Distributed photovoltaic-oriented district-level voltage coordinated regulation and control method and system

Abstract

The invention provides a distributed photovoltaic-oriented district-level voltage coordinated regulation and control method and system, relates to the technical field of power distribution network operation control, and aims to solve the problems of response lag, frequent regulation and control, large active loss and the like in the existing regulation and control method. Comprising the steps of constructing a distributed photovoltaic-load node topological graph, performing regulation and control region division and key point screening, constructing a distributed photovoltaic prediction model, generating a prospective rolling optimization regulation and control strategy through model prediction control, constructing an experience playback mode, and performing regulation and control based on the experience playback mode and a model prediction control mode. According to the invention, the problems in the prior art are solved, frequent regulation and control are effectively avoided, and the regulation and control precision and the response speed are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of power distribution network operation and control technology, and particularly relates to a method and system for coordinated voltage regulation at the distribution transformer level for distributed photovoltaic power generation. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Driven by the "dual carbon" goals and energy transition strategy, distributed photovoltaic (DPV) has experienced explosive growth and has become an important component of the new power system. However, the randomness, intermittency, and anti-peak-shaving characteristics of DPV output, as its penetration rate continues to increase, lead to problems such as voltage fluctuations and increased backfeeding in the distribution area, seriously affecting the normal operation of user-side electrical equipment and the safety and stability of the distribution network.

[0004] Currently, the main approach to addressing voltage issues caused by distributed photovoltaic (PV) power is to utilize grid-connected inverters (DPVIs) for reactive power regulation, or, in extreme cases, reduce active power, i.e., "curtailment" or even shutdown. Specifically, when the voltage is low, the inverter outputs reactive power to supplement the voltage; when the voltage is high, it absorbs reactive power to lower the voltage. However, this method has certain limitations: frequent inverter adjustments not only accelerate equipment aging but also trigger overcurrent and overheat protection mechanisms, leading to derating or even shutdown, resulting in unnecessary active power loss; the lack of dynamic adaptability and predictability to PV output fluctuations leads to an increase in the curtailment rate.

[0005] Meanwhile, while traditional PI control is simple in structure, its slow dynamic response makes it difficult to cope with sudden changes in photovoltaic output, failing to guarantee high-quality power output. Model Predictive Control (MPC), as a multi-variable, multi-constraint optimization control strategy, can predict future system states based on predictive models; by directly embedding multiple constraints such as voltage, power, and control frequency into the optimization objective, it eliminates the need for additional regulators. However, standard MPC strategies still have certain problems when applied to the control of distributed photovoltaic clusters at the substation level: they rely on repeated online solutions to the optimization problem, resulting in a heavy computational burden and high computational resource requirements for the terminal controller, making it difficult to meet the high-timeliness control needs during rapid voltage fluctuations. Therefore, existing control methods suffer from problems such as response lag, frequent control, and significant active power loss. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a method and system for coordinated voltage regulation at the distribution area level for distributed photovoltaic (PV) systems. Based on user load and PV distribution, a distributed PV-load node topology is constructed. Key nodes are selected and regulation targets are designed. An optimized regulation strategy is generated by predicting using a constructed distributed PV prediction model. A historical regulation strategy reuse mechanism is introduced to improve response speed and reduce computational burden, thereby effectively avoiding frequent regulation and significantly improving regulation accuracy and response speed.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for coordinated voltage regulation at the distribution transformer level for distributed photovoltaic systems, comprising: Based on the distribution of distributed photovoltaic and load within the transformer area, a distributed photovoltaic-load node topology diagram is constructed; The control area is divided based on the distributed photovoltaic-load node topology, and the key nodes within the control area are selected. The photovoltaic characteristic state variables at the current moment are obtained, and the photovoltaic output prediction value at the next moment is obtained through the constructed distributed photovoltaic output prediction model. Based on key nodes within the control area, a rolling optimization objective function for model predictive control is constructed; Based on the photovoltaic output prediction at the next moment, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode to regulate the voltage of each region. Based on voltage regulation feedback data from various regions, an experience playback model is constructed. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

[0008] As one implementation method, a distributed photovoltaic-load node topology diagram is constructed based on the distribution of distributed photovoltaic and loads within the transformer area. The specific process is as follows: Acquire distributed photovoltaic power source data, load user data, and electricity consumption-related data; Using distributed photovoltaic power sources and load users as points and electricity consumption associations as edges, an adjacency matrix is ​​constructed, thus obtaining the distributed photovoltaic-load node topology graph.

[0009] As one implementation method, the control area is divided based on the distributed photovoltaic-load node topology diagram, and key nodes within the control area are selected. The specific process is as follows: Based on the distributed photovoltaic-load node topology, the distribution area is divided into multiple control zones; Calculate the voltage sensitivity of each node within each control region; The node with the minimum voltage sensitivity in each control region is taken as the key node of the corresponding control region.

[0010] As one implementation method, the photovoltaic characteristic state quantities obtained at the current moment include historical power output, irradiance, temperature, humidity, and rated power output of the equipment.

[0011] As one implementation method, the predicted photovoltaic output for the next time moment is obtained through the constructed distributed photovoltaic output prediction model. The specific process is as follows: A distributed photovoltaic (PV) power output prediction model is constructed using a long short-term memory (LSTM) network. The PV characteristic state variables at the current moment are input into the distributed PV power output prediction model to obtain the PV power output prediction for the next moment.

[0012] As one implementation method, a rolling optimization objective function for model predictive control is constructed based on key nodes within the control area. The specific process is as follows: Obtain the voltage of key nodes, the total active power of the region, and the inverter control frequency; With the optimization objectives of maximizing voltage stability at key nodes, maximizing total active power output in the region, and minimizing inverter control frequency, a rolling optimization objective function for model predictive control is constructed. The constraints of the rolling optimization objective function for the design model predictive control include: voltage constraints, distributed photovoltaic power output constraints, remaining capacity utilization constraints, and regulation frequency constraints.

[0013] As one implementation method, the formula for constructing the rolling optimization objective function of model predictive control is: ; ; ; ; in, This represents the difference between the voltage at the critical node after regulation and the target voltage. This represents the derating ratio of active power at time t; This represents the difference between the region's output active power and its rated active power. This represents the rated active power of the distributed photovoltaic system served by the i-th DPVI. This represents the predicted active power output value of the distributed photovoltaic system served by the i-th DPVI at time step t. This represents the cumulative value of the minimum control operation, and its value indicates the voltage control frequency of the i-th DPVI. The binary variable characterization value representing whether distributed photovoltaic i participates in voltage regulation at time step t is 0 or 1; This represents the cumulative number of time steps. , , , These represent the weighting coefficients for regional voltage, active power output, and voltage regulation frequency, respectively.

[0014] As one implementation method, based on the photovoltaic power output prediction at the next time step, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode, which regulates the voltage of each region. The specific process is as follows: The predicted photovoltaic output at the next moment is input into the rolling optimization objective function of the model predictive control for solution, and a predictive control command sequence is obtained. The first control command in the predictive control command sequence is sent to each distributed photovoltaic inverter to regulate the voltage of each area and update the current node voltage status of each area. Moving into the next time step, the input to the objective function is updated based on the updated current node voltage state of each region and the new photovoltaic output forecast; Repeat the above steps to form a model predictive control mode and complete the voltage regulation of each region.

[0015] As one implementation method, coordinated regulation of various regions is carried out based on experience playback mode and model predictive control mode. The specific process is as follows: The current data to be predicted is matched with the data in the experience replay area, and the pattern is judged based on the matching result. If a match is found, the experience replay mode is triggered, and historical control strategies are called from the experience replay area for control. Otherwise, the model prediction control mode is used for control. Based on the results of regulation, an incentive mechanism is used to optimize the judgment.

[0016] A second aspect of the present invention provides a distribution-level voltage coordinated regulation system for distributed photovoltaic systems, comprising: The topology graph construction module is used to construct a distributed photovoltaic-load node topology graph based on the distribution of distributed photovoltaic and loads within the transformer area; The key point screening module is used to divide the control area based on the distributed photovoltaic-load node topology map and screen out the key nodes within the control area; The prediction module is used to obtain the photovoltaic characteristic state quantities at the current moment, and to obtain the photovoltaic output prediction quantity at the next moment through the constructed distributed photovoltaic output prediction model. The objective construction module is used to construct the rolling optimization objective function of model predictive control based on key nodes within the control area; The control module is used to solve the rolling optimization objective function of the model predictive control based on the photovoltaic output prediction at the next moment, form the model predictive control mode, and regulate the voltage of each region. Based on voltage regulation feedback data from various regions, an experience playback model is constructed. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

[0017] The above one or more technical solutions have the following beneficial effects: This embodiment addresses the issues of delayed response, frequent operation, and significant output derating inherent in traditional inverter control. It combines a distributed photovoltaic output prediction model with a rolling optimization objective function of model predictive control. By directly embedding the optimization objective into constraints such as voltage regulation range, active power, and regulation frequency, and employing MPC to implement a predictive and overall planning rolling optimization control strategy, this approach maximizes the active power output of distributed photovoltaic systems, avoids the output reduction problem caused by frequent inverter regulation, and simultaneously provides sufficient margin for voltage regulation in the subsequent period.

[0018] In this embodiment, a dual-mode approach of experience playback mode and model predictive control mode is adopted for coordinated regulation of each region. By combining the look-ahead optimization of model predictive control (MPC) with the rapid response of the simplified experience playback mode, the accuracy and predictability of the control are guaranteed, and millisecond-level regulation speed is achieved in most scenarios, thus resolving the contradiction between the online calculation time of MPC and the rapid voltage fluctuation.

[0019] In this embodiment, an experience playback zone with self-optimizing updates is constructed. By setting self-perception rules for the control boundary, after the distributed photovoltaic power output prediction model outputs and before model predictive control is carried out, the photovoltaic predicted output classification triggers a simplified mode, that is, the control strategy with the most similar historical control curve is directly selected from the experience playback zone and applied to the current control. The control strategy is continuously updated through a reward mechanism, thereby further improving the response speed of the voltage control at the distribution area level and ensuring the stability and predictability of the control results.

[0020] In this embodiment, by using topology-based key node screening and regional coordination, the complex global control problem is decomposed into several regional coordination sub-problems, which greatly reduces the control complexity. Furthermore, by using key nodes to coordinate the actions of each inverter within the region, regulation conflicts are avoided, voltage consistency is ensured, and the stability of system operation is significantly improved. This method is suitable for voltage regulation in various scenarios at the distribution station level.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1 This is a flowchart of a distribution-level voltage collaborative regulation method for distributed photovoltaic systems according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram illustrating the command current generation principle of a distributed photovoltaic inverter according to Embodiment 1 of the present invention; Figure 3 This is an example diagram illustrating the topological association distribution of four groups of neighboring unit nodes in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram illustrating the comparative effect of voltage regulation exceeding the lower limit by 2% in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram illustrating the comparative effect of voltage regulation exceeding the upper limit by 2% in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram comparing the voltage drop effect before and after adjusting the voltage to exceed the upper limit by 8% in Embodiment 1 of the present invention. Detailed Implementation

[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0026] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0027] Example 1 This embodiment discloses a method for coordinated voltage regulation at the distribution transformer level for distributed photovoltaic systems.

[0028] To more clearly illustrate this embodiment, a process for implementing distributed photovoltaic (PV) substation-level voltage coordinated regulation can be specifically described as follows: A method for coordinated voltage regulation at the distribution transformer substation level for distributed photovoltaic systems includes: S1. Based on the distribution of distributed photovoltaic and load within the transformer area, construct a distributed photovoltaic-load node topology diagram; S2. Divide the control area based on the distributed photovoltaic-load node topology diagram and select the key nodes within the control area; S3. Obtain the photovoltaic characteristic state variables at the current moment, and obtain the photovoltaic output prediction for the next moment through the constructed distributed photovoltaic output prediction model; S4. Based on the key nodes within the control area, construct the rolling optimization objective function for model predictive control; S5. Based on the photovoltaic output prediction at the next moment, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode to regulate the voltage of each region. Based on voltage regulation feedback data from various regions, an experience playback model is constructed. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

[0029] like Figure 1 As shown, in step S1, a distributed photovoltaic-load node topology diagram is constructed based on the distribution of distributed photovoltaic and load within the transformer area.

[0030] The distribution area microgrid includes multiple distributed photovoltaic (PV) power sources. To adapt to the randomness and intermittency of renewable energy output, each distributed PV power source is equipped with a Distributed Photovoltaic Inverter (DPVI). On the one hand, it outputs active power to the distribution area microgrid based on control commands; on the other hand, in response to the fluctuation of PV output, multiple DPVIs work together to carry out regional voltage regulation and maximize active power output, ensuring the power quality of the distribution area microgrid while supporting the maximization of the benefits of surplus capacity participating in the electricity market.

[0031] The specific process for constructing a distributed photovoltaic-load node topology is as follows: (1) Obtain distributed photovoltaic power source data, load user data and electricity consumption related data.

[0032] In this embodiment, a photovoltaic project located in a village in a certain city is used as an example. A low-voltage collection line is used to aggregate photovoltaic output, and a step-up transformer connects surplus electricity to the grid. High-precision sensors are installed on the distributed photovoltaic equipment side to collect minute-level photovoltaic output data between 10:00 and 16:00, and 15-minute-level data collection for other time periods. The village has 112 photovoltaic projects, with each household having an average of 48 monocrystalline 540W photovoltaic modules installed on their roof, totaling 2.9MW, divided into 28 units. The generated electricity, after self-consumption, is collected from every four adjacent units and then fed into a dedicated step-up transformer, which connects to the grid via a 10kV line. This ultimately improves response efficiency, ensures voltage stability, and enhances power quality.

[0033] It mainly acquires distributed photovoltaic power source data, load user data, and electricity consumption-related data.

[0034] Among them, distributed photovoltaic power source data includes photovoltaic module rated capacity, photovoltaic module data, maximum active power output, and reference voltage; load user data includes reference voltage, historical electricity consumption, and electricity consumption frequency; electricity consumption related data includes communication association and transmission line association, which are obtained from the archive information of the distribution network system.

[0035] (2) Using distributed photovoltaic power sources and load users as points and electricity consumption associations as edges, construct an adjacency matrix to obtain the distributed photovoltaic-load node topology.

[0036] Based on the distribution of regional distributed photovoltaic power sources and loads, a regional distributed photovoltaic-load node topology is defined.

[0037] In this embodiment, taking a transformer area containing multiple distributed photovoltaic power sources as an example, both distributed photovoltaic power sources and load users are treated as nodes. Then the distributed photovoltaic power generation node set is Treat any existing power consumption connections (such as power transmission lines) between nodes as edges. And use this to construct an adjacency matrix ,in, This yields the distributed photovoltaic-load node topology and builds a topology association distribution map based on Matlab / Simulink.

[0038] like Figure 1 As shown, in step S2, the control area is divided based on the distributed photovoltaic-load node topology diagram, and key nodes within the control area are selected.

[0039] Based on the size of the distribution area, it is divided into multiple regions, each containing multiple distributed photovoltaic power sources. Given the interconnectedness of electricity consumption leading to voltage correlation, a node that can sensitively respond to voltage changes in other nodes within its respective region is selected as the critical node. The specific process is as follows: (1) Based on the distributed photovoltaic-load node topology, the transformer area is divided into multiple control areas.

[0040] In this embodiment, as Figure 2 As shown, a topology distribution diagram of 32 nodes was constructed based on Matlab / Simulink. Each node is connected to an electrical load, and 16 nodes are connected to distributed photovoltaic (PV) power sources equipped with DPVIs. Taking the node distribution topology of the transformer substation divided into 4 regions as an example, the topology distribution of the 4 adjacent unit nodes is shown. Each region is equipped with 4 distributed PV power sources with DPVIs. Group 1 includes nodes 4, 6, 7, and 8; Group 2 includes nodes 9, 11, 12, and 13; Group 3 includes nodes 15, 18, 20, and 21; and Group 4 includes nodes 24, 26, 29, and 32. Table 1 shows the DPVI parameter values ​​equipped with the distributed PV power sources in this transformer substation. All DPVIs use the same parameters. The reference voltage for the low-voltage transformer substation is 220V, and the voltage control target is normalized to 1.0 pu.

[0041] Table 1 DPVI Parameter Values ​​for Distributed Photovoltaic Systems

[0042] (2) Calculate the voltage sensitivity of each node in each control area.

[0043] By calculating the correlation of voltage fluctuations among nodes within a region, voltage sensitivity is used as an evaluation index for the voltage sensitivity among nodes. The formula is: (1) Where N represents the number of nodes in the region; This represents the voltage fluctuation difference between node n and node k, obtained based on historical data statistics; and These represent the voltage changes at nodes n and k, respectively; T represents the time window for accumulating sample data, calculated with a one-year time window.

[0044] (3) The node with the minimum voltage sensitivity in each control region is taken as the key node of the corresponding control region.

[0045] Will have minimum voltage sensitivity Each node is designated as a key node, and one key node is set for each region.

[0046] After the above steps, the voltage fluctuation correlation between nodes in the region is calculated as an evaluation index of the voltage sensitivity between nodes, ensuring that the voltage deviation between nodes in the region is minimized. This ensures that the voltage of all nodes in the region operates within the voltage fluctuation threshold range monitored by the key nodes through the coordinated regulation of the voltage of the key nodes by multiple DPVIs.

[0047] like Figure 1 As shown, in step S3, the photovoltaic characteristic state quantity at the current moment is obtained, and the photovoltaic output prediction quantity at the next moment is obtained through the constructed distributed photovoltaic output prediction model.

[0048] A distributed photovoltaic (PV) power output prediction model is constructed using a Long Short-Term Memory (LSTM) network algorithm. The model is input with the current PV characteristic state variables and outputs the predicted PV power output for the next time step. The specific process is as follows: (1) Obtain the photovoltaic characteristic state variables at the current moment.

[0049] The photovoltaic characteristic state quantities at the current moment include historical power output, irradiance, temperature, humidity, and rated power output of the equipment.

[0050] (2) Construct and train a distributed photovoltaic power output prediction model.

[0051] 1) Construct a distributed photovoltaic power output prediction model.

[0052] A distributed photovoltaic power output prediction model is constructed using a Long Short-Term Memory (LSTM) network. The LSTM network consists of an input layer, two hidden feature extraction layers, and an output layer.

[0053] 2) Train and optimize the distributed photovoltaic power output prediction model.

[0054] First, the distributed photovoltaic power output prediction model is trained based on the photovoltaic characteristic state variables in historical data, that is, historical data is input into the model to obtain the predicted value.

[0055] Secondly, a loss function is constructed to train the model parameters. The mean square error is set to measure the difference between the predicted value and the true value and is used as the loss function to backpropagate and update the model parameters. When the difference between the predicted value and the true value reaches the threshold of minimizing the loss function, the formula is the same as formula (2), and the trained distributed photovoltaic power output prediction model is obtained.

[0056] Finally, a monitoring feedback correction mechanism is constructed to monitor and correct the distributed photovoltaic power output prediction model, and the accuracy of the model output is monitored, as follows: (2) Where E represents The average error of the photovoltaic output prediction model at each time step; This represents the prediction error of the photovoltaic output prediction model at the current time t; This represents the actual photovoltaic output at the current time t.

[0057] If the set threshold is reached, the model needs to be retrained. The set threshold is the maximum average error allowed between the model prediction and the actual value, i.e., the maximum value of E, which is set according to the business requirements for the accuracy of distributed photovoltaic prediction.

[0058] The above steps not only improve the prediction accuracy of the distributed photovoltaic power output prediction model, but also avoid the problem of historical sample data migration in the experience playback area over time, thus replacing manually preset fixed thresholds with dynamic thresholds. This provides effective data support for updating the cluster center sample data of various samples in the experience playback area.

[0059] (3) The photovoltaic output prediction model is constructed to obtain the photovoltaic output prediction amount at the next moment.

[0060] The photovoltaic characteristic state variables at the current moment are input into the trained distributed photovoltaic power output prediction model to obtain the photovoltaic power output prediction for the next moment.

[0061] like Figure 1As shown, in step S4, a rolling optimization objective function for model predictive control is constructed based on the key nodes within the control area.

[0062] S401. Based on the key nodes within the control area, adjust the magnitude and phase of the inverter output current and perform voltage regulation on the key nodes.

[0063] The specific process is as follows: (1) Decompose the output current of the distributed photovoltaic inverter and calculate the maximum remaining power capacity of the inverter.

[0064] like Figure 3 As shown, based on the DPVI command current generation principle, distribution area-level voltage regulation is implemented. By changing the magnitude and phase of the inverter output current, the active or reactive power output of distributed photovoltaic systems is regulated, achieving distribution area voltage regulation based on predictive methods. Specifically: exist Figure 3 In this context, the i-th DPVI current output is configured to include active power current. Harmonic control current and residual current that can be used for reactive power output regulation Specifically, it is expressed as: (3) Its corresponding maximum remaining power capacity It can be represented as: (4) in, Let i be the rated current of the i-th DPVI; Provide active power current for the i-th DPVI; The harmonic extraction module for the i-th DPVI is provided to suppress harmonic currents in the local nonlinear load current. and They are respectively and The effective value; To enable real-time monitoring of effective voltage values ​​via power quality monitoring devices, it supports recording of rapid voltage change events at the 10-millisecond level.

[0065] (2) Based on the key nodes in the control area, monitor the voltage of each node in the area, calculate the deviation between the voltage of each node in the area and the voltage control set value, and transmit the deviation value to each node in the area to adjust the magnitude and phase of the inverter output current.

[0066] Based on a key node set in each region, it is responsible for monitoring the deviation between the voltage of other nodes and the voltage control setpoint, and transmitting the deviation value to each node in the region. Each node can adjust the magnitude and phase of the above three parts of the output current by changing the output current through its own DPVI, thereby realizing the adjustment of active and reactive power output, and thus realizing the control of the voltage of the key node in the region.

[0067] At this point, the inverter's remaining power capacity and reactive power regulation are set to satisfy the power theory triangle relationship. For example, when the DPVI outputs 90% of the rated active power, the remaining 10% of the rated capacity can theoretically absorb up to about 43.6% of the reactive power corresponding to the rated active power.

[0068] (3) Based on the maximum remaining power capacity and the adjustment of inverter output current, generate the control strategy of inverter in the region.

[0069] The remaining capacity of the DPVI is proportionally related to the variable reactive power output. Changing the reactive power can alter the voltage of critical nodes within a certain range. Therefore, equation (4) shows that the remaining capacity utilization rate of the i-th DPVI can be controlled. To achieve regulation of the voltage at key nodes, the control strategy for the i-th DPVI within the region is as follows: (5) in, This represents the remaining capacity utilization rate for the next time period; Let be the remaining capacity utilization rate of the i-th DPVI within the region; m is the number of DPVIs in the region; For the distributed photovoltaic-load node topology diagram within the region The corresponding element in the i-th row and j-th column of the Laplace matrix L; It is the DPVI identifier of a critical node if and only if i is a critical node. =1, otherwise 0; The DPVI information weight of the monitoring node within the voltage control group, also known as the feedback information weight; This represents the upper limit of the remaining capacity of the monitoring node DPVI within the voltage control group; Let j be the rated capacity of the j-th DPVI; For the first Rated capacity of each DPVI; To set a ratio for the active power output at time t, This represents the predicted active power of the j-th DPVI at time t; This indicates that at time t, the first... Predicted active power of each DPVI; As a key node DPVI voltage sensitivity coefficient relative to other nodes; The setpoint for critical node voltage control is typically set to 1. pu; and Let be the harmonic mitigation currents of the j-th DPVI and the critical node DPVI, respectively. Take the maximum value to ensure that the harmonic mitigation requirements are met at time t+1.

[0070] It is determined by the parameters of the DPVI device; , All of these were predicted by the distributed photovoltaic power output prediction model.

[0071] The harmonics are given by the j-th DPVI. Provided by the harmonic module of the critical node DPVI.

[0072] Key Nodes DPVI voltage sensitivity coefficient relative to other nodes The values ​​are obtained through distribution network parameters, i.e., the relationship between equal values ​​or differences is obtained when there is a parallel or series connection of power transmission.

[0073] The remaining capacity utilization rate of the i-th distributed photovoltaic inverter (DPVI) in the region at the next moment in formula (5) is determined by three parts: the remaining capacity utilization rate at the current moment, the voltage deviation value of the key node, and the difference between the remaining capacity utilization rate of the i-th distributed photovoltaic inverter and all neighboring distributed photovoltaic inverters in the region in the distributed photovoltaic-load node topology diagram.

[0074] If the critical node voltage u is Within a certain difference threshold range (e.g., ±2%), other DPVIs in the region do not need to carry out voltage regulation. If the difference threshold range is exceeded, their power output will be changed according to the topological correlation to achieve regulation of key nodes until the fluctuation stabilizes within the difference threshold range.

[0075] This embodiment regulates the voltage of key nodes by adjusting the active and reactive power output of each node's DPVI within the region. To improve the real-time performance of regulation, a model predictive control-based look-ahead generation of the active power ratio and reactive power regulation strategy is used to maximize the active power output while ensuring the voltage stability of key nodes.

[0076] S402. Construct the rolling optimization objective function for model predictive control.

[0077] (1) Obtain the voltage of key nodes, the total active power of the region and the inverter control frequency.

[0078] (2) With the optimization objectives of voltage stability at key nodes, maximization of total active power output in the region and minimization of inverter control frequency, a rolling optimization objective function for model predictive control is constructed.

[0079] The sliding time window rolling mechanism is used to regulate the voltage and power output of regional distributed photovoltaic (PV) systems. The specific objective is to keep the voltage at key nodes close to the rated voltage within each time window, maximizing the active power output of distributed PV systems and avoiding power output degradation caused by frequent inverter adjustments. Simultaneously, it provides sufficient margin for inverter voltage regulation at key nodes in subsequent periods. The rolling optimization objective function for model predictive control is constructed as follows: (6) (7) (8) (9) in, This represents the difference between the voltage at the critical node after regulation and the target voltage. This represents the derating ratio of active power at time t; This represents the difference between the region's output active power and its rated active power. This represents the rated active power of the distributed photovoltaic system served by the i-th DPVI. This represents the predicted active power output value of the distributed photovoltaic system served by the i-th DPVI at time step t. This represents the cumulative value of the minimum control operation, and its value indicates the voltage control frequency of the i-th DPVI. The binary variable characterization value representing whether distributed photovoltaic i participates in voltage regulation at time step t is 0 or 1; This represents the cumulative number of time steps. , , , These represent the weighting coefficients for regional voltage, active power output, and voltage regulation frequency, respectively. These parameters directly affect the power quality and photovoltaic output of the distribution area.

[0080] The difference between the critical node voltage and the target voltage after regulation The smaller the value obtained from formula (7), the closer the voltage in the region is to the target voltage.

[0081] The difference between the regional output active power and the rated active power The value is calculated by formula (8). The smaller the value, the higher the photovoltaic output efficiency of the transformer area and the less curtailment of solar power.

[0082] Frequent adjustments can lead to a 3%-5% loss of active power. Therefore, the minimum cumulative value of the adjustment operation can be calculated using formula (9). This avoids frequent adjustments and thus prevents the loss of active power.

[0083] (3) Design the constraints of the rolling optimization objective function of the model predictive control, including: voltage constraints, distributed photovoltaic power output constraints, remaining capacity utilization constraints and regulation frequency constraints.

[0084] The constraints are set as follows: 1) Voltage constraint, the formula is: (10) in, , These are the maximum and minimum voltage amplitudes of the i-th DPVI, respectively, indicating that the voltage should be maintained within the allowable range for safe operation of the power grid in the distribution area. If the maximum value is exceeded, the distributed photovoltaic equipment will be shut down.

[0085] 2) Output constraint of distributed photovoltaic power generation, the formula is: (11) (12) (13) in, , These are the active and reactive power outputs of the distributed photovoltaic power source i served by the i-th DPVI at time step t; The maximum active power output of distributed photovoltaic power source i; , These are the minimum and maximum reactive power outputs of distributed photovoltaic power source i, respectively. The maximum ramp rate of distributed photovoltaic power source i; The binary variable characterization value represents whether the distributed photovoltaic power source i is connected to the grid and outputs power at time step t.

[0086] Formulas (11) and (12) represent that the active and reactive power output of distributed photovoltaic power sources does not exceed their upper and lower limits, respectively; Formula (13) is the ramping constraint for distributed photovoltaic power sources.

[0087] 3) Remaining capacity utilization constraint, the formula is: (14) Equation (5) shows that when the remaining capacity utilization rate exceeds the maximum reactive power regulation allowed by DPVI, it indicates that the inverter cannot achieve voltage reduction operation through reactive power regulation. At this time, the active power output ratio of DPVI should be reduced to ensure the voltage of critical nodes. and The objective function is minimized through iterative rolling optimization; when subsequent reactive power regulation can satisfy voltage regulation, the objective function is minimized. This will constrain the voltage reduction control method that lowers active power output, thus returning to the reactive power voltage regulation method.

[0088] 4) Adjusting frequency constraints.

[0089] Considering the frequency response constraint of the photovoltaic unit controller, the control command should not exceed the upper limit allowed by the response constraint. Therefore, the frequency constraint formula is: (15) (16) Where Δf is the maximum frequency value allowed by the photovoltaic unit controller; This represents a time step, measured in milliseconds (ms). This represents the cumulative number of control operations performed by the i-th DPVI within 1 second.

[0090] Through the above steps, the problems of insufficient foresight, difficulty in taking multiple objectives into account, and the possibility of control commands exceeding equipment safety limits in traditional control methods are solved. At the same time, considering voltage deviation, active power output, and regulation frequency, the inverter's reactive power capacity can be prioritized for voltage regulation while ensuring voltage safety, thereby minimizing active power reduction (curtailment) and improving the effective utilization rate of distributed photovoltaic power.

[0091] like Figure 1 As shown, in step S5, based on the photovoltaic output prediction at the next moment, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode to regulate the voltage of each region; based on the voltage regulation feedback data of each region, an experience playback mode is constructed; based on the experience playback mode and the distributed photovoltaic output prediction model, coordinated regulation of each region is carried out.

[0092] S501. Based on the photovoltaic output prediction at the next moment, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode to regulate the voltage of each region.

[0093] (1) Input the predicted photovoltaic output at the next moment into the rolling optimization objective function of the model predictive control to solve for the predictive control command sequence.

[0094] (2) Send the first control command in the predictive control command sequence to each distributed photovoltaic inverter to regulate the voltage of each region and update the current node voltage status of each region.

[0095] (3) Enter the next time step and update the input of the objective function based on the updated current node voltage state of each region and the new photovoltaic power output prediction.

[0096] (4) Repeat steps (1) to (3) to form a model predictive control mode and complete the voltage regulation of each region.

[0097] Within the model predictive control area, the voltage of the distribution grid at the substation level is... Predictive control is implemented step by step. Each time the objective function is solved, a set of control sequences is obtained. The control sequence with the minimum objective function value is taken as the optimal control sequence, and its first control variable is applied to each DPVI for regulation. The grid at the distribution area level updates the current voltage state of key nodes. At the next time step, the current photovoltaic characteristic state is used as the input to the predictive model to obtain the predicted photovoltaic output for the next time step. This predicted output is then combined with the actual voltage of key nodes from the previous time step to update the input to the objective function. This effectively utilizes external environment and model error information, repeating the objective function solution process from the previous time step. The first control variable of the optimal control sequence obtained at the next time step is applied to each DPVI, and this iterative rolling optimization is repeated.

[0098] S502. Based on voltage regulation feedback data of each region, an experience playback mode is constructed.

[0099] Based on voltage regulation feedback data from various regions, regulation boundary sensing rules are set, and an experience playback zone is constructed. After obtaining the predicted data output by the distributed photovoltaic power output prediction model, the predicted data is classified and similarity-evaluated with the historical regulation samples stored in the experience playback zone. According to the regulation boundary sensing rules, either the simplified mode or the model prediction control mode is triggered to execute the acquired regulation strategy. The reward value is measured based on the execution result feedback data. If the reward value is positive, the regulation sample data of the current prediction data, which is composed of the current regulation operation and the corresponding historical meteorological data, is stored in the experience playback zone, thereby continuously updating the sample data in the experience playback zone.

[0100] Given the rapid changes in distributed photovoltaic (PV) output, the current method of remotely generating and issuing control strategies via predictive control (MPC) is insufficient to meet the high timeliness requirements of control response. Therefore, a simplified model is constructed based on control feedback data to further improve the efficiency of distributed PV control at the substation level. Control boundary perception rules are set, and after the distributed PV output prediction model is output but before model predictive control is implemented, PV output time-series prediction data and meteorological data (hereinafter referred to as prediction data) are categorized, and the capacity zone of the current control operation is determined by combining the categorization rules.

[0101] S503. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

[0102] (1) Match the current data to be predicted with the data in the experience replay area, and make a pattern judgment based on the matching result. If they match, trigger the experience replay mode and call the historical control strategy from the experience replay area for control. Otherwise, use the predictive control mode for control.

[0103] Specifically, when the distance between the predicted data and the cluster center sample of the category in the experience replay unit is less than the similarity evaluation threshold... Furthermore, the duration of voltage fluctuation within ±2% after adjustment is longer than the set time. If this happens, a simplified mode is triggered, which means that the control strategy with the most similar historical control curve is directly selected from the experience replay area and applied to the current control. The experience replay area is used to store sample data of strategies that have achieved high accuracy in historical control.

[0104] The evaluation similarity value between the predicted data and the cluster center samples of the category in the experience playback unit is calculated using the Euclidean clustering algorithm. The distance is the average of the differences between the distributed photovoltaic power output prediction data and the power output data of the cluster center samples at various time points. The evaluation similarity threshold is set according to the core reward mechanism. The larger the value, the more reasonable the set evaluation similarity threshold is. Historical feedback data is statistically analyzed to calculate... The maximum value can be used to obtain the corresponding similarity assessment threshold.

[0105] When the cluster center samples of the predicted data and the curves of each category in the experience playback unit are not higher than the similarity threshold. Or, the duration of voltage fluctuation within ±2% after adjustment is shorter than the set time. If this occurs, the model predictive control mode is triggered, which generates a control strategy through rolling optimization and applies it to the current control. The model predictive control mode is, in this embodiment, a combination of the distributed photovoltaic power output prediction model and the rolling optimization objective function of model predictive control.

[0106] (2) Based on the results of regulation, the reward mechanism is used to optimize the judgment.

[0107] By repeatedly verifying the model's stability in handling similar problems, the reliability of its defined regulatory boundary-aware rules is ensured. The core reward mechanism is as follows: (17) in, Indicates the reward / punishment coefficient, guiding the calibration of cognitive boundaries; This represents the reward / penalty score for perceptual partitioning; the higher the reward value, the higher the accuracy of boundary delineation.

[0108] Through the above steps, by setting dual-mode triggering rules based on effect stability, it is ensured that the simplified mode is only activated when reliable, thus jointly guaranteeing the consistency of the system's control speed and accuracy in long-term operation. At the same time, a reward mechanism is introduced to drive the continuous evolution of the experience playback area, further ensuring the consistency of the system's control speed and accuracy in long-term operation. The control strategy is continuously updated through the reward mechanism, thereby further improving the response speed of the substation-level voltage control and ensuring the stability and predictability of the control results.

[0109] S6. Verify voltage regulation of distributed power sources at the substation level.

[0110] To verify and compare the effectiveness and applicability of the distributed voltage control strategy in this embodiment, three scenarios were set up for comparative analysis of regulation.

[0111] Scenario 1: A sudden increase in load causes the voltage at critical nodes to fall below the lower limit by 2%; Scenario 2: Distributed power output fluctuations cause the voltage of critical nodes to exceed the upper limit by 2%; Scenario 3: Distributed power output fluctuations cause the voltage of critical nodes to exceed the upper limit by 8%.

[0112] The following table selects the current mainstream distributed power source voltage control methods for comparison: Method 1: Convert voltage regulation into a minimum voltage control deviation objective function, solve for the function value based on the Hamilton-Jacobi-Bellman equation, and then execute... The voltage regulation value is obtained through iterative evaluation using a dual neural network. Method 2: Use multiple distributed photovoltaic inverters to coordinate and regulate regional voltage fluctuations, determine the regional regulation leader by minimizing the single-step convergence factor, and construct a leader-follower mode for coordinated and consistent control of node voltage. Method 3: Voltage regulation is achieved through hierarchical model predictive control. The upper layer plans the overall regulation index of each group of DPVI, while the lower layer combines nonparametric kernel density estimation to predict the output of distributed power sources and uses a sliding window to update the voltage rolling optimization decision.

[0113] S601. Comparative analysis of regulation in scenario 1 where a sudden increase in load causes the voltage at critical nodes to fall below the lower limit by 2%.

[0114] Simulation sample data was obtained based on historical monitoring data. The simulation time was set to 15 seconds. At t=7.5 seconds, the active power load of Group 1 increased by 30 kW, and the reactive power increased by 19.2 kVar. Figure 1 Table 2 shows how the DPVI in Group 1 adjusts the voltage changes of key nodes by outputting reactive power.

[0115] Table 2 Output power before and after control for each group

[0116] As shown in Table 2, at 5s, due to the sudden increase in load of Group 1, the voltage of the critical node dropped. The DPVI in the group quickly raised the voltage of the critical node to around 1.0pu by outputting reactive power.

[0117] like Figure 4 As shown, before regulation, the critical node 6 in group 1 suddenly exceeded the lower limit at 7.5s. Therefore, the four nodes in the group equipped with photovoltaic inverters boosted the voltage by outputting reactive power. After regulation using the four methods, the voltage of each node stabilized at the control target 1 pu. Among them, method 1 completed the voltage regulation task within 7s because it adopted an execution... The adversarial learning iterative solution for evaluation is more prone to instability in regulation compared to the other three methods. Its voltage regulation fluctuates significantly at 11.2s, posing a risk of exceeding the upper limit. Method 2 also completes the voltage regulation task within 7s. However, because it uses a leader-follower mode for coordinated control of node voltages, it fails to predict the output of distributed power sources in the future and adopts a conservative regulation operation based on the current output. Therefore, the regulation completion time is 2.8 times that of the method in this embodiment. Method 3 completes the voltage regulation task within 6s. Due to the use of a hierarchical control strategy, the regulation response speed is about 2s slower than other methods in the initial stage. However, due to the use of a forward-looking model predictive control in the later stage, the regulation speed is faster than that of Method 1 and Method 2.

[0118] Based on the above research, the method in this embodiment completes the voltage regulation task within 2.5 seconds. Due to its improvements in distributed photovoltaic prediction, it enhances prediction accuracy and stability, and introduces model predictive control to achieve forward-looking regulation. At the same time, it uses historical regulation response data to trigger a simplified mode through experience playback, further improving regulation accuracy and response efficiency. Under the premise of ensuring stable voltage regulation, when the voltage of key nodes exceeds the lower limit of -7% to -2% due to a sudden increase in load, the predictive control voltage completion speed is more than 2.4 times faster than other methods.

[0119] S602, Comparative analysis of regulation in scenario 2 where distributed power output fluctuations cause the voltage of critical nodes to exceed the upper limit by 2%.

[0120] Simulation sample data was obtained based on daily monitoring data. The simulation time was set to 15s. At t=5s, due to the increase in light intensity, the voltage of each group of key nodes exceeded the upper limit by 2%. Each group of DPVI regulated the voltage of each node by absorbing reactive power, as shown in Table 3.

[0121] Table 3 Output power before and after control for each group

[0122] As shown in Table 3, at 5s, due to the increased output of the distributed power source, the voltage of each critical node exceeded the limit by 2%. Each group of DPVIs absorbed reactive power to quickly reduce the voltage of the critical node to around 1.0pu.

[0123] like Figure 5 As shown, this embodiment takes group 2 as an example to illustrate the comparison before and after voltage regulation exceeding the upper limit by 2%. The key node 11 in group 2 exceeds the upper limit by 2% in 5 seconds. Therefore, the four nodes in the group equipped with photovoltaic inverters reduce the voltage by absorbing reactive power. After regulation by the four methods, the voltage of each node is stable at the control target of 1 pu. Among them, method 1 completes the voltage reduction regulation within 7.2 seconds. Due to its lack of predictive foresight and the existence of voltage drop regulation deviation, the voltage regulation fluctuation is relatively large. Method 2 completes the voltage reduction regulation within 7.8 seconds. Its conservative regulation speed is about 1.8 seconds slower than the method in this paper. Method 3 completes the voltage reduction regulation within 7 seconds. In the initial stage, the regulation response speed has a certain lag compared with other methods, and the response speed is about 1.5 seconds slower than the method in this paper.

[0124] Based on the above research, the method in this embodiment completes the voltage reduction regulation within 6 seconds. Experiments show that the output fluctuation of distributed power sources causes the voltage of critical nodes to exceed the upper limit by 2% to 7%. Compared with other methods, the method in this embodiment has a faster response speed and more stable regulation.

[0125] S603, Comparative analysis of regulation in scenario 3 where distributed power output fluctuations cause critical node voltage to exceed the upper limit by 8%.

[0126] Simulation sample data was obtained based on monitoring data. The simulation time was set to 10s. At t=4s, due to the increase in light intensity, the voltage of each critical node exceeded the upper limit by 8%. The DPVI of each group was adjusted by reducing the active power output using the method in this paper, as shown in Table 4.

[0127] Table 4 Output power before and after control for each group

[0128] As shown in Table 4, at 150ms, due to the increase in the output of the distributed power source, the voltage of the critical node exceeds the limit. Each group of DPVI reduces the active power to quickly reduce the voltage of the critical node to around 1.0pu. Before using this method, the active power output of the traditional method is reduced by about 6.7%. After using this method, the active power output is reduced by about 3.7%, and the overall active power output is increased by about 3%.

[0129] like Figure 6As shown, taking Group 2 as an example, a comparison is made before and after the voltage exceeds the upper limit of 8%. The voltage of the key node 11 in Group 2 exceeds the limit to 8% at 4s. Therefore, the four nodes in the group equipped with photovoltaic inverters reduce the voltage by reducing the active power output. After the four methods of regulation, the voltage of each node is stable at the control target of 1 pu. Method 1 exhibits a lag in response time from 4s to 4.8s due to the time delay inherent in its iterative solution using a dual neural network. It completes voltage regulation within 3.6s, which is 1.6s slower than the method in this embodiment, representing 1.8 times the regulation time. Furthermore, its overall regulation fluctuation is significant. Method 2 completes voltage reduction regulation within 4.8s, but its conservative approach results in a 2.8s slower regulation time compared to the method in this embodiment, representing 2.4 times the regulation time. While it exhibits the smallest voltage fluctuation, its regulation completion time is the longest. Method 3 completes voltage reduction regulation within 3.8s, which is 1.8s slower than the method in this embodiment, representing 1.9 times the regulation time. In the initial stage, its regulation response speed is relatively lagging compared to other methods. Due to the delayed response, the voltage continues to rise, resulting in a 0.2s slower regulation response speed than the method in this embodiment, posing a certain risk of exceeding the limit and causing shutdown. Based on the above research, the method in this embodiment completes the voltage reduction regulation within 2 seconds, with relatively small fluctuations and the fastest regulation response speed. Experiments show that when the voltage of critical nodes exceeds the upper limit by 7%-10% due to the output fluctuation of distributed power sources, the method in this embodiment has the fastest response speed and regulation speed and is more stable than other methods. At the same time, the active power output is improved by about 3% compared with that before regulation.

[0130] This embodiment addresses the issue of voltage fluctuations in distribution areas affecting the safe and stable operation of users and distribution network equipment due to the large-scale development of distributed photovoltaic (PV) systems. First, a distributed PV-load node topology is constructed based on user load and PV distribution. Key nodes are selected, and multiple inverters are coordinated to adjust the magnitude and phase of their output current, forming a regional constraint control target. Then, a distributed PV prediction model is built, and a forward-looking rolling optimization control strategy is generated using model predictive control to achieve coordinated voltage control in various areas of the distribution area. Finally, based on feedback data from previous voltage control operations, control boundary perception rules are extracted to form a playback-simplified mode that improves control response speed. Monitoring feedback is used to correct and update the accuracy of the distributed PV prediction model output. Simulation examples verify the stability and usability of the control strategy generated by this method.

[0131] Example 2 The purpose of this embodiment is to provide a district-level voltage collaborative regulation system for distributed photovoltaic power generation, including: The topology graph construction module is used to construct a distributed photovoltaic-load node topology graph based on the distribution of distributed photovoltaic and loads within the transformer area; The key point screening module is used to divide the control area based on the distributed photovoltaic-load node topology map and screen out the key nodes within the control area; The prediction module is used to obtain the photovoltaic characteristic state quantities at the current moment, and to obtain the photovoltaic output prediction quantity at the next moment through the constructed distributed photovoltaic output prediction model. The objective construction module is used to construct the rolling optimization objective function of model predictive control based on key nodes within the control area; The control module is used to solve the rolling optimization objective function of the model predictive control based on the photovoltaic output prediction at the next moment, form the model predictive control mode, and regulate the voltage of each region. Based on voltage regulation feedback data from various regions, an experience playback model is constructed. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

[0132] Based on the provision of a distribution-level voltage collaborative regulation system for distributed photovoltaic power, the method steps in Embodiment 1 are implemented.

[0133] Example 3 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0134] Example 4 The purpose of this embodiment is to provide a computer-readable storage medium.

[0135] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0136] Example 5 The purpose of this embodiment is to provide a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods and functions involved in any of the above embodiments. The steps and methods involved in the apparatus of the above embodiments correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0137] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0138] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for coordinated voltage regulation at the distribution transformer substation level for distributed photovoltaic systems, characterized in that, include: Based on the distribution of distributed photovoltaic and load within the transformer area, a distributed photovoltaic-load node topology diagram is constructed; The control area is divided based on the distributed photovoltaic-load node topology, and the key nodes within the control area are selected. The photovoltaic characteristic state variables at the current moment are obtained, and the photovoltaic output prediction value at the next moment is obtained through the constructed distributed photovoltaic output prediction model. Based on key nodes within the control area, a rolling optimization objective function for model predictive control is constructed; Based on the photovoltaic output prediction at the next moment, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode to regulate the voltage of each region. Based on voltage regulation feedback data from various regions, an experience playback model is constructed. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

2. The method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, Based on the distribution of distributed photovoltaic (PV) systems and loads within the transformer area, a distributed PV-load node topology diagram is constructed. The specific process is as follows: Acquire distributed photovoltaic power source data, load user data, and electricity consumption-related data; Using distributed photovoltaic power sources and load users as points and electricity consumption associations as edges, an adjacency matrix is ​​constructed, thus obtaining the distributed photovoltaic-load node topology graph.

3. The method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, Based on the distributed photovoltaic-load node topology diagram, control areas are divided, and key nodes within the control areas are selected. The specific process is as follows: Based on the distributed photovoltaic-load node topology, the distribution area is divided into multiple control zones; Calculate the voltage sensitivity of each node within each control region; The node with the minimum voltage sensitivity in each control region is taken as the key node of the corresponding control region.

4. The method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, The photovoltaic characteristic state quantities acquired at the current moment include historical power output, irradiance, temperature, humidity, and rated power output of the equipment.

5. A method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, The photovoltaic output prediction model is constructed to obtain the photovoltaic output prediction amount at the next moment. Specifically, a long short-term memory network is used to construct a distributed photovoltaic output prediction model, and the photovoltaic characteristic state quantity at the current moment is input into the distributed photovoltaic output prediction model to obtain the photovoltaic output prediction amount at the next moment.

6. A method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, Based on key nodes within the control area, a rolling optimization objective function for model predictive control is constructed. The specific process is as follows: Obtain the voltage of key nodes, the total active power of the region, and the inverter control frequency; With the optimization objectives of maximizing voltage stability at key nodes, maximizing total active power output in the region, and minimizing inverter control frequency, a rolling optimization objective function for model predictive control is constructed. The constraints of the rolling optimization objective function for the design model predictive control include: voltage constraints, distributed photovoltaic power output constraints, remaining capacity utilization constraints, and regulation frequency constraints.

7. A method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, The formula for constructing the rolling optimization objective function of model predictive control is as follows: ； ； ； ； in, This represents the difference between the voltage at the critical node after regulation and the target voltage. This represents the derating ratio of active power at time t; This represents the difference between the region's output active power and its rated active power. This represents the rated active power of the distributed photovoltaic system served by the i-th DPVI. This represents the predicted active power output value of the distributed photovoltaic system served by the i-th DPVI at time step t. This represents the cumulative value of the minimum control operation, and its value indicates the voltage control frequency of the i-th DPVI. The binary variable characterization value representing whether distributed photovoltaic i participates in voltage regulation at time step t is 0 or 1; This represents the cumulative number of time steps. , , , These represent the weighting coefficients for regional voltage, active power output, and voltage regulation frequency, respectively.

8. A method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, Based on the photovoltaic power output prediction for the next time step, the rolling optimization objective function of the model predictive control is solved to form a model predictive control mode, and the voltage of each region is regulated. The specific process is as follows: The predicted photovoltaic output at the next moment is input into the rolling optimization objective function of the model predictive control for solution, and a predictive control command sequence is obtained. The first control command in the predictive control command sequence is sent to each distributed photovoltaic inverter to regulate the voltage of each area and update the current node voltage status of each area. Moving into the next time step, the input to the objective function is updated based on the updated current node voltage state of each region and the new photovoltaic output forecast; Repeat the above steps to form a model predictive control mode and complete the voltage regulation of each region.

9. A method for coordinated voltage regulation at the distribution transformer substation level as described in claim 1, characterized in that, Based on the experience playback mode and the model predictive control mode, coordinated regulation and control of various regions are carried out. The specific process is as follows: The current data to be predicted is matched with the data in the experience replay area, and the pattern is judged based on the matching result. If a match is found, the experience replay mode is triggered, and historical control strategies are called from the experience replay area for control. Otherwise, the model prediction control mode is used for control. Based on the results of regulation, an incentive mechanism is used to optimize the judgment.

10. A distribution-level voltage coordinated regulation system for distributed photovoltaic power generation, characterized in that, include: The topology graph construction module is used to construct a distributed photovoltaic-load node topology graph based on the distribution of distributed photovoltaic and loads within the transformer area; The key point screening module is used to divide the control area based on the distributed photovoltaic-load node topology map and screen out the key nodes within the control area; The prediction module is used to obtain the photovoltaic characteristic state quantities at the current moment, and to obtain the photovoltaic output prediction quantity at the next moment through the constructed distributed photovoltaic output prediction model. The objective construction module is used to construct the rolling optimization objective function of model predictive control based on key nodes within the control area; The control module is used to solve the rolling optimization objective function of the model predictive control based on the photovoltaic output prediction at the next moment, form the model predictive control mode, and regulate the voltage of each region. Based on voltage regulation feedback data from various regions, an experience playback model is constructed. Based on experience playback mode and model predictive control mode, coordinated regulation and control are carried out in various regions.

Images