A multi-parameter joint optimization control system for photovoltaic power distribution cluster

Abstract

The application discloses a kind of photovoltaic power distribution cluster multi-parameter joint optimization control systems, it is related to photovoltaic power distribution cluster optimization control technical field, the present application is aimed at the problem of high penetration rate photovoltaic cluster voltage out-of-limit, regulation accuracy and stability is insufficient, the system is constructed by electrical state data set, voltage-power sensitivity matrix generation, voltage responsibility weight quantization, regulation instruction generation and instruction issuing module collaborative work: with power distribution transformer low-voltage side as boundary aggregation data, according to photovoltaic penetration rate self-adaptive selection power flow calculation method;Precise tracing is constructed by three-dimensional responsibility weight quantization model;Distinguish remote / local dominant type out-of-limit and differential regulation, priority reactive power regulation, active power is cut when reactive power margin is insufficient;After collaborative checking and time sequence optimization, issue instruction, cooperate closed-loop verification and failure differential treatment, realize control precision, high efficiency and stabilization, adapt to the operation demand of large-scale photovoltaic power distribution cluster.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power distribution cluster optimization control technology, specifically to a multi-parameter joint optimization control system for photovoltaic power distribution clusters. Background Technology

[0002] With the energy structure shifting towards clean and low-carbon energy, distributed photovoltaic (PV) power, with its advantages of flexible deployment and environmental efficiency, has seen its penetration rate in power distribution networks continue to increase, forming large-scale PV power distribution clusters. While the widespread application of PV power distribution clusters has effectively alleviated the pressure of traditional fossil fuel consumption, it has also brought severe challenges to power distribution network voltage control: distributed PV power output is highly volatile and intermittent, and since it is mostly concentrated on the low-voltage side of the power distribution network, it is prone to problems such as node voltage rise and exceeding limits, seriously affecting the power supply quality and safe and stable operation of the power distribution network.

[0003] Existing technologies, such as patent applications with publication numbers CN103326351B, CN119765520B and CN119965886B related to the optimization control of photovoltaic power distribution clusters, show that the existing photovoltaic power distribution cluster voltage control systems have the following technical shortcomings: (1) Traditional schemes mostly adopt centralized power flow calculation strategies, without considering the impact of dynamic changes in photovoltaic penetration on calculation efficiency and accuracy. When the number of photovoltaic nodes is large, they cannot adapt to the online real-time adjustment requirements; (2) The responsibility attribution is based only on the sensitivity coefficient, without considering the impact of electrical distance attenuation and power backfeed scenarios, which increases the error of responsibility attribution. Some schemes only judge the impact of photovoltaic nodes on over-limit voltage through the sensitivity coefficient, which leads to the overestimation or underestimation of the responsibility weight of remote photovoltaic nodes and the blurring of the responsibility boundary between near-end nodes and remote nodes. At the same time, it does not distinguish between the aggravating and suppressing effects of photovoltaic node power injection on voltage over-limit, and it is easy to misjudge the node that alleviates voltage over-limit as the responsible party. (3) It does not distinguish between remote / local dominant over-limit, and uniformly adopts local regulation, which increases the probability of voltage over-limit regulation failure caused by cross-regional transmission. Summary of the Invention

[0004] To address the aforementioned technical shortcomings, the present invention aims to provide a multi-parameter joint optimization control system for photovoltaic power distribution clusters.

[0005] To solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a multi-parameter joint optimization control system for photovoltaic power distribution clusters, including: an electrical status dataset construction module, used to collect voltage, active power and reactive power data of each node in the power distribution cluster in real time from SCADA system and photovoltaic monitoring system, and construct a real-time electrical status dataset for the entire network.

[0006] The voltage-power sensitivity matrix generation module is used to generate a voltage-power sensitivity matrix that characterizes the voltage response relationship between each node to injected power, based on the real-time electrical state dataset of the entire network and in combination with the pre-stored line topology and line parameters, and adaptively selects the Newton-Raphson method or the forward-backward substitution method according to the real-time photovoltaic penetration rate.

[0007] The voltage responsibility weight quantization module is used to input the voltage exceedance magnitude of any node and the voltage-power sensitivity matrix into the responsibility tracing calculation model that includes a sign direction factor, a distance attenuation factor and a power backfeed criterion when any node voltage exceeds the limit, and output the voltage responsibility weight quantization result of each photovoltaic node for the over-limit event.

[0008] The adjustment instruction generation module is used to sort and compare the voltage responsibility weights according to the judgment rule of prioritizing responsibility weights and secondarily prioritizing adjustment costs. If it is determined that the responsibility weight of a certain remote photovoltaic node is higher than that of the local node, the preventive adjustment instruction generation process for the remote photovoltaic node is triggered.

[0009] The instruction issuance module is used to perform coordination verification and timing optimization on the adjustment instruction sequences corresponding to multiple photovoltaic nodes to be adjusted, generate a step-by-step and hierarchical collaborative control instruction set, and issue it to the corresponding photovoltaic inverter for execution.

[0010] The beneficial effects of the present invention are as follows: (1) The present invention adopts a distributed solution strategy, takes the low-voltage side of the distribution transformer as the partition unit to calculate the local power flow in parallel, and calibrates the data through the power-voltage boundary conditions of the tie node, thus avoiding the efficiency bottleneck of centralized calculation of the whole network; at the same time, it adaptively selects the Newton-Raphson method or the forward-backward substitution method according to the real-time photovoltaic penetration rate, ensuring the convergence accuracy in complex network scenarios and improving the calculation speed in large-scale sparse network scenarios, so that the calculation delay is controlled within the second level, adapting to the online real-time adjustment needs of high-penetration photovoltaic clusters, and improving the calculation efficiency compared with the traditional centralized method.

[0011] (2) This invention constructs a three-dimensional responsibility weight quantification model consisting of a sign direction factor, a distance attenuation factor, and a power backfeeding criterion. This model quantifies the responsibility of photovoltaic nodes for voltage over-limit from multiple dimensions, including effect intensity, influence correlation, direction matching, distance attenuation, and operating conditions. This model effectively distinguishes between the aggravating and suppressing effects of photovoltaic nodes on over-limit, weakens the unreasonable responsibility weight of power backfeeding nodes, reduces attribution error control, and solves the problems of ambiguous responsibility boundaries and high misjudgment rates caused by traditional single-dimensional tracing, providing a reliable basis for precise regulation.

[0012] (3) This invention establishes a remote / local-dominated limit violation judgment mechanism. It filters the adjustment targets by geographical distance, topological location and responsibility weight ratio. For remote-dominated limit violations, preventive adjustment is initiated to offset the cross-regional transmission loss and delay, thereby improving the adjustment success rate in this type of scenario. For non-remote-dominated limit violations, local adjustment is adopted, which does not require cross-regional coordination and ensures the adjustment response speed. At the same time, reactive power adjustment is given priority. When the reactive power margin is insufficient, active power output is reduced proportionally, which balances the adjustment effect and the curtailment loss and improves the adaptability of the local adjustment strategy. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0014] Figure 1 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Reference Figure 1 As shown, the present invention provides a multi-parameter joint optimization control system for photovoltaic power distribution clusters, including: an electrical status dataset construction module, used to collect voltage, active power and reactive power data of each node in the power distribution cluster in real time from the SCADA system and the photovoltaic monitoring system, and construct a real-time electrical status dataset for the entire network.

[0017] In a specific embodiment of the present invention, the real-time acquisition process synchronously acquires feeder node measurement data in the SCADA system and inverter operation data in the photovoltaic monitoring system through a standardized communication protocol to ensure timestamp alignment.

[0018] It should be noted that the feeder node measurement data includes feeder node voltage, active power, reactive power and other measurement data, and the inverter operation data includes inverter output power, operating status and other operation data.

[0019] The validity of the collected raw data is verified, and invalid records due to communication interruption, sensor drift, or abnormal values ​​are removed.

[0020] Using the low-voltage side of the distribution transformer as the boundary, a unified electrical status data frame is formed that includes all photovoltaic access points and key load nodes.

[0021] Specifically, taking the low-voltage side of the distribution transformer as the boundary, based on the pre-established transformer-photovoltaic access point-load node association mapping table, the relevant data of all photovoltaic access points and key load nodes within this boundary are aggregated to form a unified electrical status data frame containing transformer number, node unique identifier, timestamp, electrical parameter type and value.

[0022] It should be noted that the low-voltage side of the distribution transformer specifically refers to the line closest to the power consumption side of the distribution transformer as the dividing boundary. The function of the distribution transformer is to step down the voltage, reducing the high voltage of the power grid to a low voltage that can be used by users or equipment. The low-voltage side is the side of the distribution transformer that outputs low-voltage electricity, that is, the side closer to the photovoltaic access point and key load nodes. The high-voltage side is connected to the upstream power grid and is far away from the end users. It is used as the boundary because the photovoltaic nodes and load nodes covered by the low-voltage side of the same transformer belong to the same power supply area, with strong data correlation, which facilitates centralized management and calculation.

[0023] The transformer-PV access point-load node association mapping table is constructed using distribution transformer ledger information, clarifying the affiliation of PV access points and key load nodes covered by each transformer's low-voltage side.

[0024] It should be noted that the critical load nodes specifically refer to power consumption nodes in the distribution network that have high requirements for power supply reliability and voltage quality, such as nodes where industrial production loads, important residential area loads, and public service facility loads are located.

[0025] The data frames are continuously updated according to a set sampling period and serve as the sole input source for subsequent power flow calculations and sensitivity analysis, thereby ensuring the integrity and timeliness of the entire network's state awareness.

[0026] The voltage-power sensitivity matrix generation module is used to generate a voltage-power sensitivity matrix that characterizes the voltage response relationship between each node to injected power, based on the real-time electrical state dataset of the entire network and in combination with the pre-stored line topology and line parameters, and adaptively selects the Newton-Raphson method or the forward-backward substitution method according to the real-time photovoltaic penetration rate.

[0027] In a specific embodiment of the present invention, the power flow calculation adopts a distributed solution strategy based on the Newton-Raphson method or the forward-backward substitution method. Specifically, the calculation method is automatically selected according to the real-time photovoltaic penetration rate. When the photovoltaic penetration rate is greater than or equal to the preset photovoltaic penetration rate, the forward-backward substitution method is selected, and when the photovoltaic penetration rate is less than the preset photovoltaic penetration rate, the Newton-Raphson method is selected.

[0028] It should be noted that the advantages of the forward-backward substitution method are low computational cost and fast response in scenarios with many distributed nodes and sparse networks, while the advantages of the Newton-Raphson method are high convergence accuracy when the network structure is complex and the nodes are highly correlated. The preset photovoltaic penetration rate is determined by simulation tests on typical distribution network topologies and is not arbitrarily set. The adaptation threshold determined based on the performance inflection points of the two calculation methods can be reproduced by those skilled in the art through conventional simulation tests to demonstrate the rationality of the threshold.

[0029] It should also be noted that the photovoltaic penetration rate is the ratio of photovoltaic access capacity to total load capacity.

[0030] The distributed solution strategy uses the low-voltage side of the distribution transformer as the partition unit, dividing the distribution network into multiple independent calculation sub-regions. Each sub-region calculates the local power flow in parallel, and the power-voltage boundary condition data of the tie nodes are used to aggregate the power flow results of the entire network.

[0031] It should be noted that the distributed solution strategy is adopted to avoid the efficiency bottleneck of centralized computing across the entire network, especially when there are many photovoltaic nodes and the cluster size is large. At the same time, it ensures the specificity of computing for each partition and the stronger correlation between nodes covered by the same transformer.

[0032] It should be noted that a tie node is a connection node between different sub-regions, i.e., different low-voltage side zones of distribution transformers. For example, if sub-region A and sub-region B are connected by a line, the nodes at both ends of this line are tie nodes, which are equivalent to the data interfaces between the two zones. Power-voltage boundary conditions: After each sub-region calculates its local power flow, it will obtain two key data points for its tie node: the power flowing through the tie node and the voltage of the tie node, including amplitude and phase angle. These two data points are the boundary conditions. The data calculated by both sides must be consistent; otherwise, the results for the entire network will be contradictory. For example, the data exchange process is as follows: If sub-region A calculates the power of tie node N to be 30kW and the voltage to be 0.405kV, and sub-region B calculates the power of the same node N to be 28kW and the voltage to be 0.408kV, the two regions will exchange data through a communication protocol, calibrate based on line parameters, and finally unify them to consistent values, such as 30kW and 0.405kV, to ensure that the data connection between the zones is error-free.

[0033] Using pre-stored line impedance parameters, topology connections, and node type information, a distribution network admittance matrix adapted to distributed partitioning is constructed. The voltage amplitude, phase angle, and branch power distribution data of each node are obtained by solving the admittance matrix.

[0034] It should be noted that line impedance parameters refer to the resistance and reactance of distribution network lines, which are key electrical characteristics during power transmission. Resistance leads to power loss, and reactance affects voltage phase. These are the basic physical parameters for calculating voltage and power. Topology connection relationship refers to the connection method between all nodes in a partition, such as photovoltaic nodes, load nodes, tie nodes, and lines. For example, photovoltaic node A is connected to feeder node B through line L1, and feeder node B is connected to load node C through line L2. This is equivalent to the connection map of the power grid, which determines the power transmission path. Node type information refers to the attributes of each node, such as photovoltaic node, load node, and tie node. Photovoltaic nodes are the power generation end that outputs power, load nodes are the power consumption end that consumes power, and tie nodes are the partition interface that transmits power. The node type determines the power flow and calculation logic. For example, the power of photovoltaic nodes is outflow, and the power of load nodes is inflow.

[0035] It should also be noted that the admittance matrix is ​​essentially a mathematical table, i.e., a matrix, that quantifies the electrical relationships between all nodes within a partition. The dimension of the matrix is ​​the total number of nodes in the partition multiplied by the total number of nodes in the partition. For example, if there are 20 nodes in a partition, it is a 20*20 matrix. Each element represents the electrical admittance between nodes, where admittance = 1 / impedance, reflecting the ease of power transmission between nodes. The larger the admittance, the easier the transmission. The admittance matrix is ​​also constructed partition by partition. Each admittance matrix only contains the node and line information within the corresponding partition, avoiding excessive dimensionality of the overall network admittance matrix, which would lead to low computational efficiency. By transforming the physical connections and electrical parameters of the power grid into a mathematical model that can be recognized by a computer, the solution process is simplified.

[0036] It should be noted again that the admittance matrix contains all the key information of the partition. Substituting it into the power flow equation of the distribution network, the unknowns of the equation can be solved through numerical calculation, i.e., the Newton-Raphson method or the forward-backward substitution method. These unknowns include the actual voltage magnitude, voltage phase angle, and branch power distribution data of each node, including the transmission direction and specific value of active and reactive power of each line. The voltage phase angle reflects the phase relationship of the voltage and is the key to calculating the power transmission direction of the line. The branch power distribution data reflects the flow path and distribution of electrical energy within the partition.

[0037] The power flow equations are equations that describe the laws of electrical energy transmission and are the foundation of power system analysis.

[0038] By using the differential perturbation method, a unit active / reactive power perturbation is applied to each photovoltaic node, and the voltage change of each node before and after the perturbation is calculated. The sensitivity coefficient is formed by the ratio of voltage change to power perturbation. A voltage-power sensitivity matrix is ​​constructed according to the dimensions of photovoltaic node - affected node - sensitivity coefficient.

[0039] It should be noted that the logic of applying the disturbance separately is to apply the disturbance to only one photovoltaic node at a time, while keeping the power of all other photovoltaic nodes unchanged. This avoids the cumulative effect caused by disturbing multiple nodes at the same time, which would make it impossible to distinguish the role of a single node.

[0040] It should be noted that the specific implementation process of constructing the voltage-power sensitivity matrix using the differential perturbation method is as follows: Define power disturbance amplitude , Rated active power of photovoltaic inverter, , The rated reactive power of the photovoltaic inverter is given, and the disturbance step size is given. This ensures that disturbances remain within the system's stable operating range and avoids affecting the normal power supply of the distribution network.

[0041] For each photovoltaic access node, an active power perturbation is applied individually in sequence. Reactive power disturbance Keep the power output of other nodes constant; collect the voltage amplitude changes of each node before and after the disturbance in real time through the SCADA system. .

[0042] According to the formula Let i be the affected node number and j be the photovoltaic disturbance node number. Calculate the active power-voltage sensitivity coefficient using the formula... Calculate the reactive power-voltage sensitivity coefficient.

[0043] The affected nodes are represented by row vectors, which include all feeder nodes and photovoltaic access nodes in the entire network. The column vectors represent the photovoltaic disturbance nodes in the entire network. The matrix elements are the sensitivity coefficients of the affected nodes in the corresponding rows and the photovoltaic nodes in the corresponding columns, forming a voltage-power sensitivity matrix with dimensions M×N, where M is the total number of nodes in the entire network and N is the total number of photovoltaic access nodes.

[0044] The voltage-power sensitivity matrix is ​​stored in a sparse form, retaining only the non-zero elements and their corresponding node indices for subsequent efficient accountability calculations.

[0045] In a specific embodiment of the present invention, the voltage-power sensitivity matrix generation module supports online and offline dual-mode operation: in online mode, the power flow and voltage-power sensitivity matrix are updated in real time at a cycle of seconds or minutes, which is suitable for scenarios with rapid voltage fluctuations.

[0046] In offline mode, a multi-condition sensitivity library is generated in batches based on historical typical daily data for training the responsibility tracing model or rehearsing the adjustment strategy. The two modes are managed uniformly by the dispatch center and automatically switched according to the power grid operation status to ensure that the system response speed and resource consumption are taken into account while ensuring the calculation accuracy.

[0047] The sensitivity matrix is ​​updated synchronously with the power flow calculation results. In online mode, it is updated with power flow data at the second / minute level, and in offline mode, it is updated with power flow data at the hour level, ensuring that the sensitivity coefficient can reflect the changes in the operating status of the distribution network in real time.

[0048] This invention employs a distributed solution strategy, using the low-voltage side of the distribution transformer as a partition unit to calculate local power flow in parallel. It calibrates data through power-voltage boundary conditions at tie nodes, avoiding the efficiency bottleneck of centralized calculation across the entire network. Simultaneously, it adaptively selects either the Newton-Raphson method or the forward-backward substitution method based on the real-time photovoltaic penetration rate, ensuring convergence accuracy in complex network scenarios and improving calculation speed in large-scale sparse network scenarios. This keeps the calculation latency within the second level, adapting to the online real-time adjustment needs of high-penetration photovoltaic clusters and improving calculation efficiency compared to traditional centralized methods.

[0049] The voltage responsibility weight quantization module is used to input the voltage exceedance magnitude of any node and the voltage-power sensitivity matrix into the responsibility tracing calculation model that includes a sign direction factor, a distance attenuation factor and a power backfeed criterion when any node voltage exceeds the limit, and output the voltage responsibility weight quantization result of each photovoltaic node for the over-limit event.

[0050] In a specific embodiment of the present invention, the responsibility tracing calculation model identifies the current voltage over-limit node and its over-limit magnitude. .

[0051] It should be noted that the specific method for detecting any node voltage exceeding the limit is as follows: Based on the unified electrical state data frame, the real-time voltage of all nodes within the low-voltage side boundary of the distribution transformer is extracted. Using the rated voltage of the distribution network as a benchmark, nodes exceeding the limit are determined according to a preset standard. When the real-time voltage exceeds the preset standard, it is marked as a voltage exceeding the limit node, and the exceedance range is calculated. For example, if the preset standard is (0.93 * rated voltage of the distribution network, 1.07 * rated voltage of the distribution network), the exceedance range is specifically the relative deviation value.

[0052] Traverse the corresponding row vectors in the voltage-power sensitivity matrix to extract the sensitivity coefficient of each photovoltaic node to the over-limit node.

[0053] Multiply the current injected active power of each photovoltaic node by its sensitivity coefficient, and then multiply by... The sign direction factor is used to obtain the preliminary responsibility contribution value.

[0054] It should be noted that the purpose of the preliminary responsibility contribution value is to construct a scientific responsibility attribution logic from three dimensions: intensity of action, correlation of impact, and direction matching. This avoids the attribution bias caused by only looking at sensitivity or ignoring direction in traditional responsibility tracing, and ensures that the preliminary responsibility contribution value can truly reflect the actual impact of each photovoltaic node on voltage over-limit. This lays the foundation for the accurate correction of the subsequent introduction of distance attenuation factor and power backfeed criterion.

[0055] The injected active power represents the intensity of the photovoltaic (PV) node's impact on the power grid. The greater the injected active power, the more significant the disturbance to the grid caused by changes in its operating state. For example, the output fluctuation of a 100kW PV node has a much greater impact on voltage than that of a 10kW node. The sensitivity coefficient represents the correlation between the PV node and over-limit nodes, reflecting the magnitude of the impact of a unit change in active power on the voltage of the over-limit node. The larger the absolute value of the sensitivity coefficient, the more sensitive the PV node is to voltage regulation from the over-limit node. The sign direction factor is used to distinguish between the aggravating and suppressing effects of photovoltaic nodes on voltage over-limit. When the upper voltage limit is exceeded, The sign direction factor is +1. At this point, the injected active power at the photovoltaic node will further increase the voltage, exacerbating the voltage limit exceedance. The multiplied contribution is positive. Quantifying its negative impact, when the lower voltage limit is exceeded... The sign direction factor is -1. At this time, the active power injected into the photovoltaic node will also raise the voltage, but it belongs to suppressing over-limit. After multiplication, the contribution value is negative.

[0056] The increase in voltage due to active power injection from photovoltaic nodes is an inherent characteristic of distribution networks. This solution uses sign correction to match this unified physical phenomenon of voltage increase with the current demand for over-limit conditions. When the voltage has exceeded the upper limit, voltage increase is an adverse factor, and its responsibility attribute is marked with +1; when the voltage is below the lower limit, voltage increase is an advantageous factor, and its non-responsibility attribute is marked with -1. This ensures that responsibility tracing only targets nodes that exacerbate over-limit conditions, avoids misjudging nodes that alleviate over-limit conditions as the responsible party, and improves the accuracy of attribution.

[0057] By introducing a distance attenuation factor and a power backfeed criterion, the initial responsibility contribution value is weighted and corrected, and finally normalized to generate the voltage responsibility weight of each photovoltaic node. This weight reflects the causal correlation strength of each photovoltaic node to the over-limit event, providing a quantitative basis for subsequent regulation decisions.

[0058] It should be noted that the initial responsibility contribution value of a certain photovoltaic node is assumed to be... The distance decay factor is The distance-weighted contribution value is obtained after correction. .

[0059] , This refers to the electrical distance between the photovoltaic node and the over-limit node. This refers to the maximum electrical distance within the low-voltage side boundary of the distribution transformer. The value range is [0.5, 1].

[0060] The power backfeed correction factor is β=0.3 if the photovoltaic node experiences power backfeed (i.e., injected active power P < 0), to weaken the responsibility weight of backfeed power on voltage exceedances. Otherwise, β=1. The specific value of 0.3 is determined through typical distribution network topology simulation tests. By constructing distribution network models with different photovoltaic penetration rates (e.g., 10%-50%) and different load distributions, the power backfeed scenario of the photovoltaic node under P = -1%Pn to -10%Pn is simulated, and its contribution to voltage exceedances is statistically analyzed to be 25%-35% of that under normal injection conditions. To balance the accuracy of responsibility quantification with engineering operability, the intermediate value of 0.3 is taken as the correction factor for the power backfeed condition. Under normal injection conditions (P ≥ 0), the node's contribution to voltage exceedances is direct, hence β=1. This value can reduce the responsibility weight error in the power backfeed scenario, ensuring a scientific three-dimensional responsibility weight quantification system in conjunction with the sign direction factor and distance attenuation factor.

[0061] This invention constructs a three-dimensional responsibility weighting quantification model based on a sign direction factor, a distance attenuation factor, and a power backfeeding criterion. This model quantifies the responsibility of photovoltaic nodes for voltage exceedances from multiple dimensions, including effect intensity, influence correlation, direction matching, distance attenuation, and operating conditions. The model effectively distinguishes between the aggravating and suppressing effects of photovoltaic nodes on voltage exceedances, weakens the unreasonable responsibility weights of power backfeeding nodes, reduces attribution error control, and solves the problems of ambiguous responsibility boundaries and high misjudgment rates caused by traditional single-dimensional tracing, providing a reliable basis for precise regulation.

[0062] The adjustment instruction generation module is used to sort and compare the voltage responsibility weights according to the judgment rule of prioritizing responsibility weights and secondarily prioritizing adjustment costs. If it is determined that the responsibility weight of a certain remote photovoltaic node is higher than that of the local node, the preventive adjustment instruction generation process for the remote photovoltaic node is triggered.

[0063] In a specific embodiment of the present invention, the determination rule of prioritizing responsibility weight and secondarily prioritizing adjustment cost is specifically implemented as follows: all photovoltaic nodes are arranged in descending order of responsibility weight. If the node with the highest weight is located upstream of the over-limit node and the geographical distance exceeds the set geographical distance threshold, and the responsibility weight is greater than that of the local node, it is determined to be a remote-dominated over-limit event; otherwise, it is determined to be a non-remote-dominated over-limit event.

[0064] It should be noted that the upstream of the over-limit node is based on the electrical topology of the distribution network and the direction of power transmission. The power receiving side of the over-limit node, that is, the side from which the power flows to the over-limit node, is the upstream. When the low-voltage side node of the distribution transformer is the over-limit node, the feeder node and photovoltaic access point corresponding to the high-voltage side of the transformer are the upstream, and the load side node is the downstream.

[0065] It should also be noted that the straight-line distance calculated using the coordinates of the photovoltaic node and the over-limit node is the geographical distance. The setting of the geographical distance threshold is specifically based on simulation tests of a typical distribution network topology. When the geographical distance between the upstream highest weight node and the over-limit node exceeds the geographical distance threshold, the output fluctuation of that node needs to be transmitted through multiple levels of lines to affect the over-limit node. This is a case of remote disturbance-driven over-limit. Line loss and delay need to be considered in subsequent adjustments. The geographical distance threshold can be dynamically adjusted according to the distribution network line length and topology density. For example, in dense urban distribution networks, the geographical distance threshold can be set to 3km; in rural distribution networks with longer lines, the geographical distance threshold can be set to 8km.

[0066] When a remote-dominated over-limit event is identified, the responsibility weight ratio between the remote node and the local node is compared. If the ratio is greater than the preset significance threshold, the adjustable capacity, historical curtailment rate, and equipment health status of the remote node are first evaluated. If the adjustment feasibility conditions are met, the local adjustment is skipped, and the preventive adjustment process for the remote node is directly initiated to generate a preventive adjustment instruction. If the conditions are not met, the process regresses to the second-highest weight node for recursive judgment.

[0067] It should be noted that the local node specifically refers to a photovoltaic node located downstream of the over-limit node, whose geographical distance from the over-limit node is less than the set geographical distance threshold, and which belongs to the same low-voltage side boundary of the distribution transformer. It is a regular regulation candidate relative to the remote node.

[0068] The preset significance threshold is determined through simulation of typical distribution network topologies with different photovoltaic penetration rates, such as 10%-50%. For example, if it is 1.5, when the ratio of the responsibility weight of the remote node to the local node is >1.5, the remote node has a significant dominant role in the limiting event, and preventive regulation should be initiated first to offset the cross-regional transmission loss and delay.

[0069] It should be noted that the conditions for determining whether the adjustment is feasible are as follows: Condition 1, the adjustable capacity, i.e. the remaining adjustable reactive capacity, is greater than or equal to the target adjustment amount * redundancy coefficient or the remaining adjustable active capacity is greater than or equal to the target adjustment amount * redundancy coefficient, where the redundancy coefficient is as shown in 1.2.

[0070] Condition 2: The average light curtailment rate of the remote node during the first target period is less than a preset average light curtailment rate threshold, such as 5%. Condition 3: The inverter has no current fault alarms, the number of adjustment actions in the second cycle (e.g., 30 days) is less than or equal to the adjustment action threshold (e.g., 100 times), and the operating temperature of key components is less than or equal to the operating temperature threshold (e.g., 85℃).

[0071] It should be noted again that the recursive judgment is performed by falling back to the second highest weight node, and the termination rules are as follows: ① Find a photovoltaic node that meets the adjustment feasibility conditions and start the adjustment process; ② If no feasible node is found after traversing all photovoltaic nodes with a responsibility weight ≥ 10%, an operation and maintenance alarm is triggered and the recursion is terminated.

[0072] When the event is determined to be a non-remote-driven over-limit event, local adjustment is selected.

[0073] It should be noted that the local regulation specifically involves: referring to the preventive regulation instruction generation process, determining the target regulation amount based on the responsibility weight of the local photovoltaic node and the voltage exceedance of the over-limit node, prioritizing reactive power regulation, reducing active power output proportionally when the reactive power regulation margin is insufficient, and simultaneously verifying whether the regulation action causes other nodes to generate new over-limit risks. If so, the regulation range is dynamically adjusted, and after generating the preliminary regulation instruction, it enters the coordination verification and timing optimization process.

[0074] Local regulation for non-remotely-driven limit-crossing events follows the same logic as remote preventative regulation. It only identifies local photovoltaic nodes as the regulation targets based on geographical distance and topological location, without considering cross-regional line losses and delays. The specific methods for calculating target regulation, prioritizing reactive / active power regulation, and verifying cascading risks are identical to the corresponding steps in remote preventative regulation, ensuring the accuracy and safety of the regulation actions.

[0075] In a specific embodiment of the present invention, the preventive adjustment instruction generation process includes: determining the target adjustment amount based on the responsibility weight of the remote photovoltaic node and the voltage overshoot of the over-limit node; combining the reactive power regulation capability curve of the photovoltaic inverter, using reactive power absorption to suppress voltage rise; if the reactive power regulation margin is insufficient, reducing the active power output proportionally.

[0076] Specifically, the method for determining the target adjustment amount is as follows: the responsibility weight of the remote photovoltaic node and the voltage exceedance range of the over-limit node are considered. The target regulation amount is obtained by multiplying the inverter's maximum reactive power regulation capacity by the safety factor. The safety factor is 0.8. The value of the safety factor is determined based on the simulation of the voltage regulation stability of the distribution network and the test of the inverter response redundancy. This ensures that the regulation amount is sufficient to offset the over-limit, while avoiding the new risks caused by over-regulation.

[0077] Based on the reactive power regulation capability curve of the photovoltaic inverter, the regulation range and response time are obtained. Reactive power absorption is prioritized to suppress voltage rise, and the remaining reactive power regulation capacity is calculated. , This represents the current reactive power output of the inverter. The maximum reactive power regulation capacity is defined as follows: when the remaining reactive power regulation capacity is less than the target regulation amount multiplied by the reactive power regulation coefficient, it is considered that the reactive power regulation margin is insufficient, and the formula is applied accordingly. Calculate the reduction in active power output. The reactive power-voltage sensitivity coefficient is denoted by k, which is the active power regulation coefficient. The value range is 0.8-1.0, preferably 0.9. This coefficient is determined based on the linearity simulation of the active power regulation of the photovoltaic inverter to ensure that the active power reduction is matched with the voltage over-limit compensation requirement and to avoid excessive curtailment of solar power. The redundancy coefficient of the reactive power regulation coefficient is consistent and is set to 1.2 to ensure the consistency and safety of the reactive power regulation margin judgment and to avoid the failure to alleviate the over-limit due to insufficient regulation.

[0078] It should also be noted that the insufficient reactive power regulation margin is specifically caused by dynamic changes during the regulation process. For example, due to the need to avoid new limits during cascading risk verification, the target reactive power regulation amount is dynamically increased, resulting in the initial remaining reactive power capacity being unable to cover the load. Alternatively, the inverter's current reactive power output may be affected by power distribution network disturbances. Increase, remaining reactive power capacity Passive reductions can occur when multiple nodes to be regulated work together, causing the total reactive power demand to exceed the total reactive power regulation capacity of the low-voltage side of the distribution transformer. Other examples include excessive active power output, excessive temperature of key components, or equipment de-capacity operation, all of which can lead to a decrease in the maximum reactive power regulation capacity.

[0079] When generating instructions, it is simultaneously verified whether the adjustment action will cause other nodes to have new risks of exceeding limits. If there are chain risks, the target adjustment amount is dynamically adjusted according to the adjustment range or auxiliary nodes are introduced to participate in the coordination.

[0080] Specifically, the target adjustment amount is substituted into the voltage-power sensitivity matrix to simulate and verify the voltage and branch power of all nodes within the low-voltage side boundary of the distribution transformer. The voltage is verified to be within the preset standard, and the branch power is verified to be less than or equal to the rated capacity. If either exceeds the limit, it is determined that there is a cascading risk. The adjustment range is reduced in increments of 10%.

[0081] Specifically, the introduction of auxiliary nodes requires the following conditions to be met: The responsibility weight is ≥ 30% of the responsibility weight of the target remote node, ensuring a certain contribution to the over-limit event and effective adjustment.

[0082] The remaining adjustable capacity is greater than or equal to the required additional adjustment amount × 1.1, with a redundancy coefficient of 1.1, ensuring sufficient adjustment capability. The geographical distance from the node exceeding the limit should be ≤50% of the set threshold, such as ≤4km in rural power distribution networks, to ensure rapid response; If there is no power backflow and no equipment fault alarm, the conditions for adjustment feasibility are met.

[0083] It should also be noted that when the voltage recovers but a new limit exceedance point appears, the real-time electrical status dataset and voltage-power sensitivity matrix of the entire network are updated first, and then the responsibility weight of each photovoltaic node for the new limit exceedance point is requantified according to the recorded responsibility tracing calculation model, and then the subsequent process is followed.

[0084] Output a draft of adjustment instructions with timing markers and execution priorities for the instruction issuing module to perform global coordination.

[0085] It should be noted that the draft adjustment instruction includes the execution object identifier, adjustment action type, target adjustment amount, timing mark, execution priority, safety boundary conditions, and feedback confirmation requirements. The execution priority is determined based on responsibility weight and influence path length. Specifically, it can be normalized and processed to obtain a priority score, and then the priority is determined according to the priority score range corresponding to each priority, ensuring that nodes with higher responsibility and faster response are executed first. The safety boundary conditions are the constraints that must be met during the adjustment process, and the feedback confirmation requirements are the status information that the inverter needs to return after execution, such as the actual adjustment amount and whether there is no fault.

[0086] This invention establishes a remote / local-dominated limit violation determination mechanism. It filters regulation targets based on geographical distance, topological location, and responsibility weight ratio. For remote-dominated limit violations, preventative regulation is initiated to offset cross-regional transmission losses and delays, improving the success rate of regulation in such scenarios. For non-remote-dominated limit violations, local regulation is used, eliminating the need for cross-regional coordination and ensuring rapid regulation response. Simultaneously, reactive power regulation is prioritized, and active power output is proportionally reduced when reactive power margin is insufficient, balancing regulation effectiveness with curtailment losses and enhancing the adaptability of the local regulation strategy.

[0087] The instruction issuance module is used to perform coordination verification and timing optimization on the adjustment instruction sequences corresponding to multiple photovoltaic nodes to be adjusted, generate a step-by-step and hierarchical collaborative control instruction set, and issue it to the corresponding photovoltaic inverter for execution.

[0088] In a specific embodiment of the present invention, the coordination check performs conflict detection on the preliminary adjustment instructions of all photovoltaic nodes to be adjusted, including conflicting adjustment directions, overlapping time periods, or resource competition.

[0089] It should be noted that the aforementioned conflicting adjustment directions specifically refer to the opposite adjustment directions of the initial adjustment commands of different nodes to be adjusted for the same target, namely the voltage of the over-limit node. For example, the command of node A is to absorb reactive power to lower the voltage, while the command of node B is to generate reactive power to raise the voltage. Both are aimed at the same over-limit node, and are therefore determined to be conflicting in direction.

[0090] The overlapping time periods specifically refer to the intersection of the planned execution time windows of multiple instructions, and the duration of the intersection is greater than or equal to a duration threshold. The duration threshold is, for example, 50ms for the minimum response period of the inverter, which is set based on the industry standard value of 30-60ms for the minimum response period of photovoltaic inverters, to ensure that the adjustment actions do not interfere with each other during the overlapping time periods.

[0091] The resource competition specifically refers to the sum of the adjustment capacities required by the initial adjustment commands of all nodes to be adjusted, which exceeds the sum of the rated adjustment capacities of the corresponding inverters. For example, if the total maximum reactive power adjustment capacity of the inverter is 100kVar, and the initial adjustment commands cumulatively require 120kVar, it is determined to be a resource competition.

[0092] If a conflict exists, the adjustment tasks are redistributed based on the weight of responsibility and the comprehensive score of adjustment cost obtained from the assessment. Then, the timing is optimized, and the adjustment actions are sorted in layers according to the impact propagation path, with priority given to executing the instructions of the nodes with the shortest impact path and the fastest response to the nodes that exceed the limit.

[0093] For example, the comprehensive scoring based on the weight of responsibility and the adjustment cost. ,in The responsibility weight of the node to be adjusted accounts for 60%, prioritizing the principle of responsibility first. Let be the total adjustment cost of the node to be adjusted. The maximum adjustment cost for all nodes to be adjusted. This is a cost optimization coefficient, weighted at 40%, reflecting the principle of suboptimal cost. The value ranges from [0,1]. The higher the score, the higher the priority of task allocation.

[0094] It should be noted that the timing optimization involves hierarchically sorting adjustment actions according to their impact propagation paths, prioritizing the execution of instructions for nodes with the shortest impact path and fastest response time for nodes exceeding the limit. The path length is used as a dimension. Where R and X are the equivalent resistance and reactance from the node to be regulated to the over-limit node. The branch power distribution data obtained by solving the distribution network admittance matrix is ​​extracted. Specifically, the resistance and reactance of all branches on the shortest electrical path from the node to be regulated to the over-limit node are summed to obtain the equivalent resistance R and reactance X. The smaller L is, the faster the adjustment action responds to the over-limit node and the higher the ranking.

[0095] If multiple nodes to be adjusted have the same comprehensive score for adjustment cost, priority is determined according to the rule of ascending order of L → descending order of the number of adjustment actions.

[0096] In a specific embodiment of the present invention, the adjustment cost includes the loss of abandoned solar power, equipment wear coefficient, and communication delay cost.

[0097] It should be noted that the power loss due to curtailment... , For active power regulation, To adjust the duration, The benchmark feed-in tariff for photovoltaic power, and the equipment wear coefficient. , The wear coefficient per unit capacity per single operation. To adjust the number of movements, The rated capacity of the inverter; communication delay cost , Adjusting costs for communication delay time .

[0098] The communication delay cost coefficient of 0.01 is determined by simulation based on the correlation between the average delay of the distribution network communication system (e.g., 10-50ms) and the loss of abandoned solar power, thus quantifying the impact of communication delay on the economic efficiency of regulation.

[0099] Simultaneously, a stepped adjustment step size is set to avoid system oscillation caused by a single large adjustment; finally, a step-by-step, hierarchical collaborative control instruction set is generated, which includes the execution order, target adjustment amount, tolerance range and feedback confirmation mechanism.

[0100] For example, the specific method for setting the stepped adjustment step size is as follows: the single adjustment amplitude does not exceed 5% of the inverter's rated capacity, the total adjustment amount is divided into steps, the interval between two adjacent steps is greater than or equal to 100ms, and it is divided into 3 steps, each step 5%, with an interval of 100ms, to avoid voltage oscillation caused by a single large adjustment. The tolerance range is set according to ±5%-±10% of the target adjustment amount, and the specific value is determined based on the adjustment accuracy level of the photovoltaic inverter. For example, ±5% is used for high-precision inverters, and ±10% is used for conventional inverters. The core function of the tolerance range is to adapt to the execution error of the equipment and avoid system oscillation caused by pursuing absolute precision. It works in conjunction with the stepped adjustment step size to ensure the stability of step-by-step adjustment and the efficiency of coordinated control. When the actual adjustment amount exceeds the tolerance range, the inverter reports through a feedback confirmation mechanism, and the system starts the next fine adjustment in the execution sequence until the adjustment amount falls within the tolerance range.

[0101] The stepped adjustment step size is set based on the inverter's adjustment accuracy, such as ±5%, and the allowable range of voltage fluctuations in the distribution network. The single adjustment amplitude does not exceed 5% of the rated capacity to avoid voltage oscillation. The 100ms interval between two adjacent steps meets the hardware constraints of the inverter's continuous adjustment.

[0102] In a specific embodiment of the present invention, a closed-loop verification mechanism is also included: after the cooperative control command is issued and executed, the system re-collects the electrical status data of the entire network in the next sampling cycle, and executes the voltage-power sensitivity matrix generation module and the voltage responsibility weight quantization module again to verify whether the voltage of the over-limit node has recovered to the safe range.

[0103] If the voltage still fails to recover, determine whether the regulation failure is caused by model error or external disturbance. If so, start the secondary responsibility weight recalculation and expand the range of regulation nodes. If the voltage has recovered but a new limit point appears, start a new round of multi-node collaborative optimization process to realize dynamic iteration and adaptive adjustment of the control strategy.

[0104] For example, by comparing the electrical status data of the entire network before and after adjustment, if the deviation between the sensitivity matrix, responsibility weight and actual status is greater than 5%, it is determined to be a model error; if there are sudden situations such as a sudden increase in load or a sudden change in photovoltaic output, it is determined to be an external disturbance. The deviation threshold for model error judgment is specifically determined by comparing the difference between the calculated values ​​and the measured values ​​of the sensitivity matrix and responsibility weight before and after adjustment.

[0105] It should be noted that the specific execution rules for the recalculation of the secondary responsibility weights and the expansion of the adjustment node range are directly determined by the cause of the adjustment failure. The differentiated policy logic for model errors and external disturbances is as follows: Model error leading to failure: Update and refresh the voltage-power sensitivity matrix, fine-tune the parameter thresholds of the distance attenuation factor α and the power backfeed correction factor β, and requantify the responsibility weight using a dynamic weight calculation rule; adjust the node range to include only photovoltaic nodes with a responsibility weight ≥ 20% after the second recalculation, and supplement high responsibility nodes that were initially missed due to model deviation. The dynamic weight calculation rule is as follows: introduce the real-time grid fluctuation coefficient Y, Y = current voltage fluctuation standard deviation / rated voltage fluctuation standard deviation, with a value range of 0.8-1.2; the corrected responsibility weight = initial responsibility contribution value × α × β × Y, where Y is calculated using voltage data from the first 3 sampling periods, reflecting the impact of grid fluctuations on responsibility attribution in real time.

[0106] External disturbances causing failure: Keep the sensitivity matrix and various factor parameters unchanged, calculate the new over-limit amplitude based on the new voltage value after the disturbance, and proportionally increase the target adjustment amount on the basis of the initial responsibility weight; on the basis of the initial adjustment node, add auxiliary adjustment nodes with reactive power adjustable capacity ≥10kVar, no current fault alarm, and average curtailment rate ≤5% within the set period, such as 72 hours, to quickly supplement the adjustment amount to offset the impact of external disturbances.

[0107] The aforementioned differentiated measures ensure the accuracy and efficiency of secondary adjustments, enabling dynamic iteration and adaptive adjustment of the control strategy.

[0108] It should be added that the formulas mentioned above, through the principle of dimensional consistency and mathematical standardization methods such as normalization, dimensionless parameter conversion, or unit system unification, can translate physical quantities with different properties into unitless standard values ​​or parameters that can be superimposed in the same dimension. This eliminates the interference of different dimensions on the computational logic, allowing the formulas to retain the original data distribution characteristics while possessing mathematical rationality and adaptability to objective laws. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention.

[0109] It should also be noted that the various threshold settings described in this invention are specifically determined based on simulation tests of typical power distribution network topologies, industry standards and specifications, equipment technical parameters, actual engineering operation data, and the experience of experts in the field. All thresholds have a solid scientific basis and are practically operable, and those skilled in the art can reproduce their rationality through conventional simulation tests.

[0110] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0111] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0112] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0113] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A multi-parameter joint optimization control system for photovoltaic power distribution clusters, characterized in that, include: The electrical status dataset construction module is used to collect voltage, active power and reactive power data of each node in the power distribution cluster in real time from the SCADA system and photovoltaic monitoring system to construct a real-time electrical status dataset for the entire network. The voltage-power sensitivity matrix generation module is used to generate a voltage-power sensitivity matrix that characterizes the voltage response relationship between each node to injected power based on the real-time electrical status dataset of the entire network and in combination with the pre-stored line topology and line parameters, and to adaptively select the Newton-Raphson method or the forward-backward substitution method according to the real-time photovoltaic penetration rate. The voltage responsibility weight quantization module is used to input the voltage exceedance magnitude of any node and the voltage-power sensitivity matrix into the responsibility tracing calculation model that includes a sign direction factor, a distance attenuation factor and a power backfeed criterion when any node voltage exceeds the limit, and output the voltage responsibility weight quantization result of each photovoltaic node for the over-limit event. The adjustment instruction generation module is used to sort and compare the voltage responsibility weights according to the judgment rule of priority of responsibility weight and second-best adjustment cost. If it is determined that the responsibility weight of a certain remote photovoltaic node is higher than that of the local node, the preventive adjustment instruction generation process for the remote photovoltaic node is triggered. The instruction issuance module is used to perform coordination verification and timing optimization on the adjustment instruction sequences corresponding to multiple photovoltaic nodes to be adjusted, generate a step-by-step and hierarchical collaborative control instruction set, and issue it to the corresponding photovoltaic inverter for execution.

2. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, The real-time acquisition process synchronously acquires feeder node measurement data from the SCADA system and inverter operation data from the photovoltaic monitoring system through a standardized communication protocol, ensuring timestamp alignment. The validity of the collected raw data is verified, and invalid records with communication interruptions, sensor drift, or abnormal values ​​are removed. Using the low-voltage side of the distribution transformer as the boundary, a unified electrical status data frame is formed that includes all photovoltaic access points and key load nodes; The data frames are continuously updated according to a set sampling period and serve as the sole input source for subsequent power flow calculations and sensitivity analysis.

3. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, The power flow calculation employs a distributed solution strategy based on the Newton-Raphson method or the forward-backward substitution method, the specific details of which are as follows: The calculation method is automatically selected based on the real-time photovoltaic penetration rate. When the photovoltaic penetration rate is greater than or equal to the preset photovoltaic penetration rate, the forward-backward substitution method is used, and when the photovoltaic penetration rate is less than the preset photovoltaic penetration rate, the Newton-Raphson method is used. The distributed solution strategy uses the low-voltage side of the distribution transformer as the partition unit to divide the distribution network into multiple independent calculation sub-regions. Each sub-region calculates the local power flow in parallel, and the power-voltage boundary condition data of the tie nodes are used to aggregate the power flow results of the entire network. Using pre-stored line impedance parameters, topology connections, and node type information, a distribution network admittance matrix adapted to distributed partitioning is constructed. The voltage amplitude, phase angle, and branch power distribution data of each node are obtained by solving the admittance matrix. By using the differential perturbation method, a unit active / reactive power perturbation is applied to each photovoltaic node, and the voltage change of each node before and after the perturbation is calculated. The sensitivity coefficient is formed by the ratio of voltage change to power perturbation. A voltage-power sensitivity matrix is ​​constructed according to the dimensions of photovoltaic node-affected node-sensitivity coefficient. The voltage-power sensitivity matrix is ​​stored in a sparse form, retaining only the non-zero elements and their corresponding node indices.

4. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, The responsibility tracing calculation model identifies the current voltage over-limit node and its exceeding range. ; Traverse the corresponding row vectors in the voltage-power sensitivity matrix and extract the sensitivity coefficient of each photovoltaic node to the over-limit node; Multiply the current injected active power of each photovoltaic node by its sensitivity coefficient, and then multiply by... The sign direction factor is used to obtain the preliminary responsibility contribution value; By introducing a distance attenuation factor and a power backfeed criterion, the initial responsibility contribution value is weighted and corrected, and finally normalized to generate the voltage responsibility weight of each photovoltaic node. This weight reflects the causal relationship strength of each photovoltaic node with the over-limit event, providing a quantitative basis for subsequent adjustment decisions.

5. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, The specific implementation of the determination rule of prioritizing responsibility weight and subordinating adjustment cost is as follows: All photovoltaic nodes are sorted in descending order of responsibility weight. If the node with the highest weight is located upstream of the node that exceeds the limit and the geographical distance exceeds the set geographical distance threshold, and the responsibility weight is greater than that of the local node, it is determined to be a remote-dominated limit event. Otherwise, it is determined to be a non-remote-dominated limit event. When a remote-dominated over-limit event is identified, the responsibility weight ratio between the remote node and the local node is compared. If the ratio is greater than the preset significance threshold, the adjustable capacity, historical curtailment rate and equipment health status of the remote node are first evaluated. If the adjustment feasibility conditions are met, the local adjustment is skipped and the preventive adjustment process for the remote node is directly initiated to generate a preventive adjustment instruction. If the conditions are not met, the process regresses to the second highest weight node for recursive judgment. When the event is determined to be a non-remote-driven over-limit event, local adjustment is selected.

6. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, The process for generating the preventative adjustment instructions includes: Based on the responsibility weight of remote photovoltaic nodes and the voltage exceedance of over-limit nodes, the target regulation amount is determined. Combined with the reactive power regulation capability curve of photovoltaic inverters, reactive power absorption is adopted to suppress voltage rise. If the reactive power regulation margin is insufficient, the active power output is reduced proportionally. When generating instructions, it is simultaneously verified whether the adjustment action will cause other nodes to have new risks of exceeding the limit. If there are chain risks, the target adjustment amount is dynamically adjusted according to the adjustment range or auxiliary nodes are introduced to participate in the coordination. Output a draft of adjustment instructions with timing markers and execution priorities for the instruction issuing module to perform global coordination.

7. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, The coordination check performs conflict detection on the preliminary adjustment commands of all photovoltaic nodes to be adjusted, including conflicting adjustment directions, overlapping time periods, or resource competition. If a conflict exists, the adjustment task will be redistributed based on the weight of responsibility and the comprehensive score of adjustment cost obtained from the assessment. Then, timing optimization was performed, and the adjustment actions were sorted in layers according to the impact propagation path, prioritizing the execution of the node instructions with the shortest impact path and the fastest response for the nodes that exceeded the limit. Simultaneously, a stepped adjustment step size is set to avoid system oscillation caused by a single large adjustment; finally, a step-by-step, hierarchical collaborative control instruction set is generated, which includes the execution order, target adjustment amount, tolerance range and feedback confirmation mechanism.

8. The photovoltaic power distribution cluster multi-parameter joint optimization control system according to claim 1, characterized in that, It also includes a closed-loop verification mechanism: After the coordinated control command is issued and executed, the system will re-collect the electrical status data of the entire network in the next sampling cycle, and execute the voltage-power sensitivity matrix generation module and the voltage responsibility weight quantization module again to verify whether the voltage of the over-limit node has recovered to the safe range. If the voltage still fails to recover, determine whether the regulation failure is caused by model error or external disturbance. If so, start the secondary responsibility weight recalculation and expand the range of regulation nodes. If the voltage has recovered but a new limit point appears, start a new round of multi-node collaborative optimization process to realize dynamic iteration and adaptive adjustment of the control strategy.

9. A multi-parameter joint optimization control system for photovoltaic power distribution clusters according to claim 1, characterized in that, The voltage-power sensitivity matrix generation module supports online and offline dual-mode operation: In online mode, the power flow and voltage-power sensitivity matrix is ​​updated in real time at a cycle of seconds or minutes, which is suitable for scenarios with rapid voltage fluctuations. In offline mode, a multi-condition sensitivity library is generated in batches based on historical typical daily data for training the responsibility tracing model or rehearsing the adjustment strategy. The two modes are managed uniformly by the dispatch center and automatically switched according to the power grid operation status to ensure that the system response speed and resource consumption are taken into account while ensuring the calculation accuracy.

10. A multi-parameter joint optimization control system for photovoltaic power distribution clusters according to claim 7, characterized in that, The adjustment costs include the loss of curtailed solar power, equipment wear and tear, and communication delay costs.

Images