Virtual power plant power regulation method based on hybrid order damping control

Abstract

The application discloses a virtual power plant power regulation method based on hybrid-order damping control and belongs to the technical field of virtual power plant power control. The method comprises the following steps: a virtual power plant hierarchical control model containing photovoltaic, energy storage and flexible load nodes is constructed; according to the response time constant difference of each node, a first-order damping control law is configured for the photovoltaic node, a second-order damping control law is configured for the energy storage node, and a fractional-order damping control law is configured for the flexible load node; a plurality of nodes with a connection degree exceeding a preset threshold are selected as damping nodes, and real-time power deviation and frequency deviation of the damping nodes are collected; the deviation is input into the hybrid-order damping control law, and the hybrid-order control amount of each node is calculated through distributed iteration; the power output of each node is adjusted according to the hybrid-order control amount, and the damping coefficient is dynamically updated according to the state feedback of the damping nodes. The method realizes fine and differentiated collaborative regulation of the power output of multiple types of distributed resources.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant power control technology, specifically a virtual power plant power regulation method based on hybrid-order restraint control. Background Technology

[0002] Virtual power plants participate in grid power regulation by aggregating distributed power sources and controllable loads. The construction method of their control model directly affects the regulation accuracy and response speed. In existing virtual power plant control architectures, a uniform control law is typically used to regulate all nodes, without considering the significant differences in response time constants among different types of equipment such as photovoltaics, energy storage, and flexible loads. The maximum power point tracking of photovoltaic inverters is completed in milliseconds, the power conversion of energy storage converters involves the coupling of electrochemical response and power electronic control, and the aggregated regulation of flexible loads is limited by user response behavior and communication delays, with response time windows reaching seconds or even minutes. This indiscriminate treatment of node dynamic characteristics makes it difficult for the parameters of a single-order control law to simultaneously adapt to the dynamic behavior of all types of nodes, leading to overshoot or response lag in some nodes, affecting the overall regulation convergence speed and steady-state accuracy.

[0003] Existing control strategies often employ a fixed-order selection method for control nodes, typically determining the set of control nodes statically based on node degree or betweenness centrality, and keeping it unchanged throughout the control process. This static selection mechanism fails to consider the drift of critical network topology nodes caused by load fluctuations and changes in renewable energy output during the operation of a virtual power plant. This results in the control nodes' ability to perceive the overall system state decaying over time, reducing the efficiency of control in guiding global dynamics. Furthermore, the setting of control coefficients often relies on offline tuning or empirical values, lacking a mechanism for dynamic adjustment based on real-time power and frequency deviations. Consequently, it cannot proactively strengthen the control to accelerate convergence when the system state deviates.

[0004] The problem this method aims to solve is how to simultaneously handle the differentiated control requirements of three devices with different response time scales—photovoltaics, energy storage, and flexible loads—in virtual power plant power regulation, and how to enable the selection of restraint nodes and the updating of restraint coefficients to have dynamic adaptive capabilities based on the real-time operating status of the system. Summary of the Invention

[0005] The purpose of this invention is to provide a virtual power plant power regulation method based on hybrid-order constraint control. This method addresses the inherent response time constants of three types of nodes: photovoltaic, energy storage, and flexible loads. It constructs first-order, second-order, and fractional-order constraint control laws to match these time constants, achieving order adaptation between the control model and the dynamic characteristics of the nodes. During the regulation process, constraint nodes are selected based on real-time changes in network connectivity. The power deviation and frequency deviation of the constraint nodes are used as feedback quantities to dynamically adjust the constraint coefficients in the hybrid-order constraint control law, so that the control intensity automatically adjusts with the degree of system deviation, thereby improving the response consistency of multi-type resource aggregation regulation.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a virtual power plant power regulation method based on hybrid-order restraint control, the method comprising: A hierarchical control model of a virtual power plant, including photovoltaic (PV), energy storage, and flexible load nodes, is constructed. Based on the differences in response time constants of each node, a first-order tethered control law is configured for the PV nodes, a second-order tethered control law for the energy storage nodes, and a fractional-order tethered control law for the flexible load nodes. Several nodes with connectivity exceeding a preset threshold are selected as tethered nodes, and their real-time power and frequency deviations are collected. These real-time power and frequency deviations are input into the hybrid-order tethered control law, and the hybrid-order control quantity for each node is calculated through distributed iterative computation. The power output of each node is adjusted according to the hybrid-order control quantity, and the restraint coefficient in the hybrid-order restraint control law is dynamically updated based on the state feedback of the restraint node. This technical solution adapts to the control response differences caused by photovoltaic fast response, energy storage dual time scale response and flexible load aggregation response by configuring differentiated restraint control laws that match the dynamic characteristics of different types of nodes. In the process of coordinated regulation, it effectively avoids overshoot or oscillation caused by control law mismatch. By selecting nodes with high connectivity as restraint nodes, it achieves local control global convergence, which greatly reduces the communication burden and computational complexity.

[0007] As a technical solution of this invention, when constructing a hierarchical control model for a virtual power plant, the geographical access locations and electrical connections of all distributed power sources and controllable loads within the jurisdiction of the virtual power plant are obtained. Based on the geographical access locations, the virtual power plant is divided into several voltage level control layers. Within each voltage level control layer, a set of photovoltaic nodes, a set of energy storage nodes, and a set of flexible load nodes are extracted based on the electrical connections. The maximum power point tracking status data of each photovoltaic node, the state of charge data of each energy storage node, and the interruptible power capacity data of each flexible load node are obtained. These status and capacity data are stored as node attributes associated with the corresponding nodes, generating a hierarchical control model for the virtual power plant containing node type labels and node attribute fields. This construction method enables the model to accurately reflect the geographical distribution and electrical hierarchy of the virtual power plant and incorporates the operating status and adjustability of each node into the model attributes, providing a complete parameter basis for the precise configuration of subsequent differentiated control laws.

[0008] Preferably, when configuring control laws for different nodes, the first-order inertial time constant is determined based on the maximum power point tracking response time of each photovoltaic node in the photovoltaic node set, and a first-order restraint control transfer function is constructed based on the first-order inertial time constant. This ensures that the power regulation response of the photovoltaic node matches the inherent delay of maximum power point tracking, avoiding maximum power point tracking instability caused by excessively rapid command changes. The second-order natural oscillation frequency and damping ratio are determined based on the power conversion system response time and battery dynamic response time of each energy storage node in the energy storage node set, and a second-order restraint control state equation is constructed based on the second-order natural oscillation frequency and damping ratio. This ensures that the charging and discharging power regulation of the energy storage node can balance response speed and overshoot suppression. The fractional order is determined based on the load aggregation time window of each flexible load node in the flexible load node set, and a fractional-order restraint control differential equation is constructed based on the fractional order. This allows for smooth load reduction within the dynamic time window of load aggregation, avoiding power consumption shocks caused by rigid shutdowns.

[0009] Furthermore, the fractional order of the fractional-order constraint control differential equation is adjusted in real time according to the dynamic changes of the load aggregation time window of the flexible load node, and the adjustment range is proportional to the rate of change of the load aggregation time window. This mechanism enables the fractional-order control law to adapt to the dynamic balance between the load adjustable margin and the response speed. When the load aggregation time window shortens rapidly, the control order is increased to accelerate convergence, and when the time window is lengthened, the order is decreased to improve steady-state accuracy, thus ensuring the stability and effectiveness of flexible load regulation in complex time-varying scenarios.

[0010] As a technical solution of the present invention, when selecting control nodes, the number of direct electrical connections between each node and other nodes in the hierarchical control model of the virtual power plant is calculated as the connectivity of that node. Nodes with a connectivity greater than a preset connectivity threshold are marked as candidate control nodes, and a preset number of nodes are randomly selected from the candidate control nodes as control nodes. The node number of each control node in the model is recorded. Based on the node number, the current active power output value and the rated active power reference value of each control node are collected in real time, and the difference between the two is calculated to obtain the real-time power deviation. The current grid frequency and the rated frequency reference value are collected simultaneously, and the difference between the two is calculated to obtain the real-time frequency deviation. This selection method ensures that the control nodes have sufficient electrical influence. Only the deviation information of a small number of high connectivity nodes needs to be collected to effectively reflect the power and frequency imbalance state of the entire virtual power plant, which significantly reduces the number of sensors deployed and the data acquisition overhead.

[0011] During distributed iterative computation, a mixed-order tethering control input signal is generated for each tethering node based on its real-time power deviation and real-time frequency deviation. This signal is then broadcast to all adjacent nodes with direct electrical connections to the tethering node. For any non-tethering node, it receives all the mixed-order tethering control input signals broadcast by its adjacent tethering or non-tethering nodes. Based on its own node type, it selects the corresponding first-order, second-order, or fractional-order tethering control law. The weighted average of all received signals is then substituted into the selected control law for iterative calculation, outputting the mixed-order control quantity of the non-tethering node. For any tethering node, its own real-time power deviation and real-time frequency deviation are directly substituted into the control law corresponding to its node type for calculation, outputting the mixed-order control quantity of the tethering node. This distributed iterative mechanism allows non-tethering nodes to gradually converge to the same control target as the tethering node through interaction with adjacent nodes without needing to obtain global information, effectively reducing the complexity of the communication topology.

[0012] Preferably, when adjacent nodes broadcast mixed-order restraint control input signals, a consensus algorithm is used to filter the signals to remove phase lag components caused by communication delays. This processing method can suppress control phase distortion introduced by communication network delays, ensure the consistency of the input signals received by each node in time phase, and thus improve the convergence accuracy and robustness of the entire distributed iterative process under non-ideal communication conditions.

[0013] As a technical solution of the present invention, when adjusting the power output and updating the restraint coefficient, the mixed-order control quantity calculated by each non-restrained node is converted into a power adjustment command for that node and sent to the corresponding photovoltaic node, energy storage node, or flexible load node to adjust its active power output value; at the end of each control cycle, the real-time power deviation and real-time frequency deviation of all restrained nodes are re-acquired, and the absolute change of the real-time power deviation and the absolute change of the real-time frequency deviation of each restrained node between the current control cycle and the previous control cycle are calculated; it is determined whether both absolute changes are less than their respective preset convergence thresholds; if both are less than their respective preset convergence thresholds, the control is considered complete. If the deviation is less than the convergence threshold, all constraint coefficients in the current hybrid-order constraint control law remain unchanged. If any deviation is not less than the convergence threshold, the constraint coefficient corresponding to that constraint node is linearly increased according to the ratio of the absolute change in real-time power deviation to the absolute change in real-time frequency deviation. The updated constraint coefficient is then substituted into the hybrid-order constraint control law for iterative calculation in the next control cycle. This dynamic update strategy can proactively enhance the constraint strength to accelerate convergence when the system power or frequency deviation converges slowly, and keep the constraint coefficient unchanged when the deviation has entered a steady state to avoid over-adjustment, thus achieving an adaptive balance between control strength and system stability margin.

[0014] Furthermore, the restraint coefficient in the hybrid-order restraint control law is adaptively adjusted based on the weighted sum of the real-time power deviation and frequency deviation of the restraint node. The weighting coefficient is determined by the voltage sensitivity of the restraint node. Nodes with high voltage sensitivity contribute more to voltage support in the system. Assigning them a higher weighting coefficient enables the restraint coefficient to more sensitively reflect the actual power and frequency regulation requirements of the system, thereby improving the control accuracy.

[0015] As a preferred implementation, after determining that the absolute change in real-time power deviation and the absolute change in real-time frequency deviation are both less than their respective preset convergence thresholds, the sum of squares of the real-time power deviations of all restraining nodes in the current control cycle is further calculated to obtain the global power deviation energy value. The sum of squares of the real-time frequency deviations of all restraining nodes in the current control cycle is also calculated to obtain the global frequency deviation energy value. The historical global power deviation energy value and the historical global frequency deviation energy value stored in the previous control cycle are also obtained. The magnitudes of the current global power deviation energy value and the historical global frequency deviation energy value are compared. If the current global power deviation energy value is less than the historical global power deviation energy value and the current global frequency deviation energy value is less than the historical global frequency deviation energy value, then all restraining coefficients in the current hybrid-order restraining control law remain unchanged. If any one is not less than, then a linear increase operation of the restraining coefficient is triggered. By introducing secondary verification at the global deviation energy level, the potential instability situation where the local deviation of the restraining node fluctuates slightly but the global deviation energy shows an upward trend can be effectively identified. Early intervention is carried out when the system shows a tendency to diverge, which significantly improves the safety of virtual power plant power regulation.

[0016] As a technical solution of the present invention, when calculating the mixed-order control quantity, the non-restrained node obtains the electrical distance value of each neighboring node broadcasting a signal to it, calculates the corresponding weighting coefficient based on the electrical distance value, wherein the neighboring node with the smaller electrical distance value corresponds to a larger weighting coefficient; multiplies the mixed-order restraint control input signal of each neighboring node by the corresponding weighting coefficient, sums them, and then divides by the sum of all weighting coefficients to obtain the weighted average input signal; selects the corresponding control law transfer function, state equation, or differential equation according to the node type of the non-restrained node, and uses the weighted average input signal as the input variable. The system performs a numerical integration or differential iteration calculation to obtain the initial control quantity of the non-restrained node. It then acquires the historical control quantity of the non-restrained node from the previous control cycle and weights and fuses the initial control quantity with the historical control quantity to obtain the mixed-order control quantity of the non-restrained node in the current control cycle. This weighted and historical control quantity fusion mechanism based on electrical distance enables nodes closer to the restraining source to undertake more regulation tasks. At the same time, by introducing historical control quantities to smooth out and suppress instantaneous disturbances during the iteration process, it ensures the dual smoothness of power regulation commands in both spatial distribution and temporal dimensions.

[0017] Specifically, when the non-stuck node is a photovoltaic (PV) node, a first-order stuck control transfer function is selected as the control law. After performing one operation on the weighted average input signal, Euler difference calculations are performed to obtain the initial control quantity for the PV node. This initial control quantity characterizes the maximum power point tracking offset of the PV node, enabling efficient operation near the maximum power point while completing power response. When the non-stuck node is an energy storage node, a second-order stuck control state equation is selected as the control law. The weighted average input signal is decomposed into proportional and differential input signals, which are then substituted into the proportional and differential channels of the second-order stuck control state equation for parallel calculation. The calculation results from the two channels are then combined. By superimposing these values, the initial control quantity of the energy storage node is obtained. This initial control quantity characterizes the power charge-discharge conversion rate of the energy storage node, enabling it to respond quickly to system power shortages while preventing frequent switching of charge-discharge modes from adversely affecting battery life. When the non-restrained node is a flexible load node, a fractional-order restraint control differential equation is selected as the control law. A fractional-order numerical approximation calculation based on the Green's function is performed on the weighted average input signal to obtain the initial control quantity of the flexible load node. This initial control quantity characterizes the load reduction ratio of the flexible load node, enabling gradual and smooth power reduction within the load aggregation time window, thus reducing the impact on users' electricity consumption.

[0018] As a technical solution of this invention, after completing power regulation and dynamically updating the restraint coefficient, the actual power output values ​​of all photovoltaic nodes, energy storage nodes, and flexible load nodes after power regulation are obtained. The ratio of the actual power output value to the rated power capacity of each node is calculated to obtain the load factor of that node. It is then determined whether there are nodes in the virtual power plant hierarchical control model whose load factor exceeds a preset overload threshold. If so, the node type label of that node is temporarily modified to a restricted node, and the weight coefficient of that node in the next control cycle is reduced. Based on the distribution location of all restricted nodes, weak areas with insufficient power regulation capability in the virtual power plant hierarchical control model are identified. At the beginning of the next control cycle, the restraint coefficient of all nodes in this weak area is increased. This mechanism can monitor the operating margin of each node in real time during the regulation process, actively reduce the load of nodes approaching overload and mark them as restricted. Simultaneously, by strengthening the restraint strength of the weak area, the control pressure is diffused to the distributed resources outside the area, avoiding cascading cutoff triggered by a single node overload, effectively improving the operational reliability and collaborative safety of the virtual power plant under extreme conditions.

[0019] The technical effects and advantages provided by the present invention in the above technical solution are as follows: A first-order tethered control law is configured for the photovoltaic (PV) node. Its transfer function structure contains only a single inertial element, maintaining consistency in order with the dynamic characteristics of PV maximum power point tracking (MPPT), thus avoiding phase lag or numerical oscillations introduced by higher-order control laws in millisecond-level response scenarios. This first-order structure allows the PV node to directly adjust its output using an inertial response upon receiving a power deviation signal, eliminating overshoot caused by the mismatch between the control law order and the dynamic order of the object. The energy storage node employs a second-order tethered control state equation, mapping the power conversion system response and the battery dynamic response to proportional and differential channels, respectively. These two channels are calculated in parallel and then superimposed for output. The proportional channel handles steady-state power deviations, while the differential channel responds to the power change rate. This allows the charge / discharge conversion rate of the energy storage node to be controlled by a combination of parameters: the natural oscillation frequency and the damping ratio. This provides rapid power support while suppressing energy loss caused by frequent switching of charge / discharge directions. The flexible load nodes are configured with fractional-order constraint control differential equations, the fractional order of which is dynamically set according to the load aggregation time window. The memory characteristics of fractional-order operators are used to smooth load reduction commands, enabling the flexible load response process to exhibit a non-integer integral effect in the time dimension. This adapts to the non-uniform delay characteristics of the load aggregation time window as user behavior changes. Three different orders of constraint control laws coexist in the same virtual power plant hierarchical control model. Through independent configuration of their respective orders, differentiated control channels corresponding to the physical response time scale of the nodes are formed. Each node undertakes a regulation share matching its response speed according to its own dynamic capabilities, avoiding slow-response nodes being forced to track fast-changing commands or fast-response nodes being limited by slow coordination mechanisms.

[0020] Nodes with connectivity exceeding a preset threshold are selected as control nodes. A direct mapping relationship is established between the selection of control nodes and the real-time connectivity of the network topology, allowing the distribution of control nodes to be dynamically updated as the electrical connection structure within the virtual power plant changes. After collecting its own real-time power and frequency deviations, the control node broadcasts the deviation signal along the electrical connection edge to adjacent nodes. The control coefficient is linearly adjusted at the end of each control cycle based on the absolute changes in power and frequency deviations. When the deviation change is not less than the convergence threshold, the control coefficient increases linearly, enhancing the control node's guidance of neighboring nodes and accelerating deviation decay. When the deviation change falls below the convergence threshold and the global power or frequency deviation energy value does not meet the convergence condition, the control coefficient is also increased to prevent the system from prematurely falling into a local stability illusion. This adaptive control strength adjustment mechanism based on state feedback concentrates control energy on critical nodes when the system deviates significantly, and automatically reduces the control gain when the system approaches steady state, minimizing unnecessary control actions that disturb the normal operating points of nodes. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0022] Figure 1 This is a flowchart of a virtual power plant power regulation method based on hybrid-stage restraint control; Figure 2 This is a flowchart of the generation process for a hierarchical control model of a virtual power plant; Figure 3 This is a flowchart of the virtual power plant overload node identification and dynamic adjustment of the restraint coefficient. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0024] See Figure 1 This invention provides a virtual power plant power regulation method based on hybrid-order restraint control, comprising: constructing a hierarchical control model of a virtual power plant including photovoltaic, energy storage, and flexible load nodes; configuring a first-order restraint control law for photovoltaic nodes, a second-order restraint control law for energy storage nodes, and a fractional-order restraint control law for flexible load nodes based on the differences in the response time constants of each node; selecting several nodes in the hierarchical control model with a connectivity exceeding a preset threshold as restraint nodes, and collecting the real-time power deviation and frequency deviation of the restraint nodes; inputting the real-time power deviation and frequency deviation into the hybrid-order restraint control law, and calculating the hybrid-order control quantity of each node through distributed iteration; adjusting the power output of each node according to the hybrid-order control quantity, and dynamically updating the restraint coefficient in the hybrid-order restraint control law based on the state feedback of the restraint nodes.

[0025] In specific implementation, please refer to Figure 2This method obtains the geographical access locations and electrical connections of all distributed power sources and controllable loads within the jurisdiction of the virtual power plant. The geographical access locations are obtained by exporting the latitude, longitude, and altitude coordinates of all grid-connected points within the virtual power plant's jurisdiction from the power grid geographic information system. Each grid-connected point corresponds to one distributed power source or one controllable load. The latitude, longitude, and altitude coordinates are combined to form a three-dimensional spatial location vector, which serves as the geographical access location of the corresponding grid-connected point. The electrical connections are obtained by exporting the distribution network topology file from the power grid energy management system. The distribution network topology file includes line segment numbers, line segment start node numbers, line segment end node numbers, line segment impedance values, and admittance values. A node adjacency matrix is ​​established based on the line segment start node numbers and line segment end node numbers. In the node adjacency matrix, an element with a value of 1 in the `i`th row and `j`th column indicates a direct electrical connection between node `i` and node `j`, while a value of 0 indicates no direct electrical connection between node `i` and node `j`. The node adjacency matrix is ​​then used as the electrical connection relationship.

[0026] In practical implementation, the virtual power plant is divided into several voltage level control layers based on geographical access location. The process of dividing the voltage level control layers involves obtaining the installation location and turns ratio parameters of all transformers within the virtual power plant. The transformer installation location is represented by a three-dimensional spatial position vector of the node connected to the high-voltage side busbar of the transformer, and the turns ratio parameter is represented by the ratio of the rated voltage value on the high-voltage side to the rated voltage value on the low-voltage side of the transformer. Using the transformer installation location as the dividing boundary, nodes with the same voltage level are grouped into the same voltage level control layer. The voltage levels include 10kV, 380V, and 220V layers. Based on electrical connection relationships, photovoltaic node sets, energy storage node sets, and flexible load node sets are extracted within each voltage level control layer. The process of extracting the photovoltaic node set involves traversing all nodes within the voltage level control layer, obtaining the power type field of each node from the node grid connection registration information, and adding nodes with a power type field value of "photovoltaic power generation" to the photovoltaic node set. The process of extracting the energy storage node set involves adding nodes with a power type field value of "electrochemical energy storage" to the energy storage node set. The process of extracting the flexible load node set is to add nodes whose power type field value is "flexible load" to the flexible load node set.

[0027] In practice, the maximum power point tracking (MPPT) status data for each photovoltaic (PV) node is acquired. This MPPT data is obtained by reading the maximum power point voltage value at the current sampling moment from the PV inverter controller. With maximum power point current value Read the irradiance value at the current sampling time from the photovoltaic array data acquisition device. With solar panel temperature value The maximum power point voltage value Maximum power point current value Irradiance value With solar panel temperature value These are combined into a four-dimensional state vector, which serves as the maximum power point tracking state data. The state of charge (POC) data for each energy storage node is acquired. This POC data is obtained by reading the POC percentage value at the current sampling time from the energy storage battery management system. Maximum voltage of individual cells Minimum voltage of individual cells Read the maximum allowable charging power value at the current sampling moment from the energy storage converter controller. With the upper limit of allowable discharge power The percentage of state of charge Maximum voltage of individual cells Minimum voltage of a single unit Maximum allowable charging power With the upper limit of allowable discharge power The data is combined into a five-dimensional state vector, which serves as the state of charge data. The interruptible power capacity data for each flexible load node is then obtained. This interruptible power capacity data is acquired by reading the pre-agreed maximum interruptible active power value of the flexible load node within the current scheduling cycle from the load aggregator control platform. Maximum interruptible duration Interrupt response time constant The maximum interruptible active power value Maximum interruptible duration Interrupt response time constant They are combined into a three-dimensional state vector, which serves as interruptible power capacity data.

[0028] In practical implementation, maximum power point tracking (MPPT) status data, state of charge (SOC) data, and interruptible power capacity (EPC) data are used as node attributes and associated with the corresponding nodes for storage. The associated storage process involves creating a node attribute data table in the real-time database of the virtual power plant control center. This table includes a node ID field, a node type label field, a node attribute field, and a last update timestamp field. The node ID field stores the unique identifier of the node; the node type label field takes the value of "PV," "energy storage," or "flexible load"; the node attribute fields store the corresponding MPPT status data, SOC data, or EPC data in a structured data format; and the last update timestamp field records the last write time of the node attribute field. This generates a hierarchical control model of the virtual power plant containing the node type label and node attribute fields. The virtual power plant hierarchical control model is stored in memory using a directed attribute graph data structure. Each vertex in the directed attribute graph corresponds to a node, and each vertex carries a vertex label and a set of vertex attributes. The vertex label takes the value of one of "photovoltaic", "energy storage" or "flexible load". The vertex attributes contain all the contents of the node attribute fields of the corresponding node. The edges in the directed attribute graph correspond to the elements with a value of 1 in the adjacency matrix of the nodes. Each edge carries an electrical distance value as an edge attribute. The electrical distance value is determined by the impedance modulus of the line segment calculated from the impedance value and admittance value of the line segment.

[0029] In practical implementation, the first-order inertial time constant of each photovoltaic node is determined based on its maximum power point tracking (MPPT) response time. The MPPT response time is obtained by initiating a closed-loop step test on the photovoltaic node. Under stable operating conditions, the DC bus voltage setpoint of the photovoltaic inverter is suddenly reduced by 5% from the current operating point. The time required for the actual DC bus voltage to return to steady state from the moment of the sudden reduction, with fluctuations not exceeding ±2% of the steady-state value, is recorded. This recorded time is taken as the MPPT response time, measured in seconds. The first-order inertial time constant is determined by dividing the MPPT response time by three. The unit is seconds, and the first-order inertial time constant is... The value is equal to one-third of the maximum power point tracking response time.

[0030] In practical implementation, a first-order entrapment control transfer function is constructed based on the first-order inertial time constant. First-order entrapment control transfer function The expression is:

[0031] in, This represents the Laplace variable, with units of radians per second; The first-order inertial time constant of the photovoltaic node is calculated from one-third of the maximum power point tracking response time, and is expressed in seconds. This represents the proportional gain for photovoltaic (PV) node control, ranging from 0.5 to 2.0. The specific value is determined based on the voltage sensitivity of the control layer at the voltage level of the PV node. Voltage sensitivity is defined as the ratio of the change in active power injection at the PV node to the change in voltage amplitude at that node. Higher voltage sensitivity results in higher voltage sensitivity. The closer the value is to 2.0, the lower the voltage sensitivity. The closer the value is to 0.5.

[0032] In practical implementation, the second-order natural oscillation frequency and damping ratio of each energy storage node are determined based on the power conversion system response time and battery dynamic response time of each energy storage node in the energy storage node set. The power conversion system response time is obtained by reading the time required for the power command step response to reach 90% of the command value from the factory technical parameter table of the energy storage converter, and using this time as the power conversion system response time. The unit is seconds. The battery dynamic response time is obtained by extracting the time required for a single battery cell to stabilize and maintain a stable voltage with fluctuations not exceeding ±1% of the rated voltage when switching from zero-power command condition to rated power condition from the charge / discharge test data of the energy storage battery management system. This extracted time is taken as the battery dynamic response time. The unit is seconds. Second-order natural oscillation frequency. The method for determining this is to calculate the response time of the power conversion system. Battery dynamic response time The second-order natural oscillation frequency is obtained by taking the reciprocal of the sum and then multiplying the resulting reciprocal by an adjustment factor of 1.25. The unit is radians per second. Damping ratio. The determination method is to calculate the battery dynamic response time. With power conversion system response time The ratio of the two values ​​is used as the damping ratio of the energy storage node. Damping ratio It is a dimensionless quantity, and its value ranges from 0.3 to 0.8.

[0033] In practical implementation, a second-order restraint control state equation is constructed based on the second-order natural oscillation frequency and damping ratio. The second-order restraint control state equation uses the power deviation of the energy storage node. and the rate of change of power deviation For the state variables, the state equation expression is:

[0034] in, This represents the deviation between the actual output value of the active power of the energy storage node and the reference value of the rated active power, expressed in watts. express The first derivative with respect to time, measured in watts per second; express The second derivative with respect to time, in watts per second squared; The second-order natural oscillation frequency of the energy storage node is obtained by multiplying the reciprocal of the sum of the power conversion system response time and the battery dynamic response time by 1.25, and the unit is radians per second. The damping ratio of an energy storage node is a dimensionless quantity, derived from the ratio of the battery's dynamic response time to the power conversion system's response time. The tethering control input signal represents the energy storage node and is determined by the hybrid-order tethering control input signal received by the energy storage node. The unit is watts.

[0035] In practical implementation, the fractional order of the flexible load nodes is determined based on the load aggregation time window of each flexible load node in the flexible load node set. The load aggregation time window is obtained by reading the pre-set load aggregation time window length of the flexible load nodes within the current scheduling cycle from the load aggregator control platform. The unit is seconds, and the length of the load aggregation time window is... This represents the maximum permissible time interval between the issuance of a load shedding command by the load aggregator and the actual execution of the load shedding action by the flexible load. Fractional order. The determination method is to calculate the length of the load aggregation time window. With reference time constant The ratio is substituted into the order mapping function. Calculations are performed to obtain the fractional order. fractional order The value range is between 0.2 and 1.0, which is the reference time constant. The value is sixty seconds. This represents an exponential function with the natural constant e as its base.

[0036] In practical implementation, fractional-order constraint control differential equations are constructed based on fractional-order orders. These equations utilize the interruptible active power quantities of flexible load nodes. For the controlled variable, the differential equation expression is:

[0037] in, Indicates fractional order Caputo fractional differential operator of order 1. Based on the length of the load aggregation time window With reference time constant The ratio is calculated using an order mapping function and is a dimensionless quantity; Indicates time Interruptible active power of flexible load nodes, in watts; This represents the proportional control coefficient for flexible load nodes, with a value ranging from 0.1 to 1.0. The specific value is determined based on the maximum interruptible active power value in the interruptible power capacity data of the flexible load node. Determined, maximum interruptible active power value The larger, The closer the value is to one point zero; Indicates time The restraint control input signal of the flexible load node is determined by the mixed-order restraint control input signal received by the flexible load node, and the unit is watts.

[0038] In practice, the fractional order of the fractional-order constraint control differential equation is adjusted in real time according to the dynamic changes of the load aggregation time window of the flexible load node, and the adjustment range is proportional to the rate of change of the load aggregation time window. The dynamic change monitoring method of the load aggregation time window is as follows: at the beginning of each control cycle, the load aggregator control platform re-evaluates and updates the length of the load aggregation time window based on the current communication link delay measurement value of the flexible load node and the response status of the load terminal. The value. The rate of change of the load aggregation time window is calculated by updating the length of the load aggregation time window after the current control cycle. Load aggregation time window length compared to the previous control cycle The difference divided by the length of the control cycle The rate of change of the load aggregation time window was obtained. The unit is seconds per second, and the length of the control cycle. The value is taken as one-hundredth of a second. The adjustment magnitude is proportional to the rate of change of the load aggregation time window. This is achieved by dividing the fractional order... Adjustment amount Set as the rate of change of the load aggregation time window Multiply by adjustment factor ,Right now Adjustment coefficient The value is 0.01, and the adjusted fractional order is... Equal to the fractional order before adjustment and The sum of, i.e. ,when When it exceeds 1.0, it is limited to 1.0. When the amplitude is below 0.2, its amplitude is limited to 0.2.

[0039] In practical implementation, the number of direct electrical connections between each node and other nodes in the hierarchical control model of the virtual power plant is calculated, and this number is used as the node's connectivity degree. The process for calculating connectivity degree involves extracting a node adjacency matrix from the directed attribute graph of the hierarchical control model of the virtual power plant. The number of rows and columns in the node adjacency matrix are equal to the total number of nodes in the hierarchical control model of the virtual power plant. The node adjacency matrix Line 1 Column elements A value of 1 indicates a node With nodes There is a direct electrical connection between the elements A value of 0 indicates a node With nodes There is no direct electrical connection between them; the connectivity of the node in each row of the adjacency matrix is ​​obtained by summing the values ​​of each row. connectivity The calculation formula is:

[0040] in, Represents a node The degree of connectivity is dimensionless and takes a non-negative integer value. This represents the total number of nodes in the hierarchical control model of a virtual power plant. It is a positive integer greater than 1; Represents the node in the adjacency matrix. Line 1 Column elements, The value can be 0 or 1; Indicates subscript From 1 to All Perform a summation operation.

[0041] In practice, the connectivity of each node is compared with a preset connectivity threshold, and nodes with connectivity greater than the threshold are marked as candidate restraining nodes. (Connectivity threshold) The threshold is determined by calculating the arithmetic mean of the connectivity of all nodes in the hierarchical control model of the virtual power plant, multiplying the arithmetic mean by a scaling factor of 1.2, and obtaining the connectivity threshold. Connectivity threshold Round up to the nearest integer. During the comparison operation, read the connectivity of each node sequentially. ,when Greater than the connectivity threshold At that time, the node The node number is added to the candidate restraint node list when Less than or equal to the connectivity threshold At that time, not the node Perform the marking operation.

[0042] In practice, a predetermined number of nodes are randomly selected from the candidate control nodes as control nodes, and the node number of each control node in the virtual power plant hierarchical control model is recorded. The predetermined number is determined by calculating the total number of nodes in the photovoltaic node set, energy storage node set, and flexible load node set in the virtual power plant hierarchical control model, multiplying the total number of nodes by a scaling factor of 0.1, and obtaining the selected number. Select quantity Round down to the nearest integer. When less than 2, The value is 2. The random selection process is as follows: a random integer index value between 1 and the length of the candidate restraining node list is generated using a linear congruent pseudo-random number generator. The node number corresponding to this index value in the candidate restraining node list is extracted as the restraining node number. After each extraction, the extracted node number is deleted from the candidate restraining node list. This extraction process is repeated for a total of... Next, get A list of node numbers for each control node; store the list of all extracted node numbers in the control node configuration register of the virtual power plant control center.

[0043] In practical implementation, based on the node number, the current active power output value and the rated active power reference value of each control node are collected in real time. The difference between the current active power output value and the rated active power reference value is calculated to obtain the real-time power deviation of each control node. The current active power output value is collected by communicating with the intelligent electronic device corresponding to the grid connection point of the control node, using the Modbus / TCP protocol to read the instantaneous power value from the active power register address in the intelligent electronic device. The sampling interval of the instantaneous power value is related to the control cycle length. Consistent with a sampling interval of one-hundredth of a second, the instantaneous power value read is used as the current active power output value. The unit is watts, and the subscript is... Indicates the first The first constraint node. The rated active power reference value is obtained by extracting the first reference node from the day-ahead generation plan database of the virtual power plant energy management system. The planned output value of each control node during the current scheduling period will be used as a reference value for rated active power. The unit is watts. Real-time power deviation. The calculation formula is Real-time power deviation The unit is watts, when the real-time power deviation A positive value indicates that the current active power output is higher than the rated active power reference value. When the real-time power deviation... A negative value indicates that the current active power output is lower than the rated active power reference value.

[0044] In practical implementation, the current grid frequency and the rated frequency reference value are synchronously collected for each control node. The difference between the current grid frequency and the rated frequency reference value is calculated to obtain the real-time frequency deviation of each control node. The current grid frequency is collected by reading the instantaneous frequency value from the frequency measurement channel of the synchronous phasor measurement unit installed at the grid connection point of the control node. The data refresh rate of the synchronous phasor measurement unit and the control cycle length are also considered. The matching and acquisition operations are triggered synchronously in the same interrupt service routine as the active power acquisition operations, and the instantaneous frequency value read is used as the current grid frequency. The unit is Hertz (Hz). Rated frequency reference value. The value is 50 Hz, the reference value for the rated frequency. This is a constant value, stored in the parameter configuration table of the virtual power plant control center. Real-time frequency deviation. The calculation formula is Real-time frequency deviation The unit is Hertz, when the real-time frequency deviation A positive value indicates that the current power grid frequency is higher than the rated frequency reference value. When the real-time frequency deviation... A negative value indicates that the current grid frequency is lower than the rated frequency reference value. Real-time power deviation. Deviation from real-time frequency The data are stored in the corresponding fields of the real-time deviation database, which is implemented by the real-time memory database of the virtual power plant control center.

[0045] In practical implementation, a hybrid-order strait control input signal is generated for each strait node based on its real-time power deviation and real-time frequency deviation. The hybrid-order strait control input signal is generated by reading the first... Real-time power deviation of each restraining node Deviation from real-time frequency Real-time power deviation Multiply by power deviation weighting factor Real-time frequency deviation Multiplied by frequency deviation weighting factor Then sum the two products to get the first product. Hybrid-order strait control input signal for each strait node Power deviation weighting factor The value is determined by calculating the first... Reference value of rated active power of each restraining node Total rated power of virtual power plants The ratio is multiplied by a scaling factor of 1.0 to obtain the power deviation weighting coefficient. Power deviation weighting factor The value ranges from zero to 1.0. Frequency deviation weighting coefficient. The value is determined by starting from the first... The voltage phase angle sensitivity at the current moment is read from the synchronous phasor measurement unit of each grid-connected node. Voltage phase angle sensitivity Defined as the ratio of the change in active power injection at the grid connection point to the change in voltage phase angle, in radians per watt, the voltage phase angle sensitivity is... The absolute value divided by the reference sensitivity The frequency deviation weighting coefficient is obtained. Reference sensitivity The value is 0.01 radians per watt.

[0046] In practical implementation, the hybrid-order restraint control input signal is broadcast to all adjacent nodes that have a direct electrical connection with the restraint node. The broadcasting operation involves querying the node number of the restraint node from the directed attribute graph of the virtual power plant hierarchical control model, traversing all columns of the row corresponding to that node number in the node adjacency matrix, and when the... When the element of a column is 1, the node will be... The node number is extracted as the adjacent node number, and the mixed-order tethering control input signal is used. The message is encapsulated as a UDP communication message. The UDP communication message contains the target neighbor node number field, the source restraint node number field, the mixed-order restraint control input signal value field, and the sending timestamp field. The UDP communication message is unicast to each neighbor node through the real-time communication bus of the virtual power plant control center.

[0047] In practical implementation, any non-restrained node in the hierarchical control model of the virtual power plant receives mixed-order restraint control input signals broadcast by all adjacent restrained nodes or adjacent non-restrained nodes. The receiving process is as follows: during the control cycle, the non-restrained node... Listening on the UDP port within the receive time window, the length of the receive time window is the control period. One-tenth of the UDP communication packets arriving within the receiving time window are stored in the input signal buffer of the non-pinning node. After the receiving time window closes, the source node number field and the mixed-order pinning control input signal value field of each UDP communication packet are parsed from the input signal buffer. A key-value pair is formed by combining the source node number and the mixed-order pinning control input signal value and stored in the adjacent input signal list of the non-pinning node. When adjacent nodes broadcast the mixed-order pinning control input signal, a consensus algorithm is used to filter the signal, removing the phase lag component caused by communication delay. The consensus algorithm's filtering method involves deploying a first-order lag compensation filter in the communication interface of each adjacent node. The transfer function of the first-order lag compensation filter is:

[0048] in, The transfer function of a first-order hysteresis-compensated filter is dimensionless. This represents the Laplace variable, with units of radians per second; The average delay of the communication link between the adjacent node and the restraining node is calculated as the arithmetic mean of the difference between the sending timestamp and the local receiving time recorded by the adjacent node in the most recent ten control cycles, and the unit is seconds. This represents the lag compensation strength coefficient, with a value ranging from 0.05 to 0.2. The specific value is determined by the maximum delay volatility of the communication link delay statistics during the offline phase. A smaller maximum delay volatility indicates a higher lag compensation strength. The closer the value of is to 0.05, the greater the maximum delay volatility. The closer the value of is to zero two, the better. A first-order hysteresis compensation filter is used for the mixed-order restraint control input signal. The filtering operation is to... The time-domain signal is input to the first-order hysteresis compensation filter, and the difference equation is used for real-time filtering calculation. The output is a mixed-order control input signal with phase hysteresis components filtered out. ,by replace The mixed-order control input signal value is used to encapsulate UDP communication messages.

[0049] In practical implementation, based on the node type of the non-constrained node, the corresponding first-order, second-order, or fractional-order constrained control law is selected. The weighted average of all received mixed-order constrained control input signals is then substituted into the selected control law for iterative calculation, outputting the mixed-order control quantity of the non-constrained node. The node type is determined by reading the node type from the vertex labels stored in the directed attribute graph of the virtual power plant hierarchical control model. When the vertex label value is "photovoltaic," a first-order constrained control transfer function is selected. As the control law; when the vertex label is "energy storage", the second-order restraint control state equation is selected as the control law; when the vertex label is "flexible load", the fractional-order restraint control differential equation is selected as the control law.

[0050] In practical implementation, the process of taking the weighted average of all received mixed-order tethered control input signals and substituting it into the selected control law for iterative calculation is as follows: Obtain the electrical distance value of each adjacent node that broadcasts the mixed-order tethered control input signal to the non-tethered node. The electrical distance value is read from the edge attributes connecting the non-tethered node and the adjacent node in the directed attribute graph of the virtual power plant hierarchical control model. The electrical distance value is the impedance modulus of the corresponding line segment, in ohms. Calculate the weight coefficient corresponding to each adjacent node based on its electrical distance value. The smaller the electrical distance value, the larger the weight coefficient for the adjacent node. The specific calculation method is to take the first... Electrical distance values ​​of adjacent nodes Take the reciprocal, then divide the reciprocal by the sum of the reciprocals of the electrical distance values ​​of all adjacent nodes to obtain the nth... The weight coefficients of each neighboring node Weighting coefficient The value of is between zero and one, and the sum of all weight coefficients equals one. The weighted average input signal is obtained by multiplying the mixed-order restraint control input signal of each adjacent node by its corresponding weight coefficient, summing the results, and then dividing by the sum of all weight coefficients. Weighted average input signal The unit is the same as that of the mixed-order tether control input signal, which is watt.

[0051] In practice, based on the node type of the non-restrained node, the corresponding control law transfer function, state equation, or differential equation is selected. The weighted average input signal is used as the input variable of the control law transfer function, state equation, or differential equation. A numerical integration calculation or differential iteration calculation is performed to obtain the preliminary control quantity of the non-restrained node.

[0052] When the non-restrained node is a photovoltaic node, a first-order restraint control transfer function is selected. As a control law, the weighted average input signal After one execution, Euler differential calculations are performed to obtain the initial control parameters of the photovoltaic node. The backward Euler difference calculation process involves converting the first-order control transfer function... This is converted to a discrete-time difference equation, and the expression for the discrete-time difference equation is:

[0053] in, Indicates the current control cycle The initial control quantity, Indicates the current control cycle The weighted average input signal, Indicates the previous control cycle The initial control quantity, This indicates the control cycle length, with a value of one-hundredth of a second. This represents the first-order inertial time constant of the photovoltaic node. This represents the proportional gain for photovoltaic node restraint control. Preliminary control quantity. Characterizes the maximum power point tracking offset of a photovoltaic node, in watts.

[0054] When the non-restrained node is an energy storage node, a second-order restrained control state equation is selected as the control law, and the weighted average input signal is used. Decompose into proportional input signal With differential input signal The values ​​are substituted into the proportional and differential channels of the second-order control state equation for parallel calculation. The results from the two channels are then superimposed to obtain the preliminary control parameters for the energy storage node. The decomposition method is as follows: proportional input signal. Equal to weighted average input signal Differential input signal Equal to weighted average input signal The difference between the current control cycle and the previous control cycle is divided by the control cycle length. The parallel computing process is as follows, and the proportional channel calculation formula is: The formula for calculating the differential channel is: ,in The second-order natural oscillation frequency of the energy storage node is expressed in radians per second; the formula for superposition operation is: Preliminary control quantity The power charge-discharge conversion rate characterizes the energy storage node, measured in watts.

[0055] When the non-restrained node is a flexible load node, a fractional-order restraint control differential equation is selected as the control law, and the weighted average input signal is applied. Perform a fractional numerical approximation calculation based on the Green's function to obtain the preliminary control quantity of the flexible load node. The operation process of fractional numerical approximation calculation based on Green's function is as follows: The fractional-order constraint control differential equation... Transform into discrete-time convolution form ,in The Green's function memory kernel represents the fractional differential operator of Caputo. The calculation formula is , Represents the number of steps in memory. Get the memory length from 1 , The value is one hundred. Indicates Let gamma function be the independent variable. This indicates the fractional order of the flexible load node. This represents the proportional coefficient for the flexible load node restraint control. Indicates the control cycle length. Green's function memory kernel. When a flexible load node first participates in regulation, the calculation is pre-calculated offline and stored in the controller memory of the flexible load node. During online calculation, the pre-stored Green's function memory kernel is directly called. with weighted average input signal Perform convolution operations. Initial control variables. The load reduction ratio of a flexible load node is characterized by a value between zero and one.

[0056] In practice, the historical control quantities of the non-controlled node in the previous control cycle are obtained. The preliminary control quantities and historical control quantities are then weighted and fused to obtain the mixed-order control quantity of the non-controlled node in the current control cycle. The historical control quantities are obtained by reading the mixed-order control quantities stored for the non-controlled node in the previous control cycle from the control quantity history database of the virtual power plant control center. The formula for weighted fusion is:

[0057] in, This represents the mixed-order control quantity of the current control cycle. This indicates the initial control quantity for the current control cycle. This represents the smoothing coefficient, with a value of 0.3. Smoothing coefficient The value of 0.3 is based on offline simulation comparison testing of the decay rate of the oscillation amplitude of the non-restrained node control quantity under different smoothness coefficient values. When the value is 0.3, the control oscillation decays to less than 10% of the steady-state amplitude within five control cycles, balancing control response speed and steady-state accuracy.

[0058] In practical implementation, for any constrained node, the real-time power deviation and real-time frequency deviation of the constrained node are directly substituted into the control law corresponding to the node type of the constrained node for calculation, and the mixed-order control quantity of the constrained node is output. The method for determining the node type of the constrained node is the same as that for non-constrained nodes, which is read from the vertex labels of the directed attribute graph. When the node type of the constrained node is "photovoltaic", the real-time power deviation of the constrained node is... Deviation from real-time frequency The equivalent input signal of the restraining node is synthesized by weighted summation. The equivalent input signal As a first-order restraint control transfer function The input is used to calculate the mixed-order control quantity of the restraining node through backward Euler difference calculation; when the node type of the restraining node is "energy storage", the equivalent input signal is... Substituting the decomposition results into the proportional and differential channels of the second-order restraint control state equation yields the mixed-order control quantity; when the restraint node type is "flexible load", the equivalent input signal is... The mixed-order control quantity is obtained by substituting the fractional numerical approximation based on the Green's function.

[0059] In practical implementation, the hybrid-order control quantity calculated by each non-constrained node is converted into a power adjustment command for the node, and the power adjustment command is sent to the corresponding photovoltaic node, energy storage node, or flexible load node to adjust the active power output value of the node. The conversion method of the power adjustment command is as follows: when the node is a photovoltaic node, the content of the power adjustment command is to offset the maximum power point tracking set voltage value of the photovoltaic inverter from the hybrid-order control quantity. The corresponding voltage adjustment amount; when the node is an energy storage node, the power adjustment command is the sum of the active power command value of the energy storage converter and the hybrid-stage control amount. When the node is a flexible load node, the power adjustment command is to reduce the active power consumption of the flexible load terminal by a mixed-order control quantity. The product of the rated active power. Power adjustment commands are encapsulated using the IEC61850 MMS protocol and sent to the corresponding intelligent electronic devices at the appropriate nodes.

[0060] In practice, at the end of each control cycle, the real-time power deviation and real-time frequency deviation of all control nodes are re-acquired. The re-acquisition operation is the same as the real-time acquisition operation. In the sampling interrupt service routine aligned with the control cycle boundary, the current active power output value and the current grid frequency of each control node are synchronously read, and the real-time power deviation is recalculated. Deviation from real-time frequency .

[0061] In practical implementation, the absolute change in real-time power deviation and the absolute change in real-time frequency deviation of each restraint node between the current control cycle and the previous control cycle are calculated. The absolute change in real-time power deviation... The formula for calculation is:

[0062] in: This indicates the number of times the data was re-acquired during the current control cycle. Real-time power deviation of each restraining node This indicates the number stored in the previous control cycle. Real-time power deviation of each restraining node This indicates the absolute value operation. The absolute change in real-time frequency deviation. The calculation formula is ,in This indicates the number of times the data was re-acquired during the current control cycle. Real-time frequency deviation of each restraining node This indicates the number stored in the previous control cycle. Real-time frequency deviation of each restraining node.

[0063] In practical implementation, it is determined whether the absolute change in real-time power deviation and the absolute change in real-time frequency deviation are both less than their respective preset convergence thresholds. The convergence threshold for real-time power deviation is... The value is the total rated power of the virtual power plant. One-thousandth. The convergence threshold for real-time frequency deviation. The value is 0.01 Hz. The judgment operation involves comparing each constraint node sequentially. and Size relationship and and The size relationship, when all restraining nodes simultaneously satisfy and When all conditions are met, it is determined that all values ​​are less than the convergence threshold; when any constraint node does not meet any of the conditions, it is determined that any value is not less than the convergence threshold.

[0064] In practice, if all values ​​are less than the convergence threshold, the sum of squares of the real-time power deviations of all restraining nodes within the current control cycle is further calculated to obtain the global power deviation energy value. The calculation formula is ,in This represents the total number of restraining nodes. This indicates the number of times the data was re-acquired during the current control cycle. The real-time power deviation of each restraint node is calculated. The sum of the squares of the real-time frequency deviations of all restraint nodes within the current control cycle is then calculated to obtain the global frequency deviation energy value. The calculation formula is ,in This indicates the number of times the data was re-acquired during the current control cycle. Real-time frequency deviation of each control node. Obtain historical global power deviation energy values ​​stored in the previous control cycle. Energy value deviating from historical global frequency Historical global power deviation energy value Energy value deviating from historical global frequency Read from the energy history database of the virtual power plant control center. Compare the current global power deviation energy value. Energy value of deviation from historical global power The magnitude of the deviation and the current global frequency deviation energy value are compared. Energy value deviating from historical global frequency The magnitude of the current global power deviation energy value. Less than the historical global power deviation energy value And the current global frequency deviation energy value Less than the historical global frequency deviation energy value Then all the restraint coefficients in the current hybrid-order restraint control law remain unchanged. If the current global power deviation energy value... Not less than the historical global power deviation energy value Or the current global frequency deviation energy value Not less than the historical global frequency deviation energy value This triggers a linear increase in the restraint coefficient.

[0065] In practical implementation, if any value is not less than the convergence threshold, the constraint coefficient corresponding to the constraint node is linearly increased based on the ratio of the absolute change in real-time power deviation to the absolute change in real-time frequency deviation. The updated constraint coefficient is then substituted into the hybrid-order constraint control law for iterative calculation in the next control cycle. The process of linearly increasing the constraint coefficient is as follows: for the ... Each constraint node is used to calculate the ratio. ,in To prevent small quantities from being divided by zero, It takes the value of 10 to the power of negative 6. The formula for updating the restraint coefficient is: ,in Indicates the updated number The constraint coefficient of each constraint node. Indicates the number before the update The constraint coefficient of each constraint node. This indicates that the step size of the confinement coefficient is increased. The value is 0.05. (Control coefficient) In a mixed-order restraint control law, the restraint coefficient is applied in the following way: Multiplied by the mixed-order tethering control input signal of the tethering node This represents the effective restraining signal strength of the restraining node on its adjacent nodes. The restraining coefficients in the hybrid-order restraining control law are adaptively adjusted based on the weighted sum of the real-time power deviation and frequency deviation of the restraining node. These weighting coefficients are determined by the voltage sensitivity of the restraining node. (Weighting coefficients) The calculation method is to obtain the first... Voltage sensitivity of each control node and grid connection point Voltage sensitivity Defined as the ratio of the change in reactive power injection at the grid connection point to the change in voltage amplitude, in volts per volt (V / var), and includes voltage sensitivity. The absolute value divided by the reference voltage sensitivity Reference voltage sensitivity The value is taken as 0.1 volts per volt, and the weighting coefficient is obtained. Real-time power deviation Deviation from real-time frequency weighted sum As a basis for adaptive adjustment, the constraint coefficient increases when the weighted sum increases. Increasing the step size is used in the next control cycle update. Multiply by 1.5 for calculation.

[0066] In specific implementation, please refer to Figure 3 After adjusting the power output of each node according to the hybrid-order control quantity and dynamically updating the restraint coefficient in the hybrid-order restraint control law based on the state feedback of the restraint node, the actual power output values ​​of all photovoltaic nodes, energy storage nodes, and flexible load nodes after power adjustment are obtained. The actual power output values ​​are obtained as follows: at the end of each control cycle, the power acquisition module of the virtual power plant control center sends a power read request message to the intelligent electronic devices of all photovoltaic nodes, energy storage nodes, and flexible load nodes in the virtual power plant hierarchical control model via the IEC61850MMS protocol. The power read request message includes the node number field of the target node and the active power register address field of the requested node. After receiving the power read request message, the intelligent electronic device reads the instantaneous active power value at the current sampling time from the active power register address and encapsulates the instantaneous active power value into a power read response message and sends it back to the power acquisition module. The power acquisition module uses the received instantaneous active power value as the actual power output value of the corresponding node. The unit is watts, and the subscript is... Indicates the node number.

[0067] In practical implementation, the load factor of each node is calculated by dividing its actual power output by its rated power capacity, based on the rated power capacity of each node. The rated power capacity is obtained by reading the node's parameters from the equipment parameter database of the virtual power plant energy management system. Rated power capacity Rated power capacity Represents a node The maximum active power output allowed under continuous operating conditions, expressed in watts. For photovoltaic nodes, rated power capacity. The value is the rated AC output power of the photovoltaic inverter; for energy storage nodes, the rated power capacity is... The value is taken as the rated charging and discharging power of the energy storage converter; for flexible load nodes, the rated power capacity is... The value is the rated active power consumption of the flexible load terminal when it is not participating in power reduction. Load rate The calculation formula is:

[0068] in, Represents a node The load factor is a dimensionless quantity, ranging from zero to a value exceeding 1.0. Represents a node The actual power output value, in watts; Represents a node The rated power capacity, expressed in watts. When the node is a photovoltaic node or an energy storage node. The positive power value represents the active power injected into the grid. When the node is a flexible load node, this is taken as an example. Taking the positive power value represents the active power absorbed from the power grid.

[0069] In practical implementation, it is determined whether there are photovoltaic nodes, energy storage nodes, or flexible load nodes in the virtual power plant hierarchical control model whose load rates exceed a preset overload threshold. Preset overload threshold. The setting method is to determine the overload threshold based on the equipment thermal capacity limits and safe operation specifications of various types of nodes within the virtual power plant. Set to 1.0. When the node load rate When the load exceeds 1.0, it indicates that the actual power output of the corresponding node has exceeded its rated power capacity, and the node is in an overload operation state; when the node load rate A value less than or equal to 1.0 indicates that the node is operating within its rated power capacity range. The specific process for this judgment operation involves iterating through the load factor values ​​of all photovoltaic nodes, energy storage nodes, and flexible load nodes in the virtual power plant hierarchical control model, and setting the load factor of each node... With overload threshold Compare them one by one, when there exists any node Greater than When a node in the virtual power plant hierarchical control model has a load rate exceeding a preset overload threshold is identified; when all nodes... All less than or equal to When the load rate exceeds the preset overload threshold, it is determined that there are no nodes.

[0070] In practice, if a node's load rate exceeds a preset overload threshold, its node type label is temporarily changed to a restricted node, and its weight coefficient in the next control cycle is reduced. The temporary modification of the node type label involves locating the vertex of the node with the load rate exceeding the preset overload threshold in the directed attribute graph of the virtual power plant hierarchical control model. The current value of the vertex label field is stored in the temporary label backup area. Then, the value of the vertex label field is changed from the original "photovoltaic," "energy storage," or "flexible load" to "restricted," and an overload flag is written to the vertex attribute field of the restricted node, with the overload flag value set to 1. Simultaneously, the control cycle number where the overload flag is set to 1 is recorded. The reduction of the node's weight coefficient in the next control cycle involves, before the start of the next control cycle, when the restricted node broadcasts a hybrid-order restraint control input signal to its adjacent non-restrained nodes, the weight coefficient calculated based on the reciprocal normalization of the electrical distance value is changed when calculating the weight coefficient corresponding to the restricted node. Multiply by the reduction factor Reduce factor The value is set to 0.5, reducing the factor. The value of 0.5 is chosen based on the impact of different reduction factor values ​​on the convergence time of power oscillations in adjacent regions of a restricted node under overload conditions simulated in an offline simulation environment. When the value is 0.5, the power oscillation convergence time of the topology region where the restricted node is located is shortened by more than 40% compared with the case where the weight coefficient is not reduced. At the same time, it avoids the restricted node from leaving the power regulation coordination system due to the excessive reduction of the weight coefficient.

[0071] In practical implementation, based on the distribution locations of all restricted nodes, weak areas with insufficient power regulation capability in the virtual power plant hierarchical control model are identified. The distribution locations are obtained by reading the three-dimensional spatial location vectors of all restricted nodes from the directed attribute graph of the virtual power plant hierarchical control model. These three-dimensional spatial location vectors consist of latitude and longitude coordinates and altitude data from their geographical access locations. The identification process for weak areas involves using a density-based spatial clustering algorithm to cluster the three-dimensional spatial location vectors of the restricted nodes. The neighborhood radius parameter of the density-based spatial clustering algorithm... The value is twice the average electrical distance between all restricted nodes; the minimum neighbor number parameter for the core point. A value of three marks the electrical region covered by the voltage level control layer where all nodes of each cluster are located, as well as the electrical region covered by unrestricted nodes that have direct electrical connections with the nodes of that cluster within the adjacent voltage level control layer, as a weak region.

[0072] In practical implementation, at the start of the next control cycle, the restraint coefficients of all nodes within the weak region are increased. The scope of nodes within the weak region is determined by using each cluster output by a density-based spatial clustering algorithm as a benchmark. The list of neighboring nodes for each restricted node within a cluster is queried from the directed attribute graph of the virtual power plant hierarchical control model. The set of nodes formed by all restricted nodes within a cluster and all their neighboring nodes in the neighboring node list is considered the weak region node set. The process of increasing the restraint coefficients involves, for each node in the weak region node set, before executing the iterative calculation of the mixed-order restraint control law at the start of the next control cycle, the currently stored restraint coefficient of the node is increased. With the increase of the restraint coefficient Add them together to get the increased restraint coefficient. Increment of the restraint coefficient The value is 0.1, and the increment of the restraint coefficient is... The basis for setting the value to 0.1 is that offline testing in the parameter tuning module of the virtual power plant control center revealed the impact of different incremental values ​​on the convergence speed of power deviation in the weak area. When the incremental value was 0.1, the global power deviation energy value in the weak area decreased to less than 10% of the initial value within ten control cycles, without causing power control oscillations in adjacent non-weak areas. The increased restraint coefficient... The added constraint coefficient is written to the vertex attribute field of the corresponding node in the directed attribute graph of the virtual power plant hierarchical control model, and used in the iterative calculation of the hybrid-order constraint control law. Participate in the calculation of the control law.

[0073] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes 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.

Claims

1. A virtual power plant power regulation method based on hybrid-order restraint control, characterized in that, include: A hierarchical control model of a virtual power plant, including photovoltaic (PV), energy storage, and flexible load nodes, is constructed. Based on the differences in the response time constants of each node, a first-order restraint control law is configured for the PV nodes, a second-order restraint control law for the energy storage nodes, and a fractional-order restraint control law for the flexible load nodes. Several nodes in the virtual power plant hierarchical control model with connectivity exceeding a preset threshold are selected as restraint nodes, and their real-time power deviation and frequency deviation are collected. The real-time power deviation and frequency deviation are input into a hybrid-order restraint control law, and the hybrid-order control quantity of each node is calculated through distributed iteration. The power output of each node is adjusted according to the hybrid-order control quantity, and the restraint coefficient in the hybrid-order restraint control law is dynamically updated based on the state feedback of the restraint nodes. The specific steps for selecting several nodes in the virtual power plant hierarchical control model with connectivity exceeding a preset threshold as control nodes and collecting the real-time power deviation and frequency deviation of the control nodes are as follows: Calculate the number of direct electrical connections between each node and other nodes in the hierarchical control model of the virtual power plant, and use the number of direct electrical connections as the connectivity of that node. Compare the connectivity of each node with a preset connectivity threshold, and mark nodes with connectivity greater than the connectivity threshold as candidate restraining nodes; A preset number of nodes are randomly selected from the candidate restraining nodes as restraining nodes, and the node number of each restraining node in the virtual power plant hierarchical control model is recorded. Based on the node number, the current active power output value and the rated active power reference value of each restraining node are collected in real time, and the difference between the current active power output value and the rated active power reference value is calculated to obtain the real-time power deviation of each restraining node. The current power grid frequency and the rated frequency reference value of each control node are collected synchronously, and the difference between the current power grid frequency and the rated frequency reference value is calculated to obtain the real-time frequency deviation of each control node.

2. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 1, characterized in that, The specific steps for constructing a hierarchical control model for a virtual power plant that includes photovoltaic, energy storage, and flexible load nodes are as follows: Obtain the geographical access locations and electrical connections of all distributed power sources and controllable loads within the area covered by the virtual power plant; The virtual power plant is divided into several voltage level control layers based on the geographical access location, and photovoltaic node sets, energy storage node sets, and flexible load node sets are extracted in each voltage level control layer according to the electrical connection relationship. Acquire the maximum power point tracking status data of each photovoltaic node, acquire the state of charge data of each energy storage node, and acquire the interruptible power capacity data of each flexible load node. The maximum power point tracking status data, state of charge data, and interruptible power capacity data are used as node attributes and associated with the corresponding nodes for storage, generating a virtual power plant hierarchical control model that includes node type labels and node attribute fields.

3. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 2, characterized in that, The specific steps for configuring a first-order restraint control law for the photovoltaic nodes, a second-order restraint control law for the energy storage nodes, and a fractional-order restraint control law for the flexible load nodes in the hierarchical control model of the virtual power plant based on the differences in the response time constants of each node are as follows: Based on the maximum power point tracking response time of each photovoltaic node in the photovoltaic node set, the first-order inertial time constant of the photovoltaic node is determined, and a first-order restraint control transfer function is constructed based on the first-order inertial time constant. Based on the power conversion system response time and battery dynamic response time of each energy storage node in the energy storage node set, the second-order natural oscillation frequency and damping ratio of the energy storage node are determined, and a second-order restraint control state equation is constructed based on the second-order natural oscillation frequency and damping ratio. Based on the load aggregation time window of each flexible load node in the set of flexible load nodes, the fractional order of the flexible load node is determined, and a fractional-order constraint control differential equation is constructed based on the fractional order.

4. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 3, characterized in that, The fractional order of the fractional-order constraint control differential equation is adjusted in real time according to the dynamic changes of the load aggregation time window of the flexible load node, and the adjustment range is proportional to the rate of change of the load aggregation time window.

5. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 4, characterized in that, The specific steps for inputting the real-time power deviation and frequency deviation into the hybrid-order restraint control law, and calculating the hybrid-order control quantity of each node through distributed iterative calculation are as follows: Based on the real-time power deviation and real-time frequency deviation of each restraining node, a hybrid-order restraining control input signal for that restraining node is generated. The hybrid-order tethering control input signal is broadcast to all adjacent nodes that have a direct electrical connection with the tethering node; For any non-restrained node in the hierarchical control model of the virtual power plant, it receives mixed-order restraint control input signals broadcast by all adjacent restraint nodes or adjacent non-restrained nodes. Based on the node type of the non-restrained node, select the corresponding first-order restraint control law, second-order restraint control law, or fractional-order restraint control law, and take the weighted average of all received mixed-order restraint control input signals and substitute it into the selected control law for iterative calculation, outputting the mixed-order control quantity of the non-restrained node. For any constrained node, the real-time power deviation and real-time frequency deviation of the constrained node are directly substituted into the control law corresponding to its node type for calculation, and the mixed-order control quantity of the constrained node is output.

6. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 5, characterized in that, When adjacent nodes broadcast mixed-order tethering control input signals, a consensus algorithm is used to filter the signals and remove phase lag components caused by communication delays.

7. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 6, characterized in that, The specific steps for adjusting the power output of each node according to the hybrid-order control quantity and dynamically updating the restraint coefficient in the hybrid-order restraint control law based on the state feedback of the restraint node are as follows: The hybrid-order control quantity calculated by each non-restrained node is converted into a power adjustment command for that node, and the power adjustment command is sent to the corresponding photovoltaic node, energy storage node or flexible load node to adjust the active power output value of that node. At the end of each control cycle, the real-time power deviation and real-time frequency deviation of all restraint nodes are reacquired. Calculate the absolute change in real-time power deviation and the absolute change in real-time frequency deviation for each restraint node between the current control cycle and the previous control cycle. Determine whether the absolute change of the real-time power deviation and the absolute change of the real-time frequency deviation are both less than their respective preset convergence thresholds. If all are less than the convergence threshold, then all constraint coefficients in the current mixed-order constraint control law remain unchanged. If any value is not less than the convergence threshold, then the restraining coefficient corresponding to the restraining node is linearly increased according to the ratio of the absolute change of the real-time power deviation to the absolute change of the real-time frequency deviation, and the updated restraining coefficient is substituted into the hybrid-order restraining control law for iterative calculation in the next control cycle.

8. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 7, characterized in that, The restraint coefficient in the hybrid-order restraint control law is adaptively adjusted based on the weighted sum of the real-time power deviation and frequency deviation of the restraint node, and the weighting coefficient is determined by the voltage sensitivity of the restraint node.

9. The virtual power plant power regulation method based on hybrid-order restraint control according to claim 8, characterized in that, After determining whether the absolute change in the real-time power deviation and the absolute change in the real-time frequency deviation are both less than their respective preset convergence thresholds, the following operation is also included: If all values ​​are less than the convergence threshold, the sum of the squares of the real-time power deviations of all restraining nodes in the current control cycle is further calculated to obtain the global power deviation energy value. Calculate the sum of squares of the real-time frequency deviations of all restraining nodes within the current control cycle to obtain the global frequency deviation energy value; Obtain the historical global power deviation energy value and historical global frequency deviation energy value stored in the previous control cycle; Compare the current global power deviation energy value with the historical global power deviation energy value, and compare the current global frequency deviation energy value with the historical global frequency deviation energy value; If the current global power deviation energy value is less than the historical global power deviation energy value and the current global frequency deviation energy value is less than the historical global frequency deviation energy value, then all the restraining coefficients in the current hybrid-order restraining control law remain unchanged. If the current global power deviation energy value is not less than the historical global power deviation energy value or the current global frequency deviation energy value is not less than the historical global frequency deviation energy value, then the linear increase operation of the restraint coefficient is triggered.

Images