy, Scheme III effectively realizes the coordination of different edge clusters and local adjustment for real-time active power fluctuations, achieving similar optimization performance to the centralized control method.

[0091] In summary, the distributed-local two-stage voltage control method for distributed photovoltaic clusters in distribution networks proposed in this invention can achieve efficient control of the reactive power output of distributed photovoltaics and limit the voltage of the distribution network within a reasonable range, effectively addressing the voltage over-limit problem of the distribution network under high-penetration distributed photovoltaic access.

[0092] Example 3, referring to Figure 9 This embodiment also provides a distributed-local two-stage voltage control system, including: The cluster partitioning module is used to divide the distribution network into multiple photovoltaic cluster areas based on the network topology, line parameters, access location and capacity of each distributed photovoltaic cluster, and load data. The first optimization model building module is used to build a local optimization model for each cluster, with the goal of minimizing the sum of the absolute values ​​of the voltage deviations of the nodes within the cluster, and includes power flow constraints, voltage safety constraints, and inverter capacity constraints. The numerical calculation module is used to iteratively exchange the coordination variables of the boundary nodes between the clusters through the alternating direction multiplier algorithm, complete the collaborative optimization of the reactive power output strategy of the multi-cluster, and obtain the initial reactive power output setpoint of the inverter of each photovoltaic node. The correction module is used to configure local reactive power adjustment rules for each photovoltaic node based on the initial reactive power output setpoint. These rules correct the reactive power output in real time according to the affine relationship based on the change in the actual active power output of the photovoltaic relative to the predicted value. The second optimization model building module is used to consider the difference in the feasible operating region of the inverter when the photovoltaic active power output increases and decreases, and divides the affine relationship into two independent intervals: The first slope is used when the active power output increases, and the second slope is used when the active power output decreases. By introducing two non-negative auxiliary variables to increase the dimensionality of the original control variables and performing convex hull relaxation on the feasible region after the dimensionality increase, a robust optimization model containing piecewise control characteristics is constructed. The solver module is used to perform dual transformation on the voltage deviation uncertainty constraint in the robust optimization model, transform it into a linear constraint form, and solve for the piecewise affine control parameters corresponding to each photovoltaic node. The adjustment module is used to dynamically adjust the reactive power output based on the real-time measured changes in active power during system operation, using segmented affine control parameters to suppress voltage over-limit.

[0093] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0094] This embodiment also provides an electronic device, which can be a terminal, and its internal structure diagram can be as follows: Figure 9 As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a distributed-local two-stage voltage control method. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0095] This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps: Based on the network topology, line parameters, access location and capacity of each distributed photovoltaic cluster, and load data, the distribution network is divided into multiple photovoltaic cluster areas. For each cluster, a local optimization model is established with the goal of minimizing the sum of the absolute values ​​of the voltage deviations of the nodes within the cluster. This model includes power flow constraints, voltage safety constraints, and inverter capacity constraints. By iteratively exchanging the coordination variables of the boundary nodes among the clusters using the alternating direction multiplier algorithm, the collaborative optimization of the reactive power output strategy of the multi-cluster is completed, and the initial reactive power output setpoint of the inverter of each photovoltaic node is obtained. Based on the initial reactive power output setpoint, a local reactive power adjustment rule is configured for each photovoltaic node. This rule corrects the reactive power output in real time according to the change in the actual active power output of the photovoltaic system relative to the predicted value, following an affine relationship.

[0096] Considering the difference in the feasible operating range of the inverter when photovoltaic active power output increases and decreases, the affine relationship is divided into two independent intervals: The first slope is used when the active power output increases, and the second slope is used when the active power output decreases. By introducing two non-negative auxiliary variables to increase the dimensionality of the original control variables and performing convex hull relaxation on the feasible region after the dimensionality increase, a robust optimization model containing piecewise control characteristics is constructed.

[0097] Dual transformation is performed on the voltage deviation uncertainty constraint in the robust optimization model to transform it into a linear constraint form, and the piecewise affine control parameters corresponding to each photovoltaic node are obtained by solving the solution. During system operation, reactive power is dynamically adjusted using segmented affine control parameters based on real-time measured changes in active power to suppress voltage overshoot.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0099] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0100] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A distributed-local two-stage voltage control method, characterized in that, include: Based on the network topology, line parameters, access location and capacity of each distributed photovoltaic cluster, and load data, the distribution network is divided into multiple photovoltaic cluster areas. For each cluster, a local optimization model is established with the goal of minimizing the sum of the absolute values ​​of the voltage deviations of the nodes within the cluster. This model includes power flow constraints, voltage safety constraints, and inverter capacity constraints. By iteratively exchanging the coordination variables of the boundary nodes among the clusters using the alternating direction multiplier algorithm, the collaborative optimization of the reactive power output strategy of the multi-cluster is completed, and the initial reactive power output setpoint of the inverter of each photovoltaic node is obtained. Based on the initial reactive power output setting value, a local reactive power adjustment rule is configured for each photovoltaic node. This rule corrects the reactive power output in real time according to the change in the actual active power output of the photovoltaic system relative to the predicted value, following an affine relationship.

2. The distributed-local two-stage voltage control method as described in claim 1, characterized in that, Also includes: Considering the difference in the feasible operating range of the inverter when photovoltaic active power output increases and decreases, the affine relationship is divided into two independent intervals: The first slope is used when the active power output increases, and the second slope is used when the active power output decreases. By introducing two non-negative auxiliary variables to increase the dimensionality of the original control variables and performing convex hull relaxation on the feasible region after the dimensionality increase, a robust optimization model containing piecewise control characteristics is constructed.

3. The distributed-local two-stage voltage control method as described in claim 2, characterized in that, Also includes: The voltage deviation uncertainty constraint in the robust optimization model is transformed into a linear constraint by performing a dual transformation, and the piecewise affine control parameters corresponding to each photovoltaic node are obtained by solving the solution. During system operation, reactive power is dynamically adjusted based on the real-time measured changes in active power, using the segmented affine control parameters to suppress voltage overshoot.

4. The distributed-local two-stage voltage control method as described in claim 3, characterized in that, The coordination variable of the boundary node includes the injected power of the boundary node or the transmission power of the connected branch. Each cluster only needs to exchange this coordination variable with its neighboring clusters, without relying on the global central controller.

5. The distributed-local two-stage voltage control method as described in claim 4, characterized in that, The reference point of the affine relationship is determined by the initial reactive power output setting value, and its slope reflects the reactive power adjustment caused by the unit change in active power.

6. The distributed-local two-stage voltage control method as described in claim 5, characterized in that, The convex hull relaxation is achieved by preserving all vertices of the original segmented feasible region and constructing its minimum convex set, ensuring that the relaxed model can still accurately approximate the original non-convex control characteristics.

7. The distributed-local two-stage voltage control method as described in claim 6, characterized in that, The dual transformation addresses the worst-case power fluctuation scenario by using dual variables to convert robust constraints involving maximum value operations into a set of linear inequality constraints, thereby obtaining analytical control parameters deployed on the local controller.

8. A distributed-local two-stage voltage control system, employing the method described in any one of claims 1 to 7, characterized in that, include: The cluster partitioning module is used to divide the distribution network into multiple photovoltaic cluster areas based on the network topology, line parameters, access location and capacity of each distributed photovoltaic cluster, and load data. The first optimization model building module is used to build a local optimization model for each cluster, with the goal of minimizing the sum of the absolute values ​​of the voltage deviations of the nodes within the cluster, and includes power flow constraints, voltage safety constraints, and inverter capacity constraints. The numerical calculation module is used to iteratively exchange the coordination variables of the boundary nodes between the clusters through the alternating direction multiplier algorithm, complete the collaborative optimization of the reactive power output strategy of the multi-cluster, and obtain the initial reactive power output setpoint of the inverter of each photovoltaic node. The correction module is used to configure local reactive power adjustment rules for each photovoltaic node based on the initial reactive power output setting value. The rules correct the reactive power output in real time according to the change of the actual active power output of the photovoltaic relative to the predicted value, based on an affine relationship. The second optimization model building module is used to divide the affine relationship into two independent intervals, taking into account the difference in the feasible operating region of the inverter when the photovoltaic active power output increases and decreases: The first slope is used when the active power output increases, and the second slope is used when the active power output decreases. By introducing two non-negative auxiliary variables to increase the dimensionality of the original control variables and performing convex hull relaxation on the feasible region after the dimensionality increase, a robust optimization model containing piecewise control characteristics is constructed. The solution module is used to perform dual transformation on the voltage deviation uncertainty constraint in the robust optimization model, transform it into a linear constraint form, and solve for the piecewise affine control parameters corresponding to each photovoltaic node. The adjustment module is used to dynamically adjust the reactive power output based on the real-time measured changes in active power during system operation, using the segmented affine control parameters, in order to suppress voltage over-limit.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the distributed-local two-stage voltage control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the distributed-local two-stage voltage control method according to any one of claims 1 to 7.

Images