Source network load storage dynamic balancing system and method for high-proportion new energy power grid

Abstract

The invention relates to the field of power systems, and particularly discloses a source network load storage dynamic balancing system and method for a high-proportion new energy power grid. The basic principle of the scheme of the invention is as follows: a regional collaborative optimization layer performs rolling optimization by using a distributed model predictive control algorithm based on wide-area measurement and ultra-short-term predictive data to generate a regional collaborative control target in the form of voltage or reactive power demand; and after receiving the target, each intelligent power converter in the distributed equipment execution layer performs real-time negotiation with adjacent equipment by running a distributed consistency algorithm, autonomously determines respective virtual admittance regulation quantity, and finally quickly adjusts the equivalent admittance of the grid-connected point through a virtual admittance control technology, and outputs the required reactive power. The most core technical effect of the scheme is that the second-level real-time dynamic balancing capability and the safe and stable operation level of the power grid in the high-proportion random new energy access environment are improved.

Description

Technical Field

[0001] This invention relates to the field of electrical energy, and more specifically, to a dynamic balancing system and method for a high-proportion renewable energy power grid, encompassing source, grid, load, and storage. Background Technology

[0002] With the rapid increase in the penetration rate of intermittent power sources such as wind and solar power, the physical characteristics of the power system have undergone fundamental changes, breaking the traditional stable pattern dominated by synchronous generators. The strong randomness, volatility, and weak support of new energy output have made it exponentially more difficult to maintain real-time power balance and voltage stability of the power grid. The problem of absorbing a high proportion of new energy has evolved from a simple matter of electricity intake to a dynamic equilibrium control problem involving multiple time scales from milliseconds to minutes.

[0003] To address this challenge, existing technologies are mainly developing in two directions: First, enhancing the intelligence and speed of centralized control, such as improving AGC algorithms and introducing ultra-short-term power prediction. However, this is essentially a centralized computing and terminal execution model, which suffers from inherent bottlenecks such as response lag, high computational complexity, and strong communication dependence when facing massive, dispersed distributed resources. Second, developing fully distributed local control, such as autonomous frequency and voltage regulation based on droop characteristics. While this method offers rapid response, it lacks global coordination, easily leading to regulation conflicts between devices, deviations from optimal system operating points, and an inability to solve systemic problems such as regional voltage coordination and stability. These two technological paths are fragmented in their control architecture, making it difficult to simultaneously meet the stringent requirements of global optimization and rapid coordination in high-proportion scenarios.

[0004] Therefore, a prominent problem in the current technological field is that the existing control system architecture cannot effectively coordinate the optimization decision-making capabilities of the dispatching side with the rapid response capabilities of the massive number of devices, resulting in insufficient real-time dynamic balancing capabilities of the power grid to cope with the second-level fluctuations of high proportions of new energy sources. There is an urgent need for a new solution that can connect optimization and execution, and integrate centralized guidance and distributed autonomy, to systematically improve the resilience and flexibility of the power grid. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time dynamic balancing solution for a high-proportion renewable energy power grid that integrates optimization and execution, and combines centralized guidance with distributed autonomy.

[0006] According to a first aspect of the present invention, a dynamic balancing system for a high-proportion renewable energy power grid, comprising a regional collaborative optimization layer and a distributed equipment execution layer, is proposed. The regional collaborative optimization layer is used to generate regional-level collaborative control objectives based on wide-area measurement and forecast data from the power grid through distributed optimization algorithms. The distributed device execution layer includes multiple smart power converters deployed in the power grid; the smart power converters are configured as follows: Receive regional-level coordinated control objectives; Based on the collaborative control objective and local measurement data, distributed negotiation is conducted through adjacent intelligent power converters that are connected to it to determine their respective regulation commands. Adjust the equivalent admittance of its grid connection point according to the regulation command to output the corresponding reactive power and achieve the regional coordinated control target.

[0007] According to some embodiments, in the system of the first aspect of the present invention, the regional collaborative optimization layer includes at least one regional controller; Distributed optimization algorithms include distributed model predictive control algorithms. The area controller is configured to execute distributed model predictive control algorithms, specifically including: Based on wide-area measurement data and ultra-short-term renewable energy power forecast data, rolling estimates are made of the state of the power grid in future periods. With minimizing the regional net load fluctuation as the optimization objective, the optimal control sequence within a future finite time window is solved; The optimal control sequence is converted into a regional-level coordinated control objective, which includes at least the voltage reference value of the critical bus or the total reactive power demand of the region.

[0008] According to some embodiments, in the system of the first aspect of the present invention, when the regional controller performs rolling estimation, it specifically uses the Kalman filter algorithm to perform interval estimation of the system state including the output of new energy sources.

[0009] According to some embodiments, in the system of the first aspect of the present invention, the intelligent power converter includes a virtual admittance control module and a consistency coordination module; The consensus and coordination module executes a distributed consensus algorithm, exchanges local voltage information and collaborative control objectives with neighboring smart power converters, and calculates the local consensus mechanism through iterative calculations. H02J is a local virtual admittance setpoint used by power transmission, conversion, distribution, and control equipment to bring the voltage closer to the target value, serving as a regulation command. The virtual admittance control module is used to adjust the modulation signal of the inverter bridge of the converter in real time according to the local virtual admittance setting value, so as to realize dynamic control of the equivalent admittance at the grid connection point.

[0010] According to some embodiments, in the system of the first aspect of the present invention, the distributed consensus algorithm is an average consensus algorithm or a proportional-integral consensus algorithm, so as to realize the proportional allocation of reactive power output of each intelligent power converter according to capacity or adjustable margin.

[0011] According to some embodiments, in the system of the first aspect of the present invention, the system further includes a wide-area synchronous measurement unit network; The wide-area synchronous measurement unit network consists of synchronous phasor measurement units deployed at key nodes of the power grid. It is used to collect and upload wide-area measurement data containing voltage and current phasors to the regional collaborative optimization layer at a rate higher than the power frequency.

[0012] According to some embodiments, in the system of the first aspect of the present invention, the regional-level collaborative control target is sent to each intelligent power converter through the manufacturing message specification service based on the IEC61850 standard or the general object-oriented substation event message.

[0013] According to some embodiments, in the system of the first aspect of the present invention, the intelligent power converter is any one of a photovoltaic inverter, an energy storage converter, or a wind power converter.

[0014] According to a second aspect of the present invention, a dynamic balancing method for a high-proportion renewable energy power grid is proposed, characterized in that it is applied to a system as described in the first aspect of the present invention, and the method includes: S1. The regional collaborative optimization layer generates regional collaborative control objectives through distributed optimization based on the wide-area measurement and forecast data of the power grid. S2. Each intelligent power converter in the distributed device execution layer receives the cooperative control target; S3. Based on the collaborative control objective and local measurement data, each intelligent power converter autonomously determines its own virtual admittance adjustment through distributed negotiation with adjacent devices; S4. Each intelligent power converter adjusts its output according to the virtual admittance regulation to achieve the collaborative control objective and complete the real-time dynamic balance of the power grid.

[0015] According to some embodiments, in the method of the second aspect of the present invention, step S1 specifically includes: using a distributed model predictive control algorithm to perform rolling optimization of the power grid state within a future time window, and mapping the optimization result to voltage or reactive power coordination target.

[0016] The present invention has the following beneficial effects: 1. To address the contradiction between the response delay of centralized control and the lack of global optimization guidance in distributed control, this invention constructs a two-layer control architecture that combines a regional collaborative optimization layer and a distributed device execution layer, thereby achieving an efficient unification of system-level optimization decision-making and device-level rapid response capabilities.

[0017] 2. To address the issues of complex models and high computational pressure in traditional centralized optimization algorithms when dealing with a large number of devices, this invention adopts a decoupled workflow of "macro-level collaborative goal distribution and device-level autonomous distributed negotiation," which significantly improves the system's scalability and operational robustness while ensuring optimization orientation.

[0018] 3. To address the problem of voltage instability in the distribution network caused by a high proportion of distributed power sources and the slow response of traditional voltage regulation methods, this invention deeply integrates virtual admittance control technology with distributed consensus algorithm, enabling reactive power resources to achieve millisecond-level, decentralized collaborative regulation, thereby quickly suppressing local voltage fluctuations.

[0019] 4. Addressing the challenge of effectively integrating advanced control algorithms with actual power physics systems in engineering implementation, this invention provides a clear and feasible engineering implementation path for innovative control theory by explicitly defining wide-area synchronous measurement units, intelligent converters supporting virtual admittance, and regional controllers as hardware carriers, and defining standardized data interaction processes. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without exceeding the scope of protection claimed by the present invention.

[0021] Figure 1 This is a schematic diagram of an embodiment 1000 of a dynamic balancing system for a high-proportion renewable energy power grid according to the present invention. Figure 2 for Figure 1 A schematic diagram of the structure of the intelligent power converter a in Example 1000; Figure 3 This is a schematic diagram of an embodiment 2000 of a dynamic balancing system for a high-proportion renewable energy power grid based on the present invention. Figure 4 This is a flowchart illustrating an embodiment 3000 of a dynamic balancing method for a high-proportion renewable energy power grid based on the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] See Figure 1 , Figure 1 This is a schematic diagram of an embodiment 1000 of a dynamic balancing system for a high-proportion renewable energy power grid based on the present invention. Figure 1 As shown, Embodiment 1000 includes a regional collaborative optimization layer 101 and a distributed device execution layer 102.

[0024] Optionally, the regional collaborative optimization layer 101 is used to generate regional-level collaborative control objectives based on wide-area measurement and forecast data from the power grid through distributed optimization algorithms. The regional collaborative optimization layer 101 acts as the intelligent decision-making center of the system. Its main function is to integrate global information, perform forward-looking rolling optimization calculations, and transform complex system-level stability objectives into collaborative control objectives that can be understood and executed by a massive number of distributed devices. The regional collaborative optimization layer 101 solves the problem of how to scientifically set control objectives to optimally mitigate future short-term fluctuations.

[0025] The regional collaborative optimization layer 101 includes at least one regional controller. Optionally, the regional controller is responsible for a geographically or electrically independent sub-region, such as a new energy aggregation area or a power distribution network, and its decisions are based on wide-area measurement data and ultra-short-term forecast data of that region.

[0026] Optionally, the distributed optimization algorithm includes a distributed model predictive control algorithm, and the area controller is configured to execute the distributed model predictive control algorithm, specifically including: Step S1. Based on wide-area measurement data and ultra-short-term renewable energy power forecast data, perform rolling estimation of the grid state for future periods. Optionally, step S1 aims to obtain the optimal estimate of the current and future short-term grid state based on incomplete measurement data and predictions with errors. Optionally, when performing rolling estimation, the regional controller specifically uses the Kalman filter algorithm to perform interval estimation of the system state including renewable energy output.

[0027] Optionally, the wide-area measurement data in step S1 includes: node voltage amplitude. (Unit: kV), Voltage phase angle (Unit: radians), and branch active power reactive power (Unit: MW / Mvar); Ultra-short-term renewable energy power forecast data includes: the predicted active power of all renewable energy power plants within the next N scheduling periods (e.g., the next 15 minutes, with Δt=1 minute intervals). (Unit: MW)

[0028] In some specific embodiments, in step S1, a Kalman filter algorithm is used for dynamic state estimation. Specifically: State-space model definition: Define the system state vector. This includes the voltage magnitude and phase angle of all nodes. The algorithm runs based on the following two equations: State prediction equation: Where A is the state transition matrix (usually taken from the previous time step of the grid linearization model), and B is the control input matrix (describing the effect of predicted power on the state). This represents process noise (covariance matrix Q).

[0029] Measurement update equation: = H * + Where H is the measurement Jacobian matrix, reflecting the nonlinear relationship between the state and the measurement, obtained by linearizing the current state estimate. The covariance matrix is ​​R, representing the measurement noise.

[0030] Optionally, the iterative process of the Kalman filter algorithm in step S1: a. Prediction: Based on the optimal estimate from the previous time step. and current forecasts The prior state estimate at the current moment is calculated using the state prediction equation. and its error covariance .

[0031] b. Update: Obtain actual measurements Then, calculate the Kalman gain.

[0032] Subsequently, the prior estimate is corrected using the residuals of the actual and predicted measurements to obtain the posterior optimal state estimate at the current moment: And update the error covariance. .

[0033] The output of step S1 is the optimal estimate of the system state at the current time k. That is, the voltage amplitude and phase angle, and the quantitative index of their estimation uncertainty—the error covariance matrix. . The diagonal elements represent the variance of the estimated values ​​of each state variable.

[0034] In some specific embodiments, the optimal estimation characteristics of the Kalman filter algorithm can effectively fuse noisy measured data with the prediction model, and provide key quantitative uncertainty information, namely error covariance, for subsequent robust optimization. In scenarios where the power grid model is approximately linear and the noise satisfies the Gaussian assumption, this is the optimal choice for both computational efficiency and estimation accuracy.

[0035] Optionally, in some specific embodiments, if the system is significantly nonlinear, such as containing a large number of power electronic devices, an unscented Kalman filter can be selected. The input remains the same, but the processing method is changed to selecting a "Sigma point" and passing it through the nonlinear system equations to more accurately calculate the mean and covariance of the state estimate, at the cost of increased computational complexity.

[0036] In some specific embodiments, in step S2, the system aims to minimize regional net load fluctuations and solves for the optimal control sequence within a future finite time window. Step S2, based on the latest state estimate, solves for the optimal control sequence over a future period to minimize system fluctuations.

[0037] Optionally, the input data for step S2 includes: Current system state optimal estimation and its covariance ; Ultra-short-term forecast data for the next M control windows (e.g., the next 5 minutes, with a control period of Δt = 15 seconds). and critical load forecasting ; Equipment and network constraints, including: upper and lower limits of generator / energy storage active power output. upper and lower limits of reactive power output Line transmission capacity limit Node voltage safety range .

[0038] Specifically, in step S2, the system constructs and solves a distributed robust optimization problem: Optimization problem modeling: Objective function: Minimize the fluctuation of net load in the region over the next M time periods. Specifically: ;in, , (t) represents the active power output decision variable for controllable resources (such as traditional generating units and energy storage).

[0039] Constraints: Equipment physical constraints: , .

[0040] Power grid security constraint: Line power calculated based on AC or DC power flow equations ≤ ; .

[0041] Uncertainty constraints (robust handling): Estimating the covariance of state estimates Information is transformed into a set of uncertainties in the output of new energy sources, requiring that all decisions be feasible within ±3 standard deviations of the predicted output.

[0042] Distributed solution: The alternating direction multiplier method is used. Each region controller decomposes the global problem geographically. In each iteration step i: a. Local Solution: Each controller solves its own sub-optimization problem in parallel. Its objective function includes local cost and a penalty term related to the consistency of boundary variables in neighboring regions. .

[0043] b. Information exchange: Each controller sends the calculated boundary variables (such as tie line planned power and boundary bus voltage) to the adjacent controller.

[0044] c. Global variable update and coordination: After collecting neighbor information, update the global average estimate of the boundary variables. And update the Lagrange multiplier vectors. .

[0045] d. Iteration Termination: When the difference norm of the boundary variables among the controllers is less than a preset threshold. And the change in the objective function is less than When the iteration stops, a global consensus on the optimal outcome is reached.

[0046] Finally, step S2 yields the globally optimal control sequence for the next M time periods, which represents the active / reactive power setpoint plans for all controllable resources in each time period: ; Optionally, in step S3, the system converts the optimal control sequence into a regional-level coordinated control objective, which includes at least the voltage reference value of the critical bus or the total reactive power demand of the region. Step S3 converts the detailed resource-level control plan into high-level, stable instructions suitable for the equipment execution layer.

[0047] Optionally, step S3 is based on the optimal control sequence. Especially the part about reactive power regulation (t), and the current power grid topology and parameters, are used for target mapping using voltage-reactive power sensitivity analysis, specifically including: Sensitivity matrix calculation: based on the current estimated state Based on the network admittance matrix, calculate the sensitivity coefficients of the critical voltage control bus j to each reactive power injection node i. This forms the sensitivity matrix S.

[0048] Target value calculation: If a voltage reference target is generated, then: , in, This is the planned value for the first optimization period. This is the current actual value. This calculation reverses the reactive power plan to the desired voltage change.

[0049] If the target for total reactive power demand in the generated region is to be calculated, then the summation can be performed directly: Target encapsulation: Encapsulating the calculated set of... or The value, along with the target's effective time stamp and geographic range identifier, is encapsulated into a standard format data frame.

[0050] Optionally, in step S3, a region-level collaborative control target data packet is output. This target is an instructive system operating state target, rather than a specific action command for each inverter, thus providing clear optimization guidance and necessary degrees of freedom for the autonomous collaboration of the distributed device execution layer 102.

[0051] The distributed device execution layer 102 includes multiple smart power converters a deployed in the power grid. Optionally, the smart power converters a are configured to: receive a regional-level coordinated control objective; based on the coordinated control objective and local measurement data, conduct distributed negotiation with adjacent smart power converters a through which they are communicatively connected to determine their respective regulation commands; adjust the equivalent admittance of their grid connection point according to the regulation commands to output corresponding reactive power, thereby collaboratively achieving the regional-level coordinated control objective. Optionally, the smart power converters a can be any one of photovoltaic inverters, energy storage converters, or wind power converters.

[0052] According to such Figure 1 The implementation method shown in this invention creatively integrates centralized global forward-looking optimization with distributed local rapid collaborative execution by constructing a two-layer architecture that combines a regional collaborative optimization layer and a distributed device execution layer. This enables the system to generate optimal collaborative control targets based on wide-area information and drive a large number of distributed intelligent power converters to autonomously negotiate and precisely adjust equivalent admittance through proximity communication. This achieves millisecond-level self-organized collaborative suppression of active / reactive power impacts from random fluctuations in high-proportion renewable energy sources. While ensuring the optimal operation of the power grid as a whole, this fundamentally improves the real-time dynamic balance capability and safety and stability level of the power system.

[0053] Figure 2 for Figure 1 A schematic diagram of the structure of the intelligent power converter a in Example 1000. (See attached diagram.) Figure 2 As shown, the intelligent power converter a includes a consistency coordination module a1 and a virtual admittance control module a2.

[0054] Optionally, Figure 2 The intelligent power converter 'a' shown is a distributed execution unit deployed in the power grid. Its core function is to receive the collaborative control objectives sent down from the upper layer and convert them into reactive power output that supports the stability of the power grid through autonomous negotiation and rapid control with adjacent devices.

[0055] Specifically, the consensus module a1 executes a distributed consensus algorithm, exchanges local voltage information and collaborative control objectives with adjacent smart power converters, and calculates a local virtual admittance setpoint that makes the local voltage approach the target value through iterative calculation, which serves as the adjustment command. Optionally, the distributed consensus algorithm is an average consensus algorithm or a proportional-integral consensus algorithm to achieve proportional allocation of reactive power output of each smart power converter according to capacity or adjustable margin.

[0056] Optionally, the consensus coordination module a1 supports multiple consensus algorithms. The following description uses two typical examples: the average consensus algorithm and the proportional-integral consensus algorithm. The two examples share the same data input and output interfaces, but their internal processing logic differs.

[0057] Optionally, the input data for the consistency coordination module a1 includes: The cooperative control target T is derived from the regional cooperative optimization layer 101 and is in the form of a voltage reference value. .

[0058] Local Measurement Local grid connection point voltage amplitude .

[0059] Neighbor Collaborative Information Set State data periodically received from a communicating neighbor j (j∈neighbor set Ni), including its voltage estimate. With virtual admittance setting (or equivalent susceptance) ).

[0060] Output: Local virtual admittance setting : is a complex number, and its effective part in reactive power regulation scenarios is the imaginary part, i.e., susceptance. (Unit: Siemens S), therefore ≈ j * .

[0061] In some specific embodiments, the consensus coordination module a1 is implemented based on the average consensus algorithm, aiming to quickly bring the system state estimates of all converters within the cluster to a consensus and to approximate the global voltage target in a distributed manner. The specific process includes: (1) State initialization: Define the local state vector as At the start of each control cycle (e.g., Tc = 100ms): [0] = / / Initial voltage estimation based on local measurements [0] = / / Initialize the final output admittance value of the previous cycle.

[0062] (2) Iterative calculation process: The module undergoes K iterations (e.g., K=5~10). The steps for the k-th iteration (k=1, 2, ..., K) are as follows: Information exchange: Send the local state Si[k-1] to all neighbors j∈Ni, and at the same time receive the neighbor state Sj[k-1].

[0063] b. Consistency Update: Perform average consistency calculations on the voltage estimate and admittance value respectively.

[0064]

[0065] in, These are weighting coefficients, which can typically be selected. = 1 / (1 + max( , )), The number of neighbors is used to ensure algorithm convergence.

[0066] c. Target Injection: After the consensus update, a global target is injected into the voltage estimation state. The deviation is used to guide the entire cluster state to converge to the target value.

[0067] ; α is a small positive gain (e.g., α=0.1) used to control the speed of target fusion.

[0068] (3) Output generation: After the iteration is complete, the final calculated susceptance value is taken as the output for this cycle:

[0069] In some specific embodiments, the consensus coordination module a1 is implemented based on the proportional-integral consensus algorithm. It introduces an integral term on top of average consensus, aiming to distribute reactive power proportionally according to adjustable margins and eliminate steady-state errors, thus achieving more precise and fair coordination. The specific process includes: (1) Definition of state and parameters: Define the local state vector as , where Ii is the integral term of the voltage deviation. Definition This is the maximum reactive power that the machine can currently output (calculated in real time based on DC side voltage, current limits, etc.).

[0070] (2) Iterative calculation process: The steps for the k-th iteration are as follows: a. Information exchange: Exchange extended state vectors with neighbors. .

[0071] b. Proportional-integral consistency update: The state update law is designed to simultaneously achieve neighbor state consistency and track the global objective.

[0072] ; ; , γ is the designed convergence coefficient.

[0073] c. Admittance calculation based on margin: The admittance setpoint is calculated using the integral term and local margin information to ensure that the output is proportional to the capacity.

[0074] ; in, For proportional and integral coefficients; To achieve the allocation of core items according to the adjustable margin ratio (the total cluster value can be obtained synchronously through a consensus algorithm).

[0075] (3) Output generation: After the iteration is complete, the final value is also taken as the output:

[0076] The two embodiments of this invention above provide implementation paths for the consistency and coordination module a1. The embodiment based on average consistency focuses on rapid state consistency and target tracking, and has a simple structure; the embodiment based on proportional-integral consistency introduces a dynamic proportional allocation mechanism and error-free adjustment capability, achieving a more refined allocation of collaborative benefits, and is suitable for scenarios with higher requirements for fairness and accuracy. The output of the consistency and coordination module a1. It will be transmitted to the virtual admittance control module a2 for execution.

[0077] The virtual admittance control module a2 is used to adjust the modulation signal of the inverter bridge of the converter in real time according to the local virtual admittance setting value, so as to realize the dynamic control of the equivalent admittance at the grid connection point.

[0078] Optionally, the input data for the virtual admittance control module a2 includes: Admittance setting instructions : Output from consensus coordination module a1; Instantaneous voltage at grid connection point (t): Real-time three-phase voltage value obtained by high-speed ADC sampling.

[0079] The virtual admittance control module a2 performs the following calculations and controls within each pulse width modulation cycle (e.g., Ts = 50μs): Step a: Calculate the reference current. Calculate the instantaneous value of the desired admittance current according to Ohm's law. ; like = j* This operation is equivalent to shifting the voltage signal by -90 degrees and then multiplying it by... It directly generates pure reactive current commands.

[0080] Step b: Current closed-loop tracking control. Accurate tracking is achieved in the dq rotating coordinate system. The control law is: in, The reference value for the converter bridge arm voltage to be generated; , For proportional and integral coefficients; for Command to transform to the dq coordinate system; ω is the measured current; L is the mains angular frequency; and L is the filter inductance value. This controller calculates the output voltage required to eliminate tracking errors.

[0081] Step c: Generate the modulated signal. The calculated... The voltage reference wave is converted into a three-phase stationary coordinate system through inverse Park transformation. The space vector pulse width modulation algorithm is then used to generate PWM signals to drive the power semiconductor devices (such as IGBTs) to turn on and off.

[0082] The virtual admittance control module a2 in this invention controls the current through the converter according to... The command causes it to output current on its AC side. Real-time tracking (t). From the perspective of the grid port, this converter is equivalent to a value of The linear time-varying admittance. When At 0, it exhibits sensory activity and absorbs no energy. When the voltage is less than 0, it exhibits capacitive behavior and generates reactive power. (This applies to all grid converters.) The set of components is dynamically adjusted through the consistency and coordination module a1 to jointly achieve rapid, distributed support for grid voltage.

[0083] Figure 3 This is a schematic diagram of an embodiment 2000 of a dynamic balancing system for a high-proportion renewable energy power grid based on the present invention. Figure 3As shown, embodiment 2000 includes a regional collaborative optimization layer 201, a distributed device execution layer 202, and a wide-area synchronous measurement unit network 203. The regional collaborative optimization layer 201, the distributed device execution layer 202, and... Figure 1 The regional collaborative optimization layer 101 and the distributed device execution layer 102 of the embodiment 1000 are the same, and will not be described again here.

[0084] Optionally, in embodiment 2000, the wide-area synchronous measurement unit network 203 consists of multiple synchronous phasor measurement units b deployed at key nodes of the power grid, used to collect and upload wide-area measurement data containing voltage and current phasors to the regional collaborative optimization layer at a rate higher than the power frequency.

[0085] In some specific embodiments, the wide-area synchronous measurement unit network consists of multiple synchronous phasor measurement units (hereinafter referred to as PMU devices) deployed at key nodes of the power grid interconnected through a communication network.

[0086] Deployment nodes include, but are not limited to: the beginning and end of main transmission lines, high / medium voltage busbars of important power plants and substations, outlets of major renewable energy collection stations, endpoints of regional power grid interconnection lines, and important load centers. The selection of these nodes ensures that the main power flow, voltage support points, and potential stability weaknesses of the entire network can be captured.

[0087] Network topology: PMU devices are typically connected via dedicated power fiber optic communication networks or industrial Ethernet based on the IEEE 1588 precision time protocol to form star, ring, or mesh communication networks that correspond to the physical topology of the power grid, ensuring the reliability and redundancy of data upload channels.

[0088] Specifically, the basic implementation process of the PMU device includes: (1) Signal acquisition and synchronous sampling: Input: Analog signals from the secondary side of voltage and current transformers.

[0089] After filtering, the signal is sampled by the ADC at a fixed rate (e.g., 4800 points per second). The absolute time reference for sampling is locked by a 1PPS pulse provided by the built-in BeiDou / GPS receiver, ensuring strict synchronization of sampling times across all PMUs in the network.

[0090] (2) Phasor calculation: Apply Discrete Fourier Transform to a continuous sampling sequence of one cycle (e.g., 20ms) to extract the amplitude and phase of the fundamental component.

[0091] Key formula: Fundamental phasor , where N is the number of samples per wave.

[0092] Output: Obtain the voltage phasor U∠ with precise timestamps. and current phasor I∠ .

[0093] (3) Data encapsulation and uploading: Phasors, frequencies, timestamps, etc., are encapsulated into data frames according to a standard format (such as IEEE C37.118.2). These frames are then continuously uploaded to the state estimator of the regional cooperative optimization layer via a communication network (such as fiber optic Ethernet) at a fixed frame rate (e.g., 100 frames per second).

[0094] The PMU in this scheme not only achieves sampling time synchronization, but also ensures strict periodic alignment of the generation and reporting times of data frames across the entire network at the application layer. This is achieved by deeply coupling the internal clock with the IEEE 1588 (PTP) precision network clock protocol. The PTP master clock (usually deployed in the control center) periodically sends out precision time messages, and each PMU, acting as a slave clock, dynamically calibrates its local clock, aligning the start time of the complete processing cycle of sampling, calculation, and framing to a globally unified time slot grid that is fully aligned with the master clock.

[0095] Specifically, in this invention, the distributed model predictive control algorithm used at the upper layer needs to acquire a set of network-wide data with completely consistent timestamps, reflecting a "snapshot" of the power grid at the same instant, at the beginning of each rolling optimization cycle. Although traditional PMU data has timestamps, the randomness of the upload time means that the data received by the controller is actually a mixture of states from different instants, which requires complex interpolation and alignment, introducing errors and delays.

[0096] This invention innovatively enables the uploaded wide-area measurement data to naturally form a strictly aligned spatiotemporal state matrix in the time dimension. The area controller can directly use this as input for MPC state estimation, completely eliminating the data alignment preprocessing step, reducing the sensing delay of the control loop, and significantly improving the accuracy of state estimation, making the prediction-compensation process more precise.

[0097] Optionally, regional-level collaborative control objectives can be sent to each intelligent power converter a via manufacturing message specification service based on the IEC61850 standard or general object-oriented substation event messages.

[0098] The PMU device in this solution integrates a communication stack supporting the IEC 61850 standard in hardware and is modeled as an IED device in software. It not only transmits measurement data as a client but also acts as a GOOSE subscriber and forwarding node. The coordinated control objectives generated by the area controller are encapsulated in GOOSE messages and published to designated multicast groups. Upon receiving these messages, the PMU devices deployed in the same substation or area do not simply ignore them. Instead, they utilize their existing high-speed local communication links directly to the intelligent power converters within the substation to directly forward the control objectives or trigger the converters with extremely low latency.

[0099] In scenarios with a high proportion of renewable energy, control commands need to respond to voltage fluctuations at a speed of seconds or even sub-seconds. Traditional methods relying on dispatch data networks or public wireless networks result in large delays and uncertainties. This innovation utilizes the highly reliable, low-latency fiber optic communication channels already built into the PMU network and distributed across key nodes.

[0100] The PMU device in this invention enables control commands and measurement data to share the same high-performance physical communication channel. This not only saves the cost of building a separate control communication network, but more importantly, it ensures that the path delay from decision-making to execution is minimized, fixed, and controllable. Combined with high-fidelity synchronization of state sequences for model predictive control, it forms a highly deterministic information loop of perception and control, which is the fundamental communication guarantee for achieving millisecond-level coordinated and stable control of a high proportion of renewable energy power grids.

[0101] Figure 4 This is a flowchart illustrating an embodiment 3000 of the dynamic balancing method for a high-proportion renewable energy power grid according to the present invention. Figure 4 As shown, Example 3000 includes steps S301-S304.

[0102] In step S301, the regional collaborative optimization layer generates regional-level collaborative control objectives through distributed optimization based on wide-area measurement and prediction data of the power grid. Optionally, step S301 employs a distributed model predictive control algorithm to perform rolling optimization of the power grid state within future time windows and maps the optimization results to voltage or reactive power collaborative objectives. The core of step S301 is running the distributed model predictive control algorithm, the specific implementation process of which includes: (1) Data input and state estimation: The algorithm receives high-precision synchronous phasor data (wide-area measurement data) and ultra-short-term renewable energy power prediction data from the wide-area synchronous measurement unit network. First, the Kalman filter algorithm is used to perform rolling estimation of the grid state containing renewable energy uncertainties to obtain the optimal estimates of the current and future short-term states of the system and their confidence intervals.

[0103] (2) Rolling optimization solution: Based on state estimation, with the main objective of minimizing the regional net load fluctuation, a finite time window optimization problem considering equipment and safety constraints is constructed and solved. To solve the large-scale computation problem, distributed optimization algorithms such as the alternating direction multiplier method are adopted, so that each regional controller can compute in parallel and obtain the global optimal control sequence by exchanging only boundary information.

[0104] (3) Target mapping and generation: The optimized control sequence (mainly the planned values ​​of each controllable resource in the next few minutes) is mapped into a system-level, directly achievable collaborative control target through voltage-reactive power sensitivity analysis. The typical form is the voltage reference value of the key bus. or regional total reactive power demand .

[0105] Step S301 integrates high-precision state awareness, multi-step look-ahead optimization, and distributed efficient solution. Its technical effect lies in transforming the traditional "post-event response" mode into "pre-event prediction and proactive compensation," generating optimization targets that can proactively mitigate the fluctuations of new energy sources and balance economic and safety considerations, providing clear global guidance for subsequent rapid execution.

[0106] In step S302, each intelligent power converter in the distributed device execution layer receives the cooperative control target. Optionally, step S302 is an information transmission process, in which the regional cooperative optimization layer sends the cooperative control target generated in step S301 to all relevant intelligent power converters in the distributed device execution layer via a high-speed communication network.

[0107] To achieve low latency and high reliability in data transmission, step S302 in this embodiment preferably uses the general object-oriented substation event message service based on the IEC 61850 standard. GOOSE messages feature millisecond-level transmission and multicast capabilities. This ensures that control commands can reach a massive number of distributed terminals quickly, synchronously, and reliably, laying the communication foundation for real-time collaboration.

[0108] In step S303, each intelligent power converter, based on the cooperative control target and local measurement data, autonomously determines its own virtual admittance adjustment through distributed negotiation with neighboring devices. Optionally, step S303 is executed within the consistency coordination module of each intelligent power converter. Each converter, upon receiving the global target (such as...), ... After that, combined with the local voltage measured by itself It exchanges information with adjacent converters connected via a communication network. Optionally, the exchange typically includes their respective voltage estimates and admittance states.

[0109] Subsequently, in step S303, the system runs a distributed consensus algorithm. Through multiple rounds of iterative calculations, the algorithm drives all participating converters to reach a consensus on the system state and autonomously calculates their own reactive power regulation share to approximate the global target. The final output is a specific local virtual admittance setpoint, i.e., the virtual admittance regulation. Optionally, the distributed consensus algorithm includes an average consensus algorithm or a proportional-integral consensus algorithm.

[0110] The core of step S303 is autonomous negotiation and fair allocation. Unlike traditional centralized allocation instructions, it achieves self-organizing task decomposition without centralized command through local information interaction and iterative computation. Its technical effect is to endow the system with extremely high scalability and robustness. Even if some nodes experience communication interruptions, the remaining nodes can still maintain basic coordination through local negotiation. Furthermore, the algorithm can be designed to allocate tasks according to capacity or real-time capability ratios, achieving fairness and efficiency.

[0111] In step S304, each smart power converter adjusts its output according to the virtual admittance adjustment to collaboratively achieve the collaborative control objective and complete the real-time dynamic balancing of the power grid. Optionally, step S304 is executed in the virtual admittance control module of the smart power converter. The module receives the local virtual admittance setpoint determined in step S303. Within each extremely short PWM control cycle (e.g., 50 microseconds), the module performs the following operations: First, according to Using the instantaneous value of the grid connection point voltage v(t) obtained through real-time sampling, the desired current command is calculated based on Ohm's law. Subsequently, a fast current tracking controller (such as a PI controller in the dq coordinate system) and space vector pulse width modulation technology are used to generate pulse signals to drive the power semiconductor switches, enabling the actual output current of the converter to accurately track the signal. .

[0112] Step S304 is the physical implementation of the control command. Its core is to map the admittance setpoint from the digital world to the physical output characteristics of the power electronic converter in real time and with precision through virtual admittance control technology. The technical effect is to make the converter equivalent to a programmable linear admittance element, capable of dynamically injecting or absorbing reactive current at millisecond-level speeds. When all converters in the grid adjust synchronously based on the coordinated results, reactive power deficits can be quickly compensated, voltage stabilized, and real-time dynamic power balance of the power grid can be achieved.

[0113] The embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, any changes or modifications made by those skilled in the art based on the ideas of the present invention, its specific implementation methods, and its application scope, are all within the scope of protection of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A dynamic balancing system for source-grid-load-storage in a high-proportion renewable energy power grid, characterized in that, This includes a regional collaborative optimization layer and a distributed device execution layer; The regional collaborative optimization layer is used to generate regional-level collaborative control targets based on wide-area measurement and prediction data of the power grid through a distributed optimization algorithm. The distributed device execution layer includes multiple smart power converters deployed in the power grid; the smart power converters are configured as follows: Receive the regional-level collaborative control objectives; Based on the aforementioned collaborative control objective and local measurement data, distributed negotiation is conducted through adjacent intelligent power converters that are connected in communication with the converter to determine their respective adjustment commands. The equivalent admittance of its grid connection point is adjusted according to the adjustment command to output the corresponding reactive power, thereby achieving the regional-level coordinated control objective.

2. The system according to claim 1, characterized in that, The system also includes a wide-area synchronous measurement unit network; The wide-area synchronous measurement unit network consists of multiple synchronous phasor measurement units deployed at key nodes of the power grid, used to collect and upload the wide-area measurement data containing voltage and current phasors to the regional collaborative optimization layer at a rate higher than the power frequency.

3. The system according to claim 2, characterized in that, The regional collaborative optimization layer includes at least one regional controller; The distributed optimization algorithm includes a distributed model predictive control algorithm, and the area controller is configured to execute the distributed model predictive control algorithm, specifically including: Based on the wide-area measurement data and ultra-short-term renewable energy power prediction data, the state of the power grid in future periods is estimated in a rolling manner. With minimizing the regional net load fluctuation as the optimization objective, the optimal control sequence within a future finite time window is solved; The optimal control sequence is converted into the regional-level coordinated control objective, which includes at least the voltage reference value of the key bus or the total reactive power demand of the region.

4. The system according to claim 3, characterized in that, When performing rolling estimation, the regional controller specifically uses the Kalman filter algorithm to perform interval estimation of the system state including new energy output.

5. The system according to claim 1, characterized in that, The intelligent power converter includes a virtual admittance control module and a consistency coordination module; The consensus and coordination module is used to execute a distributed consensus algorithm, exchange local voltage information and the coordination control target with adjacent smart power converters, and calculate a local virtual admittance setpoint that makes the local voltage approach the target value through iterative calculation, which serves as the adjustment command. The virtual admittance control module is used to adjust the modulation signal of the inverter bridge of the converter in real time according to the local virtual admittance setting value, so as to realize dynamic control of the equivalent admittance at the grid connection point.

6. The system according to claim 5, characterized in that, The distributed consensus algorithm is either an average consensus algorithm or a proportional-integral consensus algorithm, to achieve proportional allocation of reactive power output of each smart power converter according to capacity or adjustable margin.

7. The system according to claim 6, characterized in that, The regional-level collaborative control objectives are sent to each of the intelligent power converters via the manufacturing message specification service based on the IEC61850 standard or the general object-oriented substation event message.

8. The system according to claim 1, characterized in that, The intelligent power converter can be any one of a photovoltaic inverter, an energy storage converter, or a wind power converter.

9. A method for dynamic balancing of source, grid, load, and storage in a high-proportion renewable energy power grid, characterized in that, The method, applied to the system as described in any one of claims 1-8, comprises: S1. The regional collaborative optimization layer generates regional collaborative control objectives through distributed optimization based on the wide-area measurement and forecast data of the power grid. S2. Each intelligent power converter in the distributed device execution layer receives the cooperative control objective; S3. Each of the intelligent power converters, based on the cooperative control target and local measurement data, autonomously determines its own virtual admittance adjustment through distributed negotiation with adjacent devices; S4. Each of the intelligent power converters adjusts its output according to the virtual admittance adjustment to achieve the collaborative control objective and complete the real-time dynamic balance of the power grid.

10. The method according to claim 9, characterized in that, Step S1 specifically includes: using a distributed model predictive control algorithm to perform rolling optimization of the power grid state within a future time window, and mapping the optimization results to voltage or reactive power coordination targets.

Images