A method, device and equipment for adaptive optimization control of a network-structured multi-machine system

Abstract

The application belongs to the field of power system control, and discloses a self-adaptive optimization control method, device and equipment for a network-structured multi-machine system, constructs a frequency modulation hard constraint; constructs a virtual inertia and other decision variables based on a particle swarm optimization algorithm, performs parameter optimization iteration analysis, obtains a benchmark virtual inertia and a benchmark frequency modulation coefficient; combines real-time frequency parameters, respectively performs dynamic self-adaptive adjustment analysis according to the benchmark virtual inertia and the benchmark frequency modulation coefficient, generates adaptive virtual inertia and damping coefficients; combines SOX parameters to calculate the distribution weight of each energy storage unit; distributes the capacity and virtual inertia of each energy storage unit according to the distribution weight, the adaptive virtual inertia and the damping coefficients, and obtains an active power reference value; and realizes self-adaptive optimization control of the active power reference value. The application can solve the technical problem that the prior art relies on fixed inertia, damping parameters, distribution ratios and artificial experience, and does not perform targeted collaborative optimization based on core related parameters and constraints.

Description

Technical Field

[0001] This application relates to the field of power system control, and in particular to an adaptive optimization control method, apparatus and equipment for a grid-type multi-machine system. Background Technology

[0002] With the continuous increase in the high-proportion grid-connected penetration rate of new energy sources, the proportion of synchronous power sources in the power system is declining, and the problems of insufficient system inertia and primary frequency regulation capabilities, as well as weakened frequency support capabilities, are becoming increasingly prominent. Grid-based energy storage, by simulating the electromechanical transient characteristics of synchronous generators, can autonomously construct voltage and frequency references, possessing active inertia support, primary frequency regulation, and black-start capabilities, making it a core technological path for improving the frequency stability of new power systems. Among them, virtual synchronous machine control technology serves as the core framework, embedding rotor motion equations to make the converter exhibit characteristics outside the voltage source, providing equivalent inertia and damping support for the grid.

[0003] Current research on frequency control for grid-based energy storage has made progress in multiple directions: at the single-unit level, optimization is being carried out on parameters such as virtual inertia and damping coefficient to improve frequency dynamic response performance; at the multi-unit cluster level, power distribution and state-of-charge (SOC) balancing strategies are being explored to ensure coordinated operation of multiple units; and at the planning level, research is being conducted on energy storage capacity and frequency regulation parameter configuration to match system frequency constraints. Meanwhile, parallel-connected energy storage power station topologies are being gradually promoted, energy management systems have achieved basic power dispatching and SOC coordination functions, and the frequency support system for grid-based energy storage has a preliminary foundation for engineering applications.

[0004] However, existing technologies still have systemic shortcomings: single-machine control uses fixed inertia and damping parameters, making it difficult to balance the strength of frequency drop suppression and recovery speed; multi-machine allocation often adopts a fixed ratio allocation strategy, without comprehensively considering the state of charge and battery health, which can easily lead to some units reaching their limits and exiting prematurely, weakening the overall support continuity; at the planning level, parameter configuration relies heavily on engineering experience, and various parameters and system small disturbance stability constraints are not incorporated into the collaborative optimization framework, making it difficult to balance stability performance and investment economy, and a complete closed loop from offline planning to online control has not been formed. Summary of the Invention

[0005] This application provides an adaptive optimization control method, apparatus, and equipment for a networked multi-machine system, which addresses the technical problem that existing technologies rely on fixed inertia, damping parameters, allocation ratios, and human experience, without performing targeted collaborative optimization based on core relevant parameters and constraints, resulting in a lack of stability and reliability in control.

[0006] In view of this, the first aspect of this application provides an adaptive optimization control method for a networked multi-machine system, comprising: An equivalent model of the system frequency response is constructed for a network-type multi-machine system, and a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient is performed to obtain the frequency modulation hard constraint. Based on the particle swarm optimization algorithm, the virtual inertia, the total primary frequency regulation coefficient, and the energy storage capacity of each site are used as decision variables. The parameter optimization iterative analysis is performed based on preset constraints to generate the benchmark virtual inertia and benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints. By combining real-time frequency parameters, dynamic adaptive adjustment analysis is performed based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. The real-time frequency parameters include frequency deviation and frequency change rate. After calculating the available charge / discharge capacity of each energy storage unit based on the SOX parameters, the allocation weight is calculated based on the available charge / discharge capacity. Based on the allocation weight, the adaptive virtual inertia, and the adaptive damping coefficient, capacity allocation and virtual inertia allocation are performed for each energy storage unit to obtain the active power reference value of each energy storage unit. The active power reference value is transmitted to the converter to realize adaptive optimization control of the grid-type multi-machine system.

[0007] Preferably, the step involves constructing an equivalent model of the system frequency response for a network-type multi-machine system and performing a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain frequency modulation hard constraints, including: By taking the grid-type energy storage, conventional synchronous power supply and load in the grid-type multi-machine system as a single inertia center, a frequency dynamic time-domain equation based on the rotor swing equation is generated. Ignoring the speed governor's action delay, a Laplace transform is performed on the frequency dynamic time-domain equation to derive the s-domain transfer function of the frequency deviation to the power disturbance, generating an equivalent model of the system frequency response. After introducing a safety factor, the relationship between the steady-state frequency deviation and the total primary frequency regulation coefficient is analyzed based on the equivalent model of the system frequency response to generate the frequency regulation hard constraint.

[0008] Preferably, the step of constructing an equivalent model of the system frequency response for a network-type multi-machine system and performing a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain frequency modulation hard constraints, further includes: Constructing the system node admittance matrix based on the topology and component parameters of a network-type multi-machine system; Solve the characteristic equations based on the system node admittance matrix to obtain all the characteristic roots of the system; Based on the eigenvalues, calculate the stability margin characterizing the oscillation decay rate and the dominant mode damping ratio characterizing the oscillation suppression capability, respectively. Stability margin constraints and damping ratio constraints are constructed using the stability margin and the dominant mode damping ratio, respectively.

[0009] Preferably, the step of combining real-time frequency parameters and performing dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient includes: The actual frequency and real-time SOC of the network-type multi-machine system ports are collected in real time, and the frequency deviation and frequency change rate are calculated based on the actual frequency to determine the real-time frequency parameters. By combining the reference virtual inertia and the inertia adjustment gain, the virtual inertia is dynamically adjusted according to the amplitude of the frequency change rate and the sign of the frequency deviation to generate an adaptive virtual inertia. By combining the reference frequency modulation coefficient and the damping adjustment gain, the damping coefficient is dynamically adjusted according to the amplitude of the frequency deviation to generate an adaptive damping coefficient.

[0010] Preferably, the step of combining real-time frequency parameters and performing dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient further includes: A SOC correction factor is introduced to smoothly correct the real-time SOC based on the SOC reference range, and the frequency modulation output is softly limited. Based on the adaptive virtual inertia, the adaptive damping coefficient, and the SOC correction factor, active power adjustment calculations are performed to obtain a reference value for total active power.

[0011] Preferably, the step of calculating the available charge / discharge capacity of each energy storage unit by combining SOX parameters, and then calculating the allocation weight based on the available charge / discharge capacity, includes: The available charge / discharge capacity of each energy storage unit is calculated by combining the SOX parameters and the rated energy, wherein the SOX parameters include SOC and SOH; Calculate the ratio of the available charge / discharge capacity of each energy storage unit to the total available charge / discharge capacity, and determine the allocation weight.

[0012] Preferably, the step of allocating capacity and virtual inertia for each energy storage unit based on the allocation weight, the adaptive virtual inertia, and the adaptive damping coefficient to obtain the active power reference value for each energy storage unit includes: Calculate the total primary frequency regulation power requirement for the entire station based on the aforementioned reference frequency regulation coefficient; The capacity allocation of each energy storage unit is calculated based on the allocation weight and the total primary frequency regulation power requirement, and the active power of the unit is determined. The virtual inertia allocation is calculated based on the allocation weight and the adaptive virtual inertia to obtain the virtual inertia allocation value; The active power reference value of each energy storage unit is calculated by combining the active power of the unit and the virtual inertia allocation value.

[0013] A second aspect of this application provides an adaptive optimization control device for a networked multi-machine system, comprising: The constraint construction unit is used to build an equivalent model of the system frequency response for a network-type multi-machine system, and to perform frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain frequency modulation hard constraints. The parameter optimization unit is used to perform parameter optimization iterative analysis based on preset constraints, taking virtual inertia, total primary frequency regulation coefficient and energy storage capacity of each site as decision variables based on particle swarm optimization algorithm, and generating benchmark virtual inertia and benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints. The dynamic adjustment unit is used to combine real-time frequency parameters and perform dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. The real-time frequency parameters include frequency deviation and frequency change rate. The weight calculation unit is used to calculate the available charge / discharge capacity of each energy storage unit in combination with the SOX parameters, and then calculate the allocation weight based on the available charge / discharge capacity. The allocation calculation unit is used to allocate capacity and virtual inertia for each energy storage unit according to the allocation weight, the adaptive virtual inertia and the adaptive damping coefficient, and to obtain the active power reference value of each energy storage unit. An optimization control unit is used to transmit the active power reference value to the converter to realize adaptive optimization control of the grid-type multi-machine system.

[0014] Preferably, the constraint construction unit is specifically used for: By taking the grid-type energy storage, conventional synchronous power supply and load in the grid-type multi-machine system as a single inertia center, a frequency dynamic time-domain equation based on the rotor swing equation is generated. Ignoring the speed governor's action delay, a Laplace transform is performed on the frequency dynamic time-domain equation to derive the s-domain transfer function of the frequency deviation to the power disturbance, generating an equivalent model of the system frequency response. After introducing a safety factor, the relationship between the steady-state frequency deviation and the total primary frequency regulation coefficient is analyzed based on the equivalent model of the system frequency response to generate the frequency regulation hard constraint.

[0015] A third aspect of this application provides an adaptive optimization control device for a networked multi-machine system, the device including a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the adaptive optimization control method for the networked multi-machine system described in the first aspect according to the instructions in the program code.

[0016] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: This application provides an adaptive optimization control method for a network-type multi-machine system, comprising: constructing an equivalent model of the system frequency response for the network-type multi-machine system and performing a frequency minimum point constraint analysis based on the total primary frequency regulation coefficient to obtain frequency regulation hard constraints; using the particle swarm optimization algorithm, taking the virtual inertia, the total primary frequency regulation coefficient, and the energy storage capacity of each site as decision variables, performing parameter optimization iterative analysis based on preset constraints to generate a reference virtual inertia and a reference frequency regulation coefficient, the preset constraints including frequency regulation hard constraints; combining real-time frequency parameters, performing dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency regulation coefficient to generate an adaptive virtual inertia and an adaptive damping coefficient, the real-time frequency parameters including frequency deviation and frequency change rate; calculating the available charge / discharge capacity of each energy storage unit based on SOX parameters, and calculating the allocation weight based on the available charge / discharge capacity; allocating capacity and virtual inertia for each energy storage unit based on the allocation weight, the adaptive virtual inertia, and the adaptive damping coefficient to obtain the active power reference value of each energy storage unit; and transmitting the active power reference value to the converter to realize the adaptive optimization control of the network-type multi-machine system.

[0017] The adaptive optimization control method for a network-type multi-machine system provided in this application derives frequency regulation hard constraints by constructing an equivalent model of the system's frequency response. Then, it combines this with a particle swarm optimization algorithm to comprehensively consider the control effects of various parameters in the system. Based on preset constraints, it generates optimal baseline parameter information. Quantitative tuning replaces empirical configuration, balancing stability margin and investment economy, overcoming reliance on experience, and filling the gap in the coordinated optimization of inertia and capacity. Furthermore, using the optimized baseline parameters as the basis for adjustment, it dynamically and adaptively adjusts core parameters in conjunction with real-time frequency parameters, generating adaptive virtual inertia and damping coefficients. This fully considers the impact of dynamic frequency changes on inertia and dynamically matches the operating conditions. Moreover, it calculates the available charge / discharge capacity of each energy storage unit using SOX parameters and generates allocation weights. Active power and virtual inertia are synchronously allocated according to these weights, replacing manual experience. This allows units with sufficient energy margins to undertake more tasks, preventing premature termination due to overcharging or over-discharging of individual units, and improving the overall station's frequency regulation sustainability. Therefore, this application can solve the technical problem that the prior art relies on fixed inertia, damping parameters, allocation ratios and human experience, and does not carry out targeted collaborative optimization based on core related parameters and constraints, resulting in a lack of stability and reliability in control. Attached Figure Description

[0018] Figure 1A flowchart illustrating an adaptive optimization control method for a networked multi-machine system provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of an adaptive optimization control device for a networked multi-machine system provided in an embodiment of this application. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0020] For easier understanding, please refer to Figure 1 An embodiment of an adaptive optimization control method for a networked multi-machine system provided in this application includes: Step 101: Construct an equivalent model of the system frequency response for the network-type multi-machine system, and perform a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain the frequency modulation hard constraint; Step 102: Based on the particle swarm optimization algorithm, the virtual inertia, the total primary frequency regulation coefficient, and the energy storage capacity of each site are used as decision variables. The parameter optimization iterative analysis is performed based on preset constraints to generate the benchmark virtual inertia and the benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints. Step 103: Combining real-time frequency parameters, perform dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. The real-time frequency parameters include frequency deviation and frequency change rate. Step 104: After calculating the available charge / discharge capacity of each energy storage unit based on the SOX parameters, calculate the allocation weight according to the available charge / discharge capacity; Step 105: Based on the allocation weight, the adaptive virtual inertia and the adaptive damping coefficient, capacity allocation and virtual inertia allocation are performed for each energy storage unit to obtain the active power reference value of each energy storage unit. Step 106: Transmit the active power reference value to the converter to realize adaptive optimization control of the grid-type multi-machine system.

[0021] It should be noted that the frequency response equivalent model in this embodiment ignores the governor delay and is constructed by equating the system's internal grid-type energy storage, conventional synchronous power supply, and load to a single inertia center. Based on the system's frequency response equivalent model, frequency-modulated inertia inference analysis can be performed, specifically the frequency-inertia relationship at the lowest frequency point. Based on this, the system's total primary frequency regulation hard constraint, i.e., frequency regulation hard constraint, can be generated.

[0022] It should be noted that the subsequent process of selecting the optimal configuration parameters based on particle swarm optimization needs to satisfy not only the frequency tuning hard constraint of the configuration, but also the verification constraint of the training process, specifically the stability margin constraint and the dominant mode damping ratio constraint. These two constraints are derived step by step based on the system node admittance matrix.

[0023] It should be noted that the decision variables of the particle swarm optimization algorithm in this embodiment include virtual inertia, total primary frequency regulation coefficient, and energy storage capacity of each site. In each iteration, the particles carry a set of particles, and the iteration needs to check whether the hard frequency regulation constraint is met before calculation. Then, it can be checked whether the iteration process meets the stability margin constraint and damping ratio constraint. If all constraints are met, the total energy storage capacity, i.e., the total cost, is calculated. The objective function is to minimize the total cost; otherwise, a maximum penalty is imposed.

[0024] Parameter optimization analysis based on particle swarm optimization algorithm is an offline configuration analysis process. The optimal virtual inertia and optimal frequency modulation coefficient obtained are used as the reference virtual inertia and basic frequency modulation coefficient for subsequent dynamic inertia adjustment and online control analysis tasks, realizing a closed-loop scheme of offline planning boundary, dynamic adjustment processing, and online decomposition execution.

[0025] It should be noted that, since the virtual variable J and damping coefficient D in traditional VSG control are fixed, it is difficult to simultaneously consider frequency drop suppression and recovery speed. In this embodiment, the virtual inertia J and damping coefficient D are dynamically and adaptively adjusted according to real-time frequency parameters such as frequency deviation and frequency change rate, and then the active power control reference value is analyzed based on the adjusted parameters.

[0026] Understandably, real-time frequency parameters such as frequency deviation and rate of change are directly acquired from the system or obtained through basic calculations. The adaptive adjustment of virtual inertia and damping coefficient are two parallel branches, respectively addressing the dynamic process and steady-state deviation of frequency changes. Together, they constitute the core of frequency regulation capability and directly determine the dynamic performance of frequency control. Summarizing these adjusted control quantities generates the final executable active power command for the system, which is then distributed downstream and serves as the closed-loop endpoint of a single energy storage unit's control cycle.

[0027] It should be noted that current SOC protection mechanisms are mostly hard threshold cutoff mechanisms, which are prone to sudden power surges. Furthermore, multi-unit allocation often employs a fixed ratio strategy, failing to comprehensively consider the differences between state of charge and state of health, which can easily lead to some units reaching their limits prematurely and exiting the system, weakening overall support continuity. Therefore, this embodiment fully considers the impact of SOX parameters in its allocation strategy. SOX parameters can include relevant parameters of the battery management system such as SOC, SOH, SOP, and SOE.

[0028] The available charge and discharge capacity calculated by combining SOX parameters generates an allocation weight. Active power and inertia can be allocated according to this weight, so that units with sufficient energy margin can undertake more tasks. This can prevent single units from being overcharged or over-discharged and prematurely exiting the station, thereby improving the station's frequency regulation continuity capability.

[0029] Active power reference values ​​for each energy storage unit The data can be transmitted to the converter for system optimization control, thus achieving adaptive optimization control of a network-type multi-machine system. To adapt to engineering needs, the parameters obtained at each stage can be recorded separately, and the effectiveness of the optimization control can be verified. Specifically, the basic virtual inertia, basic frequency regulation coefficient, and total primary frequency regulation coefficient obtained based on particle swarm optimization iteration can be stored; the adaptive virtual inertia, adaptive damping coefficient, state of charge correction factor, and total active power reference value of the system obtained through dynamic adaptive adjustment can be stored; and the available charging and discharging capacity, virtual inertia allocation value, allocated frequency regulation power, and active power reference value of each energy storage unit determined during the single-machine allocation execution stage can also be stored.

[0030] Based on these optimized configuration parameters, it can be ensured that the energy storage units with more remaining energy and better lifespan output more power, while units nearing the limit output less power. This avoids individual units reaching the protection threshold first and being shut down in a chain reaction, thereby improving the continuous utilization rate of the adjustable capacity of the entire station and the overall operating life.

[0031] The adaptive optimization control method for a network-type multi-machine system provided in this application derives frequency regulation hard constraints by constructing an equivalent model of the system's frequency response. Then, it combines this with a particle swarm optimization algorithm to comprehensively consider the control effects of various parameters in the system. Based on preset constraints, it generates optimal baseline parameter information. Quantitative tuning replaces empirical configuration, balancing stability margin and investment economy, overcoming reliance on experience, and filling the gap in the coordinated optimization of inertia and capacity. Furthermore, using the optimized baseline parameters as the basis for adjustment, it dynamically and adaptively adjusts core parameters in conjunction with real-time frequency parameters, generating adaptive virtual inertia and damping coefficients. This fully considers the impact of dynamic frequency changes on inertia and dynamically matches the operating conditions. Moreover, it calculates the available charging and discharging capacity of each energy storage unit using SOX parameters and generates allocation weights. Active power and virtual inertia are synchronously allocated according to these weights, replacing manual experience. This allows units with sufficient energy margins to undertake more tasks, preventing premature termination due to overcharging or over-discharging of individual units, and improving the overall station's frequency regulation sustainability. Therefore, the embodiments of this application can solve the technical problem that the prior art relies on fixed inertia, damping parameters, allocation ratios and human experience, and does not carry out targeted collaborative optimization based on core related parameters and constraints, resulting in a lack of stability and reliability in control.

[0032] For ease of understanding, another embodiment of the adaptive optimization control method for a networked multi-machine system provided in this application includes: Step 201: Equate the grid-type energy storage, conventional synchronous power supply, and load within the grid-type multi-machine system to a single center of inertia, and generate a frequency dynamic time-domain equation based on the rotor swing equation. Step 202: Ignore the speed governor's action delay, perform a Laplace transform on the frequency dynamic time-domain equation, derive the s-domain transfer function of the frequency deviation to the power disturbance, and generate an equivalent model of the system frequency response. Step 203: After introducing the safety factor, analyze the relationship between the steady-state frequency deviation and the total primary frequency regulation coefficient at the lowest frequency point based on the equivalent model of the system frequency response, and generate frequency regulation hard constraints.

[0033] The equivalent frequency response model in this embodiment ignores the governor delay and constructs an equivalent model by equating the system's internal grid-type energy storage, conventional synchronous power supply, and load to a single inertia center. Based on this equivalent frequency response model, frequency-modulated inertia inference analysis can be performed, specifically the relationship between frequency and inertia at the lowest frequency point. Based on this, the system's total primary frequency regulation hard constraint, i.e., frequency regulation hard constraint, can be generated.

[0034] Specifically, the system frequency response equivalent model in this embodiment is a simplified equivalent model that includes grid-type energy storage, residual synchronous power sources, and loads. This model ignores inter-unit oscillations and equates the system inertia to a single inertia center. , which is the system's equivalent inertia constant, can be used to construct the frequency dynamic time-domain equation; therefore, the frequency dynamic time-domain equation in this embodiment is expressed as:

[0035] in, Indicates the system frequency deviation. The actual frequency value collected can also be directly expressed as f , This indicates the system's rated frequency, measured in Hz. The system's equivalent inertia constant is expressed as a constant value in seconds (s), and is obtained by summing the virtual inertia of each station. This represents the virtual inertia of a site. This is the rated angular frequency; The mechanical power increment of a conventional synchronous power supply can be approximated by neglecting the speed controller's operating delay. That is, a conservative estimate is that energy storage will bear the entire power deficit; This represents the disturbance power magnitude, assuming a step-type power deficit, such as the maximum single-unit failure power, in units of pu. The system baseline capacity is... , are system planning parameters; The total primary frequency regulation coefficient is used to directly define the steady-state active power-frequency droop characteristic of the entire station. Its physical meaning is the total amount of additional active power that the grid-type energy storage power station needs to provide for every 1Hz deviation in system frequency.

[0036] right Perform Laplace transform and order The s-domain transfer function of the frequency deviation to the disturbance can be obtained:

[0037] In step disturbance Under the influence of, among them, , These represent the maximum power disturbance step amplitude values, with the deficit value being positive, such as the maximum single-unit capacity. The negative sign "-" indicates the deficit. Performing an inverse Laplace transform, the time-domain frequency deviation can be obtained:

[0038] in, Monotonically decreasing, its absolute value monotonically increases, and its steady-state deviation can be expressed as: The lowest frequency point occurs immediately after the disturbance, and the deviation value at this lowest frequency point is... This can be expressed as:

[0039] Therefore, if the minimum frequency is required to be no lower than Then the following must be satisfied:

[0040] in, To allow the lowest possible frequency, a value of 49.5Hz can be chosen, that is, if , ; This formula is the stable frequency deviation constraint. The lower limit serves as a constraint on the initial total primary frequency regulation coefficient of the system.

[0041] However, in practical systems, the rate of frequency decrease cannot be too large, so inertia constraints need to be constructed. Since the initial total primary frequency regulation coefficient constraint construction process ignores the governor response and assumes the frequency monotonically decreases to the steady-state value, but in actual systems, due to the proportional effect of the governor, the frequency may decrease and then rise during the transient process, with the lowest point occurring 2-3 seconds after the disturbance, therefore, the s-domain transfer function can incorporate an equivalent model of the first-order governor:

[0042] Therefore, the closed-loop transfer function can be expressed as:

[0043] in, This represents the equivalent time constant of the speed controller, which is obtained based on the factory data or measured data of the synchronous power supply.

[0044] By solving for the extreme points of its time-domain response, a more accurate minimum point constraint can be obtained, but the formula becomes larger and more complex. Therefore, in order to balance optimization efficiency, this embodiment sets:

[0045] Then a safety factor is introduced. Also known as the transient overshoot safety factor, and This allows us to construct frequency modulation hard constraints:

[0046] Among them, the safety factor The value of is determined based on historical simulation experience, and is usually taken as 1.1~1.3; to compensate for the additional overshoot caused by the dynamics of the speed controller; this coefficient can be obtained by Monte Carlo simulation statistics before offline particle swarm optimization operation, and is used as a fixed input parameter into the PSO algorithm.

[0047] Since the minimum frequency is primarily determined by the steady-state gain, provided the inertia constraint is satisfied, the above equation can be embedded as a hard constraint into the particle swarm optimization process: during each iteration's fitness calculation, the following is checked... Check if the frequency modulation hard constraint is met; if not, apply a penalty or invalidate the solution directly.

[0048] Step 204: Construct the system node admittance matrix based on the topology parameters and component parameters of the networked multi-machine system; Step 205: Solve the characteristic equation based on the system node admittance matrix to obtain all the characteristic roots of the system; Step 206: Calculate the stability margin, which characterizes the oscillation decay rate, and the dominant mode damping ratio, which characterizes the oscillation suppression capability, based on the characteristic roots. Step 207: Construct stability margin constraints and damping ratio constraints using stability margin and dominant mode damping ratio, respectively.

[0049] It should be noted that the subsequent process of selecting the optimal configuration parameters based on particle swarm optimization needs to satisfy not only the frequency tuning hard constraint of the configuration, but also the verification constraint of the training process, specifically the stability margin constraint and the dominant mode damping ratio constraint. These two constraints are derived step by step based on the system node admittance matrix.

[0050] Specifically, the system node admittance matrix can be constructed first based on the topology and component parameters of the network-type multi-machine system. Then solve the characteristic equation. The kth eigenvalue obtained by solving can be expressed as: Based on these eigenvalues, the stability margin can be calculated respectively. And dominant mode damping ratio :

[0051] Then, stability margin constraints can be constructed based on the obtained stability margin and dominant mode damping ratio. and damping ratio constraint ,in, , These represent the lower limit of the margin and the lower limit of the damping ratio, respectively. It can take the value 0.5. The value range is 0.03 to 0.05. The stability margin constraint ensures that all oscillation modes decay at a sufficiently fast rate after a disturbance; while the damping ratio constraint prevents the system from exhibiting weakly damped or even negatively damped oscillations; these two constraints can also serve as constraints for the subsequent optimization process of configuration parameters, ensuring the stability and reliability of the optimization.

[0052] Step 208: Using the particle swarm optimization algorithm, the virtual inertia, the total primary frequency regulation coefficient, and the energy storage capacity of each site are used as decision variables. The parameter optimization iterative analysis is performed based on preset constraints to generate the benchmark virtual inertia and benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints.

[0053] It should be noted that the decision variables of the particle swarm optimization algorithm in this embodiment include virtual inertia. Total primary frequency modulation coefficient and energy storage capacity of each site In each iteration, the particle carries a set of... The iteration needs to be verified according to the frequency modulation hard constraint. Whether it meets the standard, and then calculate. Next, it is possible to verify whether the iterative process satisfies the stability margin constraint and damping ratio constraint. If all constraints are satisfied, the total energy storage capacity is then calculated. The objective function is to minimize the total cost, otherwise a maximum penalty is imposed.

[0054] Understandably, in the particle swarm optimization algorithm, this check is performed on each particle in each iteration: if any constraint is not satisfied, the particle is given a very high fitness value, i.e., a penalty, causing it to be naturally eliminated in subsequent searches. Only when both constraints are satisfied... and Only particles that meet the requirements are considered feasible solutions and included in the comparison of the objective function. In this way, the final optimization result can guarantee both the frequency and inertia requirements, as well as the dynamic stability of the system; thus obtaining the optimal combination of configuration parameters, including the optimal virtual inertia, the optimal frequency regulation coefficient, and the energy storage capacity of each site.

[0055] Parameter optimization analysis based on particle swarm optimization algorithm is an offline configuration analysis process. The optimal virtual inertia and optimal frequency modulation coefficient obtained are used as the reference virtual inertia and basic frequency modulation coefficient for subsequent dynamic inertia adjustment and online control analysis tasks, realizing a closed-loop scheme of offline planning boundary, dynamic adjustment processing, and online decomposition execution.

[0056] The obtained energy storage capacity of each site These are planning and configuration parameters that do not participate in subsequent real-time dynamic calculations. Instead, they guide the implementation of engineering plans. For example, they directly guide the site selection, capacity determination, and equipment selection for grid-type energy storage. They serve as the design basis for the rated energy configuration of each energy storage site, determining the hardware capacity boundary of the power station and are one of the core output objectives in the offline planning phase. Site selection and capacity determination refer to planning and setting the rated capacity, i.e., the rated energy, for a specific energy storage unit. The objective function of particle swarm optimization is... Minimization, in essence, means minimizing the equipment investment capacity while satisfying stability constraints.

[0057] Step 209: Real-time acquisition of the actual frequency and real-time SOC of the network-type multi-machine system ports, and calculation of frequency deviation and frequency change rate based on the actual frequency to determine the real-time frequency parameters; Step 210: Combining the reference virtual inertia and the inertia adjustment gain, dynamically adjust the virtual inertia according to the amplitude of the frequency change rate and the sign of the frequency deviation to generate an adaptive virtual inertia. Step 211: Combining the reference frequency modulation coefficient and the damping adjustment gain, dynamically adjust the damping coefficient according to the amplitude of the frequency deviation to generate an adaptive damping coefficient.

[0058] It should be noted that, due to the dummy variables in traditional VSG control... J With damping coefficient D A fixed approach makes it difficult to balance frequency drop suppression and recovery speed. This embodiment addresses this issue by adjusting the frequency deviation. With the rate of change of frequency Real-time frequency parameters affect virtual inertia J With damping coefficient D Dynamic adaptive adjustment is performed, and then the active power control reference value is analyzed based on the adjusted parameters.

[0059] Understandable is the frequency deviation. With the rate of change of frequency Real-time frequency parameters are directly acquired from the system or obtained through basic calculations; while the adaptive adjustment of virtual inertia and damping coefficient are two parallel branches, respectively addressing the dynamic process of frequency change and steady-state deviation, together forming the core of frequency regulation capability and directly determining the dynamic performance of frequency control. Summarizing these adjusted control quantities generates the final executable active power command for the system, which is then distributed downstream and serves as the closed-loop endpoint of a single energy storage unit's control cycle.

[0060] It should be noted that the actual frequency of ports in a networked multi-machine system can be collected in real time. f And real-time SOC; then, based on these parameters, frequency-related parameters are calculated. In this embodiment, the frequency deviation is mainly calculated. and rate of change of frequency Frequency deviation is , It is obtained by differentiating the actual frequency, and can also be used to calculate the rated angular frequency. ,in, This is the system's rated frequency. Real-time SOC is the real-time state of charge of the energy storage system obtained from the battery management system (BMS), and a preset reference state of charge is retrieved. Maximum allowable deviation range This refers to the SOC reference range, which can be used in SOC correction processing tasks.

[0061] The basic parameters for the dynamic adaptive adjustment analysis in this embodiment are the configuration parameters obtained from the offline optimization described above, denoted as the baseline virtual inertia. and reference frequency modulation coefficient Based on this, adaptive virtual inertia can be calculated. With global adaptive damping coefficient :

[0062] in, , These represent the inertia adjustment gain and the damping adjustment gain, respectively, sgn( ) indicates the sign taken from the frequency deviation.

[0063] The adaptive adjustment mechanism of virtual inertia is as follows: if the amplitude of the rate of change of frequency and the sign of the frequency deviation are the same, it indicates that the frequency is continuously changing towards the rated frequency value, and therefore the virtual inertia is increased. This strengthens inertia support and suppresses rapid frequency changes; if the two signs are different, it indicates that the frequency is beginning to recover to its rated value, thus reducing... This reduces inertia resistance and accelerates frequency recovery.

[0064] The adaptive adjustment mechanism of the damping coefficient is that the larger the amplitude of the frequency deviation, the higher the damping coefficient. The larger the value, the higher the active power support strength of the primary frequency regulation, which will quickly pull the frequency back to near the rated value. Therefore, the damping coefficient can be adjusted proportionally to the frequency deviation amplitude.

[0065] also, The adaptive adjustment logic superimposes a dynamic increment positively correlated with the frequency deviation amplitude onto an offline optimized steady-state benchmark. This achieves layered control where the steady-state benchmark is set by planning, and the dynamic intensity is adjusted according to operating conditions. The default base parameters for this real-time adaptive adjustment are given by offline planning, therefore... The parameter assignment relationship.

[0066] Step 212: Introduce a SOC correction factor, perform smooth correction on the real-time SOC based on the SOC reference range, and apply soft limit to the frequency modulation output. Step 213: Perform active power regulation calculations based on adaptive virtual inertia, adaptive damping coefficient, and SOC correction factor to obtain a reference value for total active power.

[0067] In addition, a SOC correction factor can be introduced. Smooth correction of real-time SOC to achieve soft limiting of frequency modulation output:

[0068] Where SOC represents real-time SOC. This represents the reference state of charge, which can be set to 50%. This is the maximum allowable deviation range, i.e., the SOC reference range. When the real-time SOC approaches the upper or lower limit deviation range, The frequency support is reduced to prevent overcharging or over-discharging. Compared with fixed parameters, this correction strategy can provide stronger inertia in the early stage of frequency drop, and the damping is adaptively enhanced during the recovery period. Moreover, the SOC protection is fundamentally different from the traditional simple limit cut-off mechanism.

[0069] Adaptive virtual inertia combined with adaptive dynamic adjustment Adaptive damping coefficient and SOC correction factor It can perform active power regulation and control analysis and calculate the corresponding reference value of the total active power of the system:

[0070] in, This indicates the system's set power, which serves as the active power baseline value.

[0071] Step 214: Calculate the available charge / discharge capacity of each energy storage unit by combining the SOX parameters and rated energy. The SOX parameters include SOC and SOH. Step 215: Calculate the ratio of the available charge / discharge capacity of each energy storage unit to the total available charge / discharge capacity, and determine the allocation weight.

[0072] It should be noted that current SOC protection mechanisms are mostly hard threshold cutoff mechanisms, which are prone to sudden power surges. Furthermore, multi-unit allocation often employs a fixed ratio strategy, failing to comprehensively consider the differences between state of charge and state of health, which can easily lead to some units reaching their limits prematurely and exiting the system, weakening overall support continuity. Therefore, this embodiment fully considers the impact of SOX parameters in its allocation strategy. SOX parameters can include relevant parameters of the battery management system such as SOC, SOH, SOP, and SOE.

[0073] The available charge and discharge capacity calculated by combining SOX parameters generates an allocation weight. Active power and inertia can be allocated according to this weight, so that units with sufficient energy margin can undertake more tasks. This can prevent single units from being overcharged or over-discharged and prematurely exiting the station, thereby improving the station's frequency regulation continuity capability.

[0074] Specifically, the SOX parameters used in this embodiment include SOC and SOH. Based on the available SOC margin and SOH inertia allocation and primary frequency modulation power allocation strategy, frequency support task balancing among multiple units can be achieved. Combining these SOX parameters, the available charge / discharge capacity of each energy storage unit can be calculated:

[0075] in, Obtained directly from BMS. , These are the lower and upper limits of the allowed working range for the SOC. For the first i The health status of each energy storage unit can be obtained based on a battery state estimation algorithm. The rated energy is the energy storage capacity of each site output based on the offline optimization algorithm described above. The plan has been completed.

[0076] The calculation process for assigning weights can be expressed as follows:

[0077] Step 216: Calculate the total primary frequency regulation power requirement for the entire station based on the reference frequency regulation coefficient; Step 217: Calculate the capacity allocation of each energy storage unit based on the allocation weight and the total primary frequency regulation power demand, and determine the active power of the unit; Step 218: Calculate the virtual inertia allocation based on the assigned weights and adaptive virtual inertia to obtain the virtual inertia allocation value; Step 219: Calculate the reference value of active power for each energy storage unit by combining the active power of the unit and the virtual inertia allocation value.

[0078] It should be noted that the total primary frequency regulation power requirement of the system is still obtained from the aggregate droop characteristic:

[0079] in, This is the reference frequency regulation coefficient, obtained through offline optimization using the particle swarm optimization algorithm. The total primary frequency regulation power demand of the system is allocated according to the proportion of available capacity in each unit. The proportion of available capacity is the allocation weight. The allocation process can be expressed as:

[0080] in, Indicates energy storage unit i The allocated frequency regulation active power increment is the active power increment.

[0081] At the same time, the virtual inertia of each energy storage unit is also distributed according to the same weighting to ensure a uniform distribution of power surges during frequency changes:

[0082] in, That is, energy storage unit i The virtual inertia allocation value, The total virtual inertia requirement for the entire site can be obtained through the adaptive virtual inertia obtained after the above adaptive dynamic adjustment. Express.

[0083] And combined with the unit's active power and virtual inertia allocation value Energy storage units can be calculated i Active power reference value:

[0084] Among them, energy storage unit i Damping coefficient This can be expressed as:

[0085] in, As the overall benchmark value for the allocation of damping coefficients among various units, each energy storage unit is decomposed according to the available capacity allocation weight to ensure the overall droop characteristics of the entire station and suppress the rated value of offline optimization. Since not every energy storage unit can achieve active power regulation, in practical applications, if it is inconvenient to adjust the active power of each unit, then the global active power is directly adjusted at the main controller, and then each unit allocates its own power. Currently, energy storage units directly use the global damping coefficient. If the active power of each unit could be precisely adjusted, then targeted and precise allocation could be achieved. In addition, based on the calculation formula for the total primary frequency regulation power demand of the system and the energy storage unit... i Allocated frequency modulation active power increment The calculation formula can be derived as follows:

[0086] Therefore, energy storage units i Active power reference value It includes not only the damping coefficient Energy storage units were also considered. i Allocated frequency modulation active power increment The impact.

[0087] When the state of charge (SOC) of an energy storage unit reaches its limit, its available charge / discharge capacity... When the frequency approaches zero, the unit automatically exits frequency regulation, and the remaining healthy units automatically take on a larger share without requiring additional logic switching. Compared to traditional average or fixed-ratio allocation, the allocation strategy in this embodiment can maximize the utilization of the adjustable capacity of the entire station, avoid local over-discharge, and complement the SOC correction factor mentioned above. While the above limits the total output from the perspective of single-unit safety, this ensures that the tasks undertaken by multiple units match their respective energy margins, achieving hierarchical collaboration from fluctuation suppression to energy balance.

[0088] Step 220: Transmit the active power reference value to the converter to realize adaptive optimization control of the grid-type multi-machine system.

[0089] Active power reference values ​​for each energy storage unit The data can be transmitted to the converter for system optimization control, thus achieving adaptive optimization control of a network-type multi-machine system. To adapt to engineering needs, the parameters obtained at each stage can be recorded separately, and the effectiveness of the optimization control can be verified. Specifically, the basic virtual inertia, basic frequency regulation coefficient, and total primary frequency regulation coefficient obtained based on particle swarm optimization iteration can be stored; the adaptive virtual inertia, adaptive damping coefficient, state of charge correction factor, and total active power reference value of the system obtained through dynamic adaptive adjustment can be stored; and the available charging and discharging capacity, virtual inertia allocation value, allocated frequency regulation power, and active power reference value of each energy storage unit determined during the single-machine allocation execution stage can also be stored.

[0090] Based on these optimized configuration parameters, it can be ensured that the energy storage units with more remaining energy and better lifespan output more power, while units nearing the limit output less power. This avoids individual units reaching the protection threshold first and being shut down in a chain reaction, thereby improving the continuous utilization rate of the adjustable capacity of the entire station and the overall operating life.

[0091] The adaptive optimization control method for a network-type multi-machine system provided in this application derives frequency regulation hard constraints by constructing an equivalent model of the system's frequency response. Then, it combines this with a particle swarm optimization algorithm to comprehensively consider the control effects of various parameters in the system. Based on preset constraints, it generates optimal baseline parameter information. Quantitative tuning replaces empirical configuration, balancing stability margin and investment economy, overcoming reliance on experience, and filling the gap in the coordinated optimization of inertia and capacity. Furthermore, using the optimized baseline parameters as the basis for adjustment, it dynamically and adaptively adjusts core parameters in conjunction with real-time frequency parameters, generating adaptive virtual inertia and damping coefficients. This fully considers the impact of dynamic frequency changes on inertia and dynamically matches the operating conditions. Moreover, it calculates the available charging and discharging capacity of each energy storage unit using SOX parameters and generates allocation weights. Active power and virtual inertia are synchronously allocated according to these weights, replacing manual experience. This allows units with sufficient energy margins to undertake more tasks, preventing premature termination due to overcharging or over-discharging of individual units, and improving the overall station's frequency regulation sustainability. Therefore, the embodiments of this application can solve the technical problem that the prior art relies on fixed inertia, damping parameters, allocation ratios and human experience, and does not carry out targeted collaborative optimization based on core related parameters and constraints, resulting in a lack of stability and reliability in control.

[0092] For easier understanding, please refer to Figure 2 This application provides an embodiment of an adaptive optimization control device for a networked multi-machine system, comprising: The constraint construction unit 201 is used to construct an equivalent model of the system frequency response for a network-type multi-machine system, and to perform constraint analysis on the minimum frequency point based on the total primary frequency modulation coefficient to obtain frequency modulation hard constraints. The parameter optimization unit 202 is used to perform parameter optimization iterative analysis based on preset constraints, taking virtual inertia, total primary frequency regulation coefficient and energy storage capacity of each site as decision variables based on particle swarm optimization algorithm, and generating benchmark virtual inertia and benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints. The dynamic adjustment unit 203 is used to combine real-time frequency parameters and perform dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. The real-time frequency parameters include frequency deviation and frequency change rate. The weight calculation unit 204 is used to calculate the available charge and discharge capacity of each energy storage unit in combination with the SOX parameters, and then calculate the allocation weight based on the available charge and discharge capacity. The allocation calculation unit 205 is used to allocate capacity and virtual inertia for each energy storage unit according to the allocation weight, adaptive virtual inertia and adaptive damping coefficient, and obtain the active power reference value of each energy storage unit. The optimization control unit 206 is used to transmit the active power reference value to the converter to realize the adaptive optimization control of the grid-type multi-machine system.

[0093] Furthermore, constraint building unit 201 is specifically used for: By taking the grid-type energy storage, conventional synchronous power supply and load in the grid-type multi-machine system as a single inertia center, a frequency dynamic time-domain equation based on the rotor swing equation is generated. Ignoring the speed governor's action delay, a Laplace transform is performed on the frequency dynamic time-domain equation to derive the s-domain transfer function of the frequency deviation to the power disturbance, generating an equivalent model of the system's frequency response. After introducing a safety factor, the relationship between the steady-state frequency deviation and the total primary frequency regulation coefficient is analyzed based on the equivalent model of the system frequency response to generate the frequency regulation hard constraint.

[0094] Furthermore, it also includes: a network system constraint building unit, specifically used for: Constructing the system node admittance matrix based on the topology and component parameters of a network-type multi-machine system; Solve the characteristic equations based on the system node admittance matrix to obtain all the characteristic roots of the system; Based on the eigenvalues, calculate the stability margin characterizing the oscillation decay rate and the dominant mode damping ratio characterizing the oscillation suppression capability, respectively. Stability margin constraints and damping ratio constraints are constructed using the stability margin and the dominant mode damping ratio, respectively.

[0095] Furthermore, the dynamic adjustment unit 203 is specifically used for: The actual frequency and real-time SOC of the network-type multi-machine system ports are collected in real time, and the frequency deviation and frequency change rate are calculated based on the actual frequency to determine the real-time frequency parameters. By combining the reference virtual inertia and the inertia adjustment gain, the virtual inertia is dynamically adjusted according to the amplitude of the frequency change rate and the sign of the frequency deviation to generate an adaptive virtual inertia. By combining the reference frequency modulation coefficient and the damping adjustment gain, the damping coefficient is dynamically adjusted according to the amplitude of the frequency deviation to generate an adaptive damping coefficient.

[0096] Furthermore, it also includes: an active power reference value calculation unit, specifically used for: A SOC correction factor is introduced to smoothly correct the real-time SOC based on the SOC reference range, and the frequency modulation output is softly limited. Based on the adaptive virtual inertia, the adaptive damping coefficient, and the SOC correction factor, active power adjustment calculations are performed to obtain a reference value for total active power.

[0097] Furthermore, the weight calculation unit 204 is specifically used for: The available charge / discharge capacity of each energy storage unit is calculated by combining the SOX parameters and the rated energy, wherein the SOX parameters include SOC and SOH; Calculate the ratio of the available charge / discharge capacity of each energy storage unit to the total available charge / discharge capacity, and determine the allocation weight.

[0098] Furthermore, the allocation calculation unit 205 is specifically used for: Calculate the total primary frequency regulation power requirement for the entire station based on the aforementioned reference frequency regulation coefficient; The capacity allocation of each energy storage unit is calculated based on the allocation weight and the total primary frequency regulation power requirement, and the active power of the unit is determined. The virtual inertia allocation is calculated based on the allocation weight and the adaptive virtual inertia to obtain the virtual inertia allocation value; The active power reference value of each energy storage unit is calculated by combining the active power of the unit and the virtual inertia allocation value.

[0099] It should be noted that the frequency response equivalent model in this embodiment ignores the governor delay and is constructed by equating the system's internal grid-type energy storage, conventional synchronous power supply, and load to a single inertia center. Based on the system's frequency response equivalent model, frequency-modulated inertia inference analysis can be performed, specifically the frequency-inertia relationship at the lowest frequency point. Based on this, the system's total primary frequency regulation hard constraint, i.e., frequency regulation hard constraint, can be generated.

[0100] Specifically, the system frequency response equivalent model in this embodiment is a simplified equivalent model that includes grid-type energy storage, residual synchronous power sources, and loads. This model ignores inter-unit oscillations and equates the system inertia to a single inertia center. , which is the system's equivalent inertia constant, can be used to construct the frequency dynamic time-domain equation; therefore, the frequency dynamic time-domain equation in this embodiment is expressed as:

[0101] in, Indicates the system frequency deviation. The actual frequency value collected can also be directly expressed as f , This indicates the system's rated frequency, measured in Hz. The system's equivalent inertia constant is expressed as a constant value in seconds (s), and is obtained by summing the virtual inertia of each station. This represents the virtual inertia of a site. This is the rated angular frequency; The mechanical power increment of a conventional synchronous power supply can be approximated by neglecting the speed controller's operating delay. That is, a conservative estimate is that energy storage will bear the entire power deficit; This represents the disturbance power magnitude, assuming a step-type power deficit, such as the maximum single-unit failure power, in units of pu. The system baseline capacity is... , are system planning parameters; The total primary frequency regulation coefficient is used to directly define the steady-state active power-frequency droop characteristic of the entire station. Its physical meaning is the total amount of additional active power that the grid-type energy storage power station needs to provide for every 1Hz deviation in system frequency.

[0102] right Perform Laplace transform and order The s-domain transfer function of the frequency deviation to the disturbance can be obtained:

[0103] In step disturbance Under the influence of, among them, , These represent the maximum power disturbance step amplitude values, with the deficit value being positive, such as the maximum single-unit capacity. The negative sign "-" indicates the deficit. Performing an inverse Laplace transform, the time-domain frequency deviation can be obtained:

[0104] in, Monotonically decreasing, its absolute value monotonically increases, and its steady-state deviation can be expressed as: The lowest frequency point occurs immediately after the disturbance, and the deviation value at this lowest frequency point is... This can be expressed as:

[0105] Therefore, if the minimum frequency is required to be no lower than Then the following must be satisfied:

[0106] in, To allow the lowest possible frequency, a value of 49.5Hz can be chosen, that is, if , ; This formula is the stable frequency deviation constraint. The lower limit serves as a constraint on the initial total primary frequency regulation coefficient of the system.

[0107] However, in practical systems, the rate of frequency decrease cannot be too large, so inertia constraints need to be constructed. Since the initial total primary frequency regulation coefficient constraint construction process ignores the governor response and assumes the frequency monotonically decreases to the steady-state value, but in actual systems, due to the proportional effect of the governor, the frequency may decrease and then rise during the transient process, with the lowest point occurring 2-3 seconds after the disturbance. Therefore, the s-domain transfer function can incorporate an equivalent model of the first-order governor:

[0108] Therefore, the closed-loop transfer function can be expressed as:

[0109] in, This represents the equivalent time constant of the speed controller, which is obtained based on the factory data or measured data of the synchronous power supply.

[0110] By solving for the extreme points of its time-domain response, a more accurate minimum point constraint can be obtained, but the formula becomes larger and more complex. Therefore, in order to balance optimization efficiency, this embodiment sets:

[0111] Then a safety factor is introduced. Also known as the transient overshoot safety factor, and This allows us to construct frequency modulation hard constraints:

[0112] Among them, the safety factor The value of is determined based on historical simulation experience, and is usually taken as 1.1~1.3; to compensate for the additional overshoot caused by the dynamics of the speed controller; this coefficient can be obtained by Monte Carlo simulation statistics before offline particle swarm optimization operation, and is used as a fixed input parameter into the PSO algorithm.

[0113] Since the minimum frequency is primarily determined by the steady-state gain, provided the inertia constraint is satisfied, the above equation can be embedded as a hard constraint into the particle swarm optimization process: during each iteration's fitness calculation, the following is checked... Check if the frequency modulation hard constraint is met; if not, apply a penalty or invalidate the solution directly.

[0114] It should be noted that the subsequent process of selecting the optimal configuration parameters based on particle swarm optimization needs to satisfy not only the frequency tuning hard constraint of the configuration, but also the verification constraint of the training process, specifically the stability margin constraint and the dominant mode damping ratio constraint. These two constraints are derived step by step based on the system node admittance matrix.

[0115] Specifically, the system node admittance matrix can be constructed first based on the topology and component parameters of the network-type multi-machine system. Then solve the characteristic equation. The kth eigenvalue obtained by solving can be expressed as: Based on these eigenvalues, the stability margin can be calculated respectively. And dominant mode damping ratio :

[0116] Then, stability margin constraints can be constructed based on the obtained stability margin and dominant mode damping ratio. and damping ratio constraint ,in, , These represent the lower limit of the margin and the lower limit of the damping ratio, respectively. It can take the value 0.5. The value range is 0.03 to 0.05. The stability margin constraint ensures that all oscillation modes decay at a sufficiently fast rate after a disturbance; while the damping ratio constraint prevents the system from exhibiting weakly damped or even negatively damped oscillations; these two constraints can also serve as constraints for the subsequent optimization process of configuration parameters, ensuring the stability and reliability of the optimization.

[0117] It should be noted that the decision variables of the particle swarm optimization algorithm in this embodiment include virtual inertia. Total primary frequency modulation coefficient and energy storage capacity of each site In each iteration, the particle carries a set of... The iteration needs to be verified according to the frequency modulation hard constraint. Whether it meets the standard, and then calculate. Next, it is possible to verify whether the iterative process satisfies the stability margin constraint and damping ratio constraint. If all constraints are satisfied, the total energy storage capacity is then calculated. The objective function is to minimize the total cost, otherwise a maximum penalty is imposed.

[0118] Understandably, in the particle swarm optimization algorithm, this check is performed on each particle in each iteration: if any constraint is not satisfied, the particle is given a very high fitness value, i.e., a penalty, causing it to be naturally eliminated in subsequent searches. Only when both constraints are satisfied... and Only particles that meet the requirements are considered feasible solutions and included in the comparison of the objective function. In this way, the final optimization result can guarantee both the frequency and inertia requirements, as well as the dynamic stability of the system; thus obtaining the optimal combination of configuration parameters, including the optimal virtual inertia, the optimal frequency regulation coefficient, and the energy storage capacity of each site.

[0119] Parameter optimization analysis based on particle swarm optimization algorithm is an offline configuration analysis process. The optimal virtual inertia and optimal frequency modulation coefficient obtained are used as the reference virtual inertia and basic frequency modulation coefficient for subsequent dynamic inertia adjustment and online control analysis tasks, realizing a closed-loop scheme of offline planning boundary, dynamic adjustment processing, and online decomposition execution.

[0120] The obtained energy storage capacity of each site These are planning and configuration parameters that do not participate in subsequent real-time dynamic calculations. Instead, they guide the implementation of engineering plans. For example, they directly guide the site selection, capacity determination, and equipment selection for grid-type energy storage. They serve as the design basis for the rated energy configuration of each energy storage site, determining the hardware capacity boundary of the power station and are one of the core output objectives in the offline planning phase. Site selection and capacity determination refer to planning and setting the rated capacity, i.e., the rated energy, for a specific energy storage unit. The objective function of particle swarm optimization is... Minimization, in essence, means minimizing the equipment investment capacity while satisfying stability constraints.

[0121] It should be noted that, due to the dummy variables in traditional VSG control... J With damping coefficient D A fixed approach makes it difficult to balance frequency drop suppression and recovery speed. This embodiment addresses this issue by adjusting the frequency deviation. With frequency change rate Real-time frequency parameters affect virtual inertia J With damping coefficient D Dynamic adaptive adjustment is performed, and then the active power control reference value is analyzed based on the adjusted parameters.

[0122] Understandable is the frequency deviation. With frequency change rate Real-time frequency parameters are directly acquired from the system or obtained through basic calculations; while the adaptive adjustment of virtual inertia and damping coefficient are two parallel branches, respectively addressing the dynamic process of frequency change and steady-state deviation, together forming the core of frequency regulation capability and directly determining the dynamic performance of frequency control. Summarizing these adjusted control quantities generates the final executable active power command for the system, which is then distributed downstream and serves as the closed-loop endpoint of a single energy storage unit's control cycle.

[0123] It should be noted that the actual frequency of ports in a networked multi-machine system can be collected in real time. f And real-time SOC; then, based on these parameters, frequency-related parameters are calculated. In this embodiment, the frequency deviation is mainly calculated. and rate of change of frequency Frequency deviation is , It is obtained by differentiating the actual frequency, and can also be used to calculate the rated angular frequency. ,in, This is the system's rated frequency. Real-time SOC is the real-time state of charge of the energy storage system obtained from the battery management system (BMS), and a preset reference state of charge is retrieved. Maximum allowable deviation range This refers to the SOC reference range, which can be used in SOC correction processing tasks.

[0124] The basic parameters for the dynamic adaptive adjustment analysis in this embodiment are the configuration parameters obtained from the offline optimization described above, denoted as the baseline virtual inertia. and reference frequency modulation coefficient Based on this, adaptive virtual inertia can be calculated. With global adaptive damping coefficient :

[0125] in, , These represent the inertia adjustment gain and the damping adjustment gain, respectively, sgn( ) indicates the sign taken from the frequency deviation.

[0126] The adaptive adjustment mechanism of virtual inertia is as follows: if the amplitude of the rate of change of frequency and the sign of the frequency deviation are the same, it indicates that the frequency is continuously changing towards the rated frequency value, and therefore the virtual inertia is increased. This strengthens inertia support and suppresses rapid frequency changes; if the two signs are different, it indicates that the frequency is beginning to recover to its rated value, thus reducing... This reduces inertia resistance and accelerates frequency recovery.

[0127] The adaptive adjustment mechanism of the damping coefficient is that the larger the amplitude of the frequency deviation, the higher the damping coefficient. The larger the value, the higher the active power support strength of the primary frequency regulation, which will quickly pull the frequency back to near the rated value. Therefore, the damping coefficient can be adjusted proportionally to the frequency deviation amplitude.

[0128] also, The adaptive adjustment logic superimposes a dynamic increment positively correlated with the frequency deviation amplitude onto an offline optimized steady-state benchmark. This achieves layered control where the steady-state benchmark is set by planning, and the dynamic intensity is adjusted according to operating conditions. The default base parameters for this real-time adaptive adjustment are given by offline planning, therefore... The parameter assignment relationship.

[0129] In addition, a SOC correction factor can be introduced. Smooth correction of real-time SOC to achieve soft limiting of frequency modulation output:

[0130] Where SOC represents real-time SOC. This represents the reference state of charge, which can be set to 50%. This is the maximum allowable deviation range, i.e., the SOC reference range. When the real-time SOC approaches the upper or lower limit deviation range, The frequency support is reduced to prevent overcharging or over-discharging. Compared with fixed parameters, this correction strategy can provide stronger inertia in the early stage of frequency drop, and the damping is adaptively enhanced during the recovery period. Moreover, the SOC protection is fundamentally different from the traditional simple limit cut-off mechanism.

[0131] Adaptive virtual inertia combined with adaptive dynamic adjustment Adaptive damping coefficient and SOC correction factor It can perform active power regulation and control analysis and calculate the corresponding reference value of the total active power of the system:

[0132] in, This indicates the system's set power, which serves as the active power baseline value.

[0133] It should be noted that current SOC protection mechanisms are mostly hard threshold cutoff mechanisms, which are prone to sudden power surges. Furthermore, multi-unit allocation often employs a fixed ratio strategy, failing to comprehensively consider the differences between state of charge and state of health, which can easily lead to some units reaching their limits prematurely and exiting the system, weakening overall support continuity. Therefore, this embodiment fully considers the impact of SOX parameters in its allocation strategy. SOX parameters can include relevant parameters of the battery management system such as SOC, SOH, SOP, and SOE.

[0134] The available charge and discharge capacity calculated by combining SOX parameters generates an allocation weight. Active power and inertia can be allocated according to this weight, so that units with sufficient energy margin can undertake more tasks. This can prevent single units from being overcharged or over-discharged and prematurely exiting the station, thereby improving the station's frequency regulation continuity capability.

[0135] Specifically, the SOX parameters used in this embodiment include SOC and SOH. Based on the available SOC margin and SOH inertia allocation and primary frequency modulation power allocation strategy, frequency support task balancing among multiple units can be achieved. Combining these SOX parameters, the available charge / discharge capacity of each energy storage unit can be calculated:

[0136] in, Obtained directly from BMS. , These are the lower and upper limits of the allowed working range for the SOC. For the first i The health status of each energy storage unit can be obtained based on a battery state estimation algorithm. The rated energy is the energy storage capacity of each site output based on the offline optimization algorithm described above. The planning was obtained. The calculation process for assigning weights can be expressed as follows:

[0137] It should be noted that the total primary frequency regulation power requirement of the system is still obtained from the aggregate droop characteristic:

[0138] in, This is the reference frequency regulation coefficient, obtained through offline optimization using the particle swarm optimization algorithm. The total primary frequency regulation power demand of the system is allocated according to the proportion of available capacity in each unit. The proportion of available capacity is the allocation weight. The allocation process can be expressed as:

[0139] in, Indicates energy storage unit i The allocated frequency regulation active power increment is the active power increment.

[0140] At the same time, the virtual inertia of each energy storage unit is also distributed according to the same weighting to ensure a uniform distribution of power surges during frequency changes:

[0141] in, That is, energy storage unit i The virtual inertia allocation value, The total virtual inertia requirement for the entire site can be obtained through the adaptive virtual inertia obtained after the above adaptive dynamic adjustment. Express.

[0142] And combined with the unit's active power and virtual inertia allocation value Energy storage units can be calculated i Active power reference value:

[0143] Among them, energy storage unit i Damping coefficient This can be expressed as:

[0144] in, As the overall benchmark value for the allocation of damping coefficients among various units, each energy storage unit is decomposed according to the available capacity allocation weight to ensure the overall droop characteristics of the entire station and suppress the rated value of offline optimization. Since not every energy storage unit can achieve active power regulation, in practical applications, if it is inconvenient to adjust the active power of each unit, then the global active power is directly adjusted at the main controller, and then each unit allocates it itself. Currently, energy storage units directly use the global damping coefficient. If the active power of each unit could be precisely adjusted, then targeted and precise allocation could be achieved. In addition, based on the system's total primary frequency regulation power demand calculation formula and energy storage unit... i Allocated frequency modulation active power increment The calculation formula can be derived as follows:

[0145] Therefore, energy storage units i Active power reference value It includes not only the damping coefficient Energy storage units were also considered. i Allocated frequency modulation active power increment The impact.

[0146] When the state of charge (SOC) of an energy storage unit reaches its limit, its available charge / discharge capacity... When the frequency approaches zero, the unit automatically exits frequency regulation, and the remaining healthy units automatically take on a larger share without requiring additional logic switching. Compared to traditional average or fixed-ratio allocation, the allocation strategy in this embodiment can maximize the utilization of the adjustable capacity of the entire station, avoid local over-discharge, and complement the SOC correction factor mentioned above. While the above limits the total output from the perspective of single-unit safety, this ensures that the tasks undertaken by multiple units match their respective energy margins, achieving hierarchical collaboration from fluctuation suppression to energy balance.

[0147] Active power reference values ​​for each energy storage unit The data can be transmitted to the converter for system optimization control, thus achieving adaptive optimization control of a network-type multi-machine system. To adapt to engineering needs, the parameters obtained at each stage can be recorded separately, and the effectiveness of the optimization control can be verified. Specifically, the basic virtual inertia, basic frequency regulation coefficient, and total primary frequency regulation coefficient obtained based on particle swarm optimization iteration can be stored; the adaptive virtual inertia, adaptive damping coefficient, state of charge correction factor, and total active power reference value of the system obtained through dynamic adaptive adjustment can be stored; and the available charging and discharging capacity, virtual inertia allocation value, allocated frequency regulation power, and active power reference value of each energy storage unit determined during the single-machine allocation execution stage can also be stored.

[0148] Based on these optimized configuration parameters, it can be ensured that the energy storage units with more remaining energy and better lifespan output more power, while units nearing the limit output less power. This avoids individual units reaching the protection threshold first and being shut down in a chain reaction, thereby improving the continuous utilization rate of the adjustable capacity of the entire station and the overall operating life.

[0149] This application also provides an adaptive optimization control device for a networked multi-machine system, the device including a processor and a memory; The memory is used to store program code and transfer the program code to the processor; The processor is used to execute the adaptive optimization control method for the networked multi-machine system in the above method embodiments according to the instructions in the program code.

[0150] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0152] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0153] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the various embodiments of this application through a computer device (which may be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0154] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An adaptive optimization control method for a networked multi-machine system, characterized in that, include: An equivalent model of the system frequency response is constructed for a network-type multi-machine system, and a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient is performed to obtain the frequency modulation hard constraint. Based on the particle swarm optimization algorithm, the virtual inertia, the total primary frequency regulation coefficient, and the energy storage capacity of each site are used as decision variables. The parameter optimization iterative analysis is performed based on preset constraints to generate the benchmark virtual inertia and benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints. By combining real-time frequency parameters, dynamic adaptive adjustment analysis is performed based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. The real-time frequency parameters include frequency deviation and frequency change rate. After calculating the available charge / discharge capacity of each energy storage unit based on the SOX parameters, the allocation weight is calculated based on the available charge / discharge capacity. Based on the allocation weight, the adaptive virtual inertia, and the adaptive damping coefficient, capacity allocation and virtual inertia allocation are performed for each energy storage unit to obtain the active power reference value of each energy storage unit. The active power reference value is transmitted to the converter to realize adaptive optimization control of the grid-type multi-machine system.

2. The adaptive optimization control method for a networked multi-machine system according to claim 1, characterized in that, The above describes the construction of an equivalent model of the system frequency response for a network-type multi-machine system, and the performance of a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain the frequency modulation hard constraints, including: By taking the grid-type energy storage, conventional synchronous power supply and load in the grid-type multi-machine system as a single inertia center, a frequency dynamic time-domain equation based on the rotor swing equation is generated. Ignoring the speed governor's action delay, a Laplace transform is performed on the frequency dynamic time-domain equation to derive the s-domain transfer function of the frequency deviation to the power disturbance, generating an equivalent model of the system frequency response. After introducing a safety factor, the relationship between the steady-state frequency deviation and the total primary frequency regulation coefficient is analyzed based on the equivalent model of the system frequency response to generate the frequency regulation hard constraint.

3. The adaptive optimization control method for a networked multi-machine system according to claim 1, characterized in that, The process involves constructing an equivalent model of the system frequency response for a network-type multi-machine system, performing a frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain the frequency modulation hard constraint, and then further including: Constructing the system node admittance matrix based on the topology and component parameters of a network-type multi-machine system; Solve the characteristic equations based on the system node admittance matrix to obtain all the characteristic roots of the system; Based on the eigenvalues, calculate the stability margin characterizing the oscillation decay rate and the dominant mode damping ratio characterizing the oscillation suppression capability, respectively. Stability margin constraints and damping ratio constraints are constructed using the stability margin and the dominant mode damping ratio, respectively.

4. The adaptive optimization control method for a networked multi-machine system according to claim 1, characterized in that, The process involves combining real-time frequency parameters and performing dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient, including: The actual frequency and real-time SOC of the network-type multi-machine system ports are collected in real time, and the frequency deviation and frequency change rate are calculated based on the actual frequency to determine the real-time frequency parameters. By combining the reference virtual inertia and the inertia adjustment gain, the virtual inertia is dynamically adjusted according to the amplitude of the frequency change rate and the sign of the frequency deviation to generate an adaptive virtual inertia. By combining the reference frequency modulation coefficient and the damping adjustment gain, the damping coefficient is dynamically adjusted according to the amplitude of the frequency deviation to generate an adaptive damping coefficient.

5. The adaptive optimization control method for a networked multi-machine system according to claim 4, characterized in that, The process involves combining real-time frequency parameters and performing dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. This process further includes: A SOC correction factor is introduced to smoothly correct the real-time SOC based on the SOC reference range, and the frequency modulation output is softly limited. Based on the adaptive virtual inertia, the adaptive damping coefficient, and the SOC correction factor, active power adjustment calculations are performed to obtain a reference value for total active power.

6. The adaptive optimization control method for a networked multi-machine system according to claim 1, characterized in that, After calculating the available charge / discharge capacity of each energy storage unit using SOX parameters, the allocation weight is calculated based on the available charge / discharge capacity, including: The available charge / discharge capacity of each energy storage unit is calculated by combining the SOX parameters and the rated energy, wherein the SOX parameters include SOC and SOH; Calculate the ratio of the available charge / discharge capacity of each energy storage unit to the total available charge / discharge capacity, and determine the allocation weight.

7. The adaptive optimization control method for a networked multi-machine system according to claim 1, characterized in that, The process of allocating capacity and virtual inertia for each energy storage unit based on the allocation weight, the adaptive virtual inertia, and the adaptive damping coefficient, to obtain the active power reference value for each energy storage unit, includes: Calculate the total primary frequency regulation power requirement for the entire station based on the aforementioned reference frequency regulation coefficient; The capacity allocation of each energy storage unit is calculated based on the allocation weight and the total primary frequency regulation power requirement, and the active power of the unit is determined. The virtual inertia allocation is calculated based on the allocation weight and the adaptive virtual inertia to obtain the virtual inertia allocation value; The active power reference value of each energy storage unit is calculated by combining the active power of the unit and the virtual inertia allocation value.

8. An adaptive optimization control device for a network-type multi-machine system, characterized in that, include: The constraint construction unit is used to build an equivalent model of the system frequency response for a network-type multi-machine system, and to perform frequency minimum point constraint analysis based on the total primary frequency modulation coefficient to obtain frequency modulation hard constraints. The parameter optimization unit is used to perform parameter optimization iterative analysis based on preset constraints, taking virtual inertia, total primary frequency regulation coefficient and energy storage capacity of each site as decision variables based on particle swarm optimization algorithm, and generating benchmark virtual inertia and benchmark frequency regulation coefficient. The preset constraints include frequency regulation hard constraints. The dynamic adjustment unit is used to combine real-time frequency parameters and perform dynamic adaptive adjustment analysis based on the reference virtual inertia and the reference frequency modulation coefficient to generate adaptive virtual inertia and adaptive damping coefficient. The real-time frequency parameters include frequency deviation and frequency change rate. The weight calculation unit is used to calculate the available charge / discharge capacity of each energy storage unit in combination with the SOX parameters, and then calculate the allocation weight based on the available charge / discharge capacity. The allocation calculation unit is used to allocate capacity and virtual inertia for each energy storage unit according to the allocation weight, the adaptive virtual inertia and the adaptive damping coefficient, and to obtain the active power reference value of each energy storage unit. An optimization control unit is used to transmit the active power reference value to the converter to realize adaptive optimization control of the grid-type multi-machine system.

9. The adaptive optimization control device for a network-type multi-machine system according to claim 8, characterized in that, The constraint construction unit is specifically used for: By taking the grid-type energy storage, conventional synchronous power supply and load in the grid-type multi-machine system as a single inertia center, a frequency dynamic time-domain equation based on the rotor swing equation is generated. Ignoring the speed governor's action delay, a Laplace transform is performed on the frequency dynamic time-domain equation to derive the s-domain transfer function of the frequency deviation to the power disturbance, generating an equivalent model of the system frequency response. After introducing a safety factor, the relationship between the steady-state frequency deviation and the total primary frequency regulation coefficient is analyzed based on the equivalent model of the system frequency response to generate the frequency regulation hard constraint.

10. An adaptive optimization control device for a network-type multi-machine system, characterized in that, The device includes a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the adaptive optimization control method for a networked multi-machine system according to any one of the instructions in the program code.

Images