A dynamic coordination control method of a micro-grid of a distributed energy source

Abstract

The application relates to the technical field of micro-grid energy management, and discloses a micro-grid dynamic coordination control method for distributed energy. The method acquires a data set of multiple monitoring nodes of a micro-grid in a historical operation interval; performs multi-resolution fluctuation analysis on an energy output data set, an energy demand data set and a system frequency data set to generate an output fluctuation index, a demand fluctuation index and a frequency fluctuation index; takes the fluctuation index as a guide variable, adopts a group optimization algorithm to explore in a micro-grid control parameter space, and locates an optimal control interval; periodically adjusts distributed power generation units and load units in the micro-grid according to the optimal control interval, generates a regulation and control instruction set and executes the regulation and control instruction set to achieve dynamic coordination control. The application realizes accurate interval positioning and efficient coordinated operation of micro-grid control, and effectively improves the stability and regulation and control efficiency of a distributed energy system.

Description

Technical Field

[0001] This invention relates to the field of microgrid energy management technology, specifically to a dynamic coordination control method for distributed energy microgrids. Background Technology

[0002] Against the backdrop of a global energy transition towards renewable energy, and with the introduction of the "dual carbon" target, countries worldwide are increasing their development and utilization of renewable energy sources such as solar, wind, and hydropower. The traditional energy system, dominated by fossil fuels, is no longer able to meet the demands of sustainable development due to environmental pollution and resource depletion. Renewable energy, with its clean, low-carbon, and sustainable characteristics, has become a core force in the energy transition.

[0003] Microgrids, as an important carrier of distributed energy, have emerged to meet this need. They organically integrate distributed power sources, energy storage devices, energy conversion devices, and loads to form a small-scale, self-controllable and manageable power generation and distribution system. In the energy transition process, microgrids play a crucial role, not only promoting the large-scale integration and efficient utilization of renewable energy, but also effectively addressing the intermittency and volatility of distributed energy generation, providing a feasible solution for achieving stable energy supply and sustainable development.

[0004] Microgrids play a vital role in urban complexes and industrial parks. Taking industrial parks as an example, microgrids can integrate distributed energy resources within the park, achieving optimized energy allocation and utilization. When the main grid fails, the microgrid can switch to independent operation mode, utilizing local distributed power sources (photovoltaics, wind power) and energy storage systems to ensure uninterrupted power supply to critical production lines, avoiding significant economic losses caused by power outages. During peak electricity consumption periods, intelligent control systems balance supply and demand in real time, alleviating voltage pressure and preventing voltage fluctuations or power rationing, ensuring normal production activities. Simultaneously, microgrids can reduce electricity costs for businesses and improve energy efficiency through peak shaving and demand control strategies, helping enterprises achieve green and low-carbon transformation.

[0005] Currently, microgrid control technology still faces many challenges. The output of distributed power sources is significantly affected by natural conditions; for example, solar power depends on sunlight intensity and duration, while wind power depends on wind speed and direction, resulting in significant intermittency and fluctuations in power generation. Load-side electricity demand also varies with time, season, and user behavior, exhibiting uncertainty. Existing control technologies struggle to accurately predict and effectively respond to these complex and changing conditions, failing to achieve efficient energy allocation and real-time supply-demand balance.

[0006] In islanded operation mode, microgrids lose the support of the main power grid and need to rely on their own control strategies to maintain stable system operation. However, existing control methods are insufficient in terms of frequency and voltage stability control. When distributed power sources and loads undergo significant changes, it can easily lead to large fluctuations in frequency and voltage, affecting power quality and potentially even causing system collapse.

[0007] With the continuous expansion of microgrid scale and increasing number, coordinated operation among multiple microgrids has become an inevitable trend. However, the current lack of effective communication and coordination mechanisms among microgrids makes it difficult to achieve resource sharing and optimized allocation, thus failing to fully leverage the overall advantages of multi-microgrid systems. Furthermore, the interaction between microgrids and the main power grid presents challenges such as power flow control difficulties arising from bidirectional power flow and potential impacts on the stability of the main power grid. Summary of the Invention

[0008] The purpose of this invention is to provide a dynamic coordination control method for microgrids of distributed energy resources to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides a dynamic coordinated control method for microgrids of distributed energy resources, the method comprising:

[0010] Acquire energy output datasets, energy demand datasets, and system frequency datasets from multiple monitoring nodes of the microgrid within its historical operating range; perform multi-resolution fluctuation analysis on the energy output datasets, energy demand datasets, and system frequency datasets to generate output fluctuation index, demand fluctuation index, and frequency fluctuation index;

[0011] Using the fluctuation index as a guiding variable, a swarm optimization algorithm is used to explore and locate the optimal control interval in the microgrid control parameter space. Based on the optimal control interval, the distributed generation units and load units in the microgrid are periodically adjusted to generate a set of control instructions and execute them to achieve dynamic coordinated control.

[0012] Preferably, acquiring energy output datasets, energy demand datasets, and system frequency datasets from multiple monitoring nodes of the microgrid within a historical operating period includes: defining the historical operating period as a sliding time window, with the window size adaptively adjusted based on load change patterns and external environmental factors in the microgrid's operating history; continuously recording energy output datasets and energy demand datasets using power sensing devices deployed at distributed energy interfaces and consumption sensing devices deployed at load connection points; and collecting system frequency datasets through a system frequency monitor.

[0013] Preferably, performing multi-resolution fluctuation analysis on the energy output dataset, energy demand dataset, and system frequency dataset includes: using dynamic mode decomposition technology to decompose each dataset into a series of dynamic mode components with different oscillation frequencies; calculating the energy distribution characteristics of each dynamic mode component; determining the contribution weight of each dynamic mode component based on the energy distribution characteristics; and weighting and fusing the fluctuation characteristics of all dynamic mode components according to the contribution weights to generate an output fluctuation index, a demand fluctuation index, and a frequency fluctuation index.

[0014] Preferably, the dynamic mode decomposition technique is used to decompose each dataset into a series of dynamic mode components with different oscillation frequencies, including: constructing a time-shift matrix of the data sequence; performing singular value decomposition on the time-shift matrix to extract the main feature patterns; solving for the oscillation frequency and decay rate of the dynamic modes through eigenvalue decomposition; retaining the main dynamic mode components according to a preset mode screening threshold; reconstructing the retained dynamic mode components and verifying their fit with the original data.

[0015] Preferably, using the fluctuation index as a guiding variable, the exploration in the microgrid control parameter space using a swarm optimization algorithm includes: constructing the microgrid control parameter space as a multi-dimensional space, with dimensions corresponding to the output fluctuation index, demand fluctuation index, and frequency fluctuation index; initializing a particle swarm, where each particle represents a candidate solution for a control interval; calculating the fitness value of each particle, with the fitness function based on the control error assessment of the control interval in the simulation environment; updating the particle position and velocity, iteratively searching until convergence, and outputting the optimal control interval.

[0016] Preferably, constructing the microgrid control parameter space into a multi-dimensional space includes: collecting multiple sets of sample output fluctuation indices, sample demand fluctuation indices, and sample frequency fluctuation indices from historical operation, as well as the corresponding multiple sets of sample control intervals as training samples; establishing a three-dimensional coordinate frame, with the X-axis corresponding to the output fluctuation index, the Y-axis corresponding to the demand fluctuation index, and the Z-axis corresponding to the frequency fluctuation index; mapping the training samples onto the three-dimensional coordinate frame to form a labeled point cloud structure, thus defining the microgrid control parameter space.

[0017] Preferably, initializing the particle swarm includes: randomly generating a certain number of particles, with the position vector of each particle randomly assigned within the microgrid control parameter space; setting the particle velocity range to a preset minimum and maximum value; and defining the swarm size and upper limit of the number of iterations.

[0018] Preferably, calculating the fitness value of each particle includes: simulating the operation of the microgrid under the control interval corresponding to the particle, calculating the energy supply and demand balance deviation and frequency stability index; and combining the deviation and index with weighted summation to obtain the fitness value, with the weights set according to the microgrid priority.

[0019] Preferably, the periodic adjustment of distributed generation units and load units in the microgrid according to the optimal control interval includes: dividing the optimal control interval into continuous control periods, allocating power generation setpoints and load adjustment values ​​to each period; using a rolling optimization strategy to calculate the power allocation scheme for each period to ensure that system constraints are met; and summarizing the setpoints of all periods to form a control instruction set.

[0020] Preferably, generating and executing the control instruction set includes: distributing the control instruction set to the distributed generation controller and load controller through the control network; monitoring the execution status of the instructions in real time and comparing the actual operating data with the expected values; if the deviation exceeds the tolerance range, restarting the process of acquiring data from multiple monitoring nodes of the microgrid within the historical operating range and performing a new round of control calculations.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] In the field of energy management, accurate understanding of energy fluctuations is the cornerstone of efficient regulation. This method comprehensively and meticulously collects key data on microgrid operation by acquiring energy output, energy demand, and system frequency datasets from multiple monitoring nodes within a microgrid's historical operating range. These datasets cover multiple dimensions, including energy production, consumption, and system operating status, providing a rich and accurate information foundation for subsequent analysis.

[0023] Multi-resolution fluctuation analysis is performed on these datasets, employing advanced algorithms and techniques to deeply mine the hidden fluctuation characteristics behind the data, thereby generating fluctuation indices, demand fluctuation indices, and frequency fluctuation indices. These fluctuation indices can intuitively and accurately reflect the fluctuations in energy output, demand, and system frequency, enabling the system to sensitively perceive subtle changes in energy.

[0024] Taking a microgrid in an industrial park as an example, before adopting this method, the inability to accurately predict energy fluctuations frequently led to energy shortages or surpluses, affecting the normal production of enterprises within the park. However, after introducing this method, through precise analysis of various datasets and the generation of fluctuation indices, energy fluctuation trends can be predicted in advance. During the high-temperature period in summer, when it is predicted that photovoltaic power generation will decrease due to cloud cover, the system can promptly adjust the output of other distributed generation units, ensuring a stable power supply within the park and avoiding production interruptions caused by energy fluctuations.

[0025] Finding the optimal control strategy is a crucial issue in the control process of microgrids. This method uses output fluctuation index, demand fluctuation index, and frequency fluctuation index as guiding variables, making full use of the energy fluctuation information contained in these indices to provide a clear search direction for the swarm optimization algorithm.

[0026] Swarm optimization algorithms are optimization algorithms that simulate the intelligent behavior of biological swarms. They possess powerful global search capabilities and efficient optimization performance. In the control parameter space of microgrids, this algorithm, through continuous exploration and iteration, can find the optimal control interval in complex parameter combinations.

[0027] In practical applications, taking a microgrid in a commercial complex as an example, the traditional fixed control interval method could not adjust the control strategy in real time according to energy fluctuations, resulting in serious energy waste. However, with this method, the swarm optimization algorithm continuously explores the optimal control interval based on the fluctuation index. During peak electricity consumption periods, the algorithm can quickly adjust the control interval, enabling distributed generation units and energy storage systems to work collaboratively, maximizing load demand while reducing energy waste and improving energy utilization efficiency. Through this intelligent exploration of optimal control, this method not only improves the control accuracy and efficiency of the microgrid but also reduces operating costs, providing a strong guarantee for the economical and stable operation of the microgrid.

[0028] Periodic adjustment of distributed generation units and load units in a microgrid according to the optimal control interval is the core link in realizing dynamic coordinated control of the microgrid. In this process, the system can flexibly adjust the output of distributed generation units and the power consumption of load units according to the real-time energy fluctuations, thereby achieving a dynamic balance between energy supply and demand.

[0029] In a microgrid on a certain island, its energy supply mainly relies on distributed energy resources, and load demand fluctuates significantly due to factors such as the tourist season. Before adopting this method, energy supply and demand imbalances frequently occurred, leading to unstable power supply. After applying this method, the system, based on the optimal control interval, promptly increases the power generation of distributed generation units when load demand increases during the peak tourist season and rationally adjusts the charging and discharging strategies of the energy storage system to ensure a stable power supply; conversely, during the off-season when load demand decreases, the output of distributed generation units is reduced, avoiding energy waste.

[0030] This dynamic coordination and control can also effectively improve the stability of the system. When the output of distributed generation units fluctuates due to changes in natural conditions, the system can quickly maintain the power balance of the system by adjusting the power distribution of the load units, avoiding frequency and voltage fluctuations caused by power imbalance, and ensuring the stable operation of the microgrid. Attached Figure Description

[0031] Figure 1 This is a schematic diagram illustrating the working principle of the microgrid dynamic coordination control method for distributed energy described in this invention.

[0032] Figure 2 A flowchart for obtaining multiple datasets for the historical operating range of a microgrid;

[0033] Figure 3 A flowchart for multi-resolution fluctuation analysis of the dataset;

[0034] Figure 4 This is a diagram illustrating the convergence process of the particle swarm optimization algorithm.

[0035] Figure 5 Initialize the distribution map for the particle swarm. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figure 1 This invention provides a dynamic coordinated control method for microgrids with distributed energy resources. The method includes: acquiring energy output datasets, energy demand datasets, and system frequency datasets from multiple monitoring nodes within the microgrid's historical operating period. These datasets form the basis for analyzing the system's dynamic behavior. Subsequently, multi-resolution fluctuation analysis is performed on the energy output datasets, energy demand datasets, and system frequency datasets. This analysis aims to extract the inherent fluctuation patterns of the data from different time scales, generating fluctuation indices, demand fluctuation indices, and frequency fluctuation indices. These fluctuation indices quantify the degree of uncertainty of the system in different aspects. Next, using the fluctuation indices as guiding variables, a swarm optimization algorithm is employed to explore the microgrid control parameter space. The swarm optimization algorithm simulates collective intelligent behavior, performing parallel searches in the multi-dimensional space defined by the fluctuation indices to locate the optimal control interval. The optimal control interval reflects the most suitable control action cycle under the current fluctuation state. Finally, based on the optimal control interval, the distributed generation units and load units in the microgrid are periodically adjusted, generating and executing a control instruction set containing specific power commands. The issuance and execution of the control instruction set constitute closed-loop control, achieving dynamic coordinated control between distributed energy resources and loads within the microgrid.

[0038] Example 1: See Figure 2Obtaining energy output, energy demand, and system frequency datasets from multiple monitoring nodes within the historical operating range of a microgrid is the data foundation for implementing dynamic coordinated control methods. The completeness and accuracy of data acquisition directly affect the effectiveness of subsequent fluctuation analysis and control decisions. The historical operating period is defined as the sliding time window. The size of the sliding time window is not fixed, but adaptively adjusted based on the load change patterns and external environmental factors in the microgrid's operating history. The analysis of load change patterns requires studying the electricity consumption patterns of the microgrid on daily, weekly, and even monthly scales, identifying peak and off-peak periods, their duration, and gradients. External environmental factors focus on key elements affecting renewable energy generation. For example, for microgrids dominated by photovoltaic power generation, the patterns of solar radiation intensity and cloud cover changes are the main considerations. For systems with a high proportion of wind power generation, the historical statistical characteristics of wind speed and direction become important bases. The adaptive adjustment mechanism of the sliding time window is implemented through an independent configuration algorithm. The configuration algorithm periodically analyzes recent historical data. When a significant change in load pattern or a sharp fluctuation in the external environment is detected, the configuration algorithm dynamically increases the capacity of the sliding time window to capture longer-term trend information. During relatively stable periods of system operation, the configuration algorithm appropriately reduces the sliding time window to focus on recent dynamics and improve the control response speed.

[0039] Power sensors deployed at distributed energy interfaces and consumption sensors at load connection points continuously record energy output and demand datasets. Power sensors typically employ high-precision smart meters or power transmitters, directly installed at the grid connection points of distributed generation units such as photovoltaic inverters, wind turbines, and energy storage converters. They measure and record active and reactive power outputs in real time, packaging the measurement data at a set sampling frequency and transmitting it to the data acquisition unit via fieldbus or industrial Ethernet. Consumption sensors are deployed on critical load feeders or key electrical equipment, also based on high-precision measurement technology, to monitor load power consumption in real time. The recorded data includes not only total active and reactive power but, in some refined management scenarios, may also include power quality parameters such as harmonic content and power factor. All sensors are equipped with a unified time synchronization module, typically using the IEEE 1588 precision clock synchronization protocol, ensuring that data from different physical locations have accurate and consistent timestamps, creating the necessary conditions for subsequent multi-point data fusion and correlation analysis. System frequency datasets are collected by a system frequency monitor. System frequency is a core parameter for measuring the power quality and stability of a microgrid. The system frequency monitor has the characteristics of high sampling rate and high accuracy, and can capture instantaneous fluctuations in system frequency. Its installation location is selected at the AC / DC conversion interface of the key bus or main control unit of the microgrid to obtain the most representative system frequency signal. The operating frequency of the system frequency monitor is much higher than that of power measurement, and it can collect hundreds or even thousands of frequency samples per second. After digital filtering, these high-frequency data are used to calculate the real-time value of the frequency and to analyze the rate of change of the frequency. The system frequency dataset, energy output dataset, and energy demand dataset are strictly aligned in the time dimension. The three sets of data together constitute a multi-dimensional time series, which fully depicts the dynamic behavior trajectory of the microgrid within the historical operating range.

[0040] Data faces risks of noise interference and data loss during transmission and aggregation, necessitating a rigorous data preprocessing procedure. This process includes data cleaning, format standardization, and timestamp alignment. Data cleaning identifies and handles outlier data points, such as using statistical methods to identify outliers outside reasonable ranges and employing interpolation algorithms to fill in data gaps caused by brief communication interruptions. Format standardization converts data from sensors with different manufacturers and protocols into a standardized data format defined within the system, eliminating data heterogeneity. Timestamp alignment, based on high-precision clock synchronization, compensates for data with potential minor transmission delays, ensuring accurate correlation of data from all monitoring nodes at the same time. The preprocessed energy output, energy demand, and system frequency datasets are stored in a distributed real-time database. This database provides efficient data query and access interfaces, offering stable and reliable data services for the subsequent multi-resolution fluctuation analysis module.

[0041] The collaborative work of the sliding window mechanism of historical operating intervals and the data acquisition system constructs a continuously updated data pool. The data pool not only serves the current control cycle, but its accumulated long-term historical data also provides valuable data resources for analyzing the operating characteristics of the microgrid and optimizing control parameters. The entire data acquisition process is an automated, periodic cycle. Whenever a new control cycle starts, the system automatically extracts the latest historical operating data from the data pool according to the updated sliding time window definition, thereby ensuring that control decisions are always based on the latest system state information, laying the cornerstone for the adaptive capability of the entire dynamic coordinated control method.

[0042] Example 2: See Figure 3The core of performing multi-resolution fluctuation analysis on energy output datasets, energy demand datasets, and system frequency datasets lies in using dynamic mode decomposition technology to deeply analyze the inherent spatiotemporal modal characteristics of the data. Dynamic mode decomposition technology decomposes each nonlinear nonstationary dataset into a series of dynamic mode components with different oscillation frequencies, and each dynamic mode component represents an inherent physical fluctuation pattern. Constructing the time-shift matrix of the data sequence is the starting point for dynamic mode decomposition (VM) techniques. The time-shift matrix consists of the original time series data and multiple data copies formed after a fixed time step delay. This matrix structure maps one-dimensional time series data to a high-dimensional state space, in which the dynamic evolution of the data can be linearly expressed. Singular value decomposition (SVD) is performed on the time-shift matrix to extract the main feature patterns. SVD decomposes the complex time-shift matrix into a product of a left singular vector matrix, a singular value matrix, and a right singular vector matrix. The magnitude of the singular values ​​directly reflects the energy proportion of the corresponding pattern in the original data. By setting a singular value threshold, patterns with lower energy that may be caused by noise can be filtered out, achieving data dimensionality reduction and noise removal while retaining the feature patterns that best represent the main dynamics of the system.

[0043] The oscillation frequency and decay rate of the dynamic modes are solved by eigenvalue decomposition. Based on the results of singular value decomposition, a low-dimensional approximate system matrix is ​​further constructed. Eigenvalue decomposition is then performed on this approximate system matrix to obtain a series of eigenvalues ​​and their corresponding eigenvectors. Each eigenvalue is a complex number. The imaginary part of the complex number determines the oscillation frequency of the corresponding dynamic mode component, while the real part reveals the decay rate or growth rate of the mode component. A negative real part means that the fluctuation of the mode will decay over time, while a positive real part means that the fluctuation will be amplified. The main dynamic mode components are retained according to a preset mode screening threshold. The mode screening threshold is usually set based on the magnitude of the eigenvalue or the energy contribution rate of the corresponding mode component. For example, mode components with a magnitude greater than one or an energy contribution rate of more than 95% can be retained. This screening process eliminates those rapidly decaying transient modes and modes representing background noise, focusing on the key modes that have a continuous impact on the long-term dynamic behavior of the system. The retained modal components are reconstructed and their fit with the original data is verified. The modal components and their time evolution functions obtained using dynamic mode decomposition (VMD) can be used to reconstruct the original time series. The error norm between the reconstructed data and the original data, such as the root mean square error (RMSE), is calculated to verify the accuracy of the VMD process and the rationality of the modality selection threshold, ensuring that the decomposed modal components can fully capture the main dynamic characteristics of the original data without introducing significant distortion. The energy distribution characteristics of each dynamic modal component are calculated. Energy distribution characteristics are typically quantified by calculating the square integral of the amplitude of each modal component over the entire analysis period or its L2 norm. Modal components with higher energy values ​​indicate a more important role in the overall data fluctuations.

[0044] The contribution weight of each modal component is determined based on the energy distribution characteristics. The contribution weight can be obtained by dividing the energy value of a single modal component by the total energy value of all retained modal components. The higher the energy proportion of a modal component, the greater its contribution weight. This weighting method enables the subsequent fusion process to highlight the influence of the dominant fluctuation mode. Based on contribution weights, the fluctuation characteristics of all modal components are weighted and fused to generate output fluctuation index, demand fluctuation index, and frequency fluctuation index. The fluctuation characteristics of each modal component can be described by parameters such as amplitude and frequency. The weighted fusion process is not a simple superposition of modal components, but rather the summation of the key fluctuation characteristic parameters of each modal component multiplied by its contribution weight, ultimately synthesizing a comprehensive scalar index. For the energy output dataset, the output fluctuation index is obtained after weighted fusion, which quantifies the uncertainty and fluctuation intensity of renewable energy power generation output. For the energy demand dataset, the demand fluctuation index is generated by weighted fusion, which reflects the severity and unpredictability of load changes. For the system frequency dataset, the same process generates the frequency fluctuation index, which characterizes the degree of disturbance to the system's stable state. These three fluctuation indices together constitute a multidimensional feature vector describing the dynamic characteristics of the current operating state of the microgrid.

[0045] The multi-resolution fluctuation analysis process transforms the raw, seemingly chaotic time-series data into a quantified fluctuation index with clear physical meaning. This fluctuation index not only reflects the overall intensity of the fluctuation but also implicitly contains information about the main frequency components of the fluctuation due to its origin from the fusion of modes at different time scales. The application of dynamic mode decomposition technology enables the analysis process to adaptively capture dynamic behaviors at different time scales, from seconds to hours, avoiding the limitations of fixed-time-window analysis methods. The generated fluctuation index provides accurate, dimensionality-reduced guiding variables for subsequent exploration using swarm optimization algorithms in the microgrid control parameter space. This allows the optimization search of the control interval to directly target the core dynamic characteristics of the system, improving the targeting and effectiveness of the coordinated control method.

[0046] Example 3: Exploring the microgrid control parameter space using a swarm optimization algorithm with fluctuation indices as the guiding variable is a core step in achieving intelligent decision-making. The fluctuation indices include output fluctuation index, demand fluctuation index, and frequency fluctuation index, which together constitute a three-dimensional feature vector characterizing the current dynamic characteristics of the system. Constructing the microgrid control parameter space as a multi-dimensional space is the basic framework for optimization search. The dimensions of the microgrid control parameter space directly correspond to the dimensions of the fluctuation indices; that is, the three orthogonal coordinate axes represent the numerical ranges of the output fluctuation index, demand fluctuation index, and frequency fluctuation index, respectively. Multiple sets of sample output fluctuation indices, sample demand fluctuation indices, and sample frequency fluctuation indices from historical operation, along with corresponding multiple sets of sample control intervals, are collected as training samples. These training samples originate from successful control cases recorded during the long-term operation of the microgrid. Each sample point contains a set of fluctuation index conditions and the control interval value proven effective under those conditions. A three-dimensional coordinate framework is established, with its axes strictly aligned with the output fluctuation index, demand fluctuation index, and frequency fluctuation index. The physical meaning and numerical range of the indices are defined: the X-axis corresponds to the output fluctuation index, the Y-axis to the demand fluctuation index, and the Z-axis to the frequency fluctuation index. Training samples are mapped onto a three-dimensional coordinate frame. Each training sample is assigned a unique spatial location within the frame based on its specific output fluctuation index, demand fluctuation index, and frequency fluctuation index values. The corresponding sample control interval is then associated with this spatial location as a label, forming a labeled point cloud structure. This spatial point cloud structure, which contains historical experience, clearly defines the microgrid control parameter space. The microgrid control parameter space is a continuous spatial region, where any point corresponds to a possible fluctuation state. The goal of the optimization algorithm is to find the optimal control interval for the current fluctuation state within this space.

[0047] The choice of Particle Swarm Optimization (PSO) algorithm stems from its powerful global search capability and ease of implementation. PSO simulates the collective behavior of flocks of birds or schools of fish, finding the optimal solution in the solution space through the interaction of individual and collective experience. Initializing the particle swarm is the starting point for PSO operation. A certain number of particles are randomly generated; these particles carry candidate solutions, and the swarm size affects the diversity of the search and the convergence speed. Each particle's position vector is randomly assigned within the microgrid control parameter space. The position vector is a three-dimensional vector, with its three components representing a candidate control interval parameter. During initialization, these components are randomly generated within a preset parameter range to ensure the initial swarm is evenly distributed throughout the microgrid control parameter space. The particle velocity range is set to preset minimum and maximum values. The velocity vector, also a three-dimensional vector, determines the step size and direction of the particle's position update in a single iteration. The velocity range limits the amplitude of particle movement, preventing the search process from diverging or getting trapped in local oscillations. The swarm size and upper limit of the number of iterations are defined. These are key parameters controlling the algorithm's computational cost and solution accuracy, requiring offline debugging to achieve a balance.

[0048] Calculating the fitness value of each particle is the core operation for evaluating particle quality and guiding the search direction. The fitness function is based on the control error assessment of the control interval in the simulation environment. The operation of the microgrid under the control interval corresponding to each particle is simulated, and a simulation model reflecting the main dynamic characteristics of the microgrid is constructed. The simulation model includes a distributed generation unit model, a load model, a network topology, and a control system model. The candidate control interval parameters represented by the particle position vector are injected into the controller of the simulation model and run in the simulation environment for a sufficiently long time. During the simulation, typical energy output fluctuations and load demand fluctuation sequences are applied to verify the effectiveness of the control intervals. The energy supply and demand balance deviation and frequency stability index are calculated. The energy supply and demand balance deviation can be quantified by calculating the integral or root mean square value of the difference between total generation power and total load power during the simulation time. The frequency stability index can be characterized by parameters such as the maximum deviation of the system frequency from the rated value and the average value of the frequency change rate. The fitness value is obtained by weighted summation of the deviation and index. The design of the fitness function directly determines the optimization objective. The specific mathematical expression of the fitness function is defined as follows:

[0049] ;

[0050] Where: symbol Represents the calculated fitness value, symbol A quantitative value representing the imbalance between energy supply and demand, with the symbol […]. The quantized value representing the frequency stability index, with the symbol... It is a weighting coefficient used for the energy supply and demand balance deviation, symbol These are weighting coefficients used in frequency stability metrics. and weighting coefficients The specific value is set according to the microgrid priority, and its magnitude reflects the relative importance of energy supply and demand balance and frequency stability in the optimization objectives.

[0051] Updating particle positions and velocities is the mechanism of iterative evolution in the Particle Swarm Optimization (PSO) algorithm. Each particle updates its velocity and position based on its historical best position and the best position discovered by the entire swarm so far. Iterative search until convergence is a continuous optimization process. In each iteration, the fitness values ​​of all particles are recalculated, and the historical best positions of individuals and the global best position of the swarm are updated, guiding the swarm towards regions with better fitness. Outputting the optimal control interval is the final result of the PSO algorithm's exploration. When the algorithm meets convergence conditions, such as reaching the preset maximum number of iterations or the global best fitness value no longer significantly improving in consecutive iterations, the algorithm terminates and outputs the control interval parameter corresponding to the current global best position of the swarm as the optimal control interval. This optimal control interval will be passed to the subsequent periodic adjustment module for execution. The exploration process of the PSO algorithm in the microgrid control parameter space essentially combines empirical knowledge based on historical data with the real-time fluctuation characteristics of the current system, automatically seeking optimization through intelligent search to dynamically determine the most suitable control time scale. This is a crucial link in achieving efficient and coordinated control of the microgrid.

[0052] See Figure 4 This paper demonstrates the convergence characteristics of the particle swarm optimization algorithm in the search space of microgrid control parameters. By observing the changes in fitness values ​​during the iteration process, the dynamic process of the algorithm gradually converging towards the optimal solution region from the initial random state can be clearly observed. The optimal fitness curve reflects the quality improvement of the best solutions in each iteration, while the average fitness curve shows the improvement in the overall performance of the swarm. This convergence characteristic verifies the effectiveness and reliability of swarm intelligence algorithms in solving complex microgrid control parameter optimization problems. The convergence trend in the graphs provides important guidance for selecting appropriate iteration numbers and algorithm parameters in practical engineering applications.

[0053] Example 4: Initializing the particle swarm optimization (PSO) algorithm is the first step in enabling the PSO algorithm to effectively explore the microgrid control parameter space. PSO is an optimization technique that simulates social behavior, and its performance largely depends on the diversity and distribution characteristics of the initial population. A certain number of particles are randomly generated; each particle is the basic unit carrying potential solutions in the algorithm, and each particle corresponds to a specific location in the microgrid control parameter space. The swarm size is a parameter that needs to be preset. The swarm size directly affects the balance between the algorithm's global exploration capability and local exploration capability. A larger swarm size helps cover a wider search area and reduces the risk of getting trapped in local optima, but it also increases the computational burden of each iteration. A smaller swarm size, while computationally efficient, may lead to insufficient population diversity and premature convergence to a suboptimal solution. Setting the swarm size requires comprehensive consideration of the complexity of the microgrid control parameter space and available computational resources, and a reasonable range of values ​​is usually determined through a series of preliminary experiments.

[0054] The position vector of each particle is randomly assigned within the microgrid control parameter space. The position vector is a multi-dimensional vector, and its dimensions are strictly consistent with the dimensions of the microgrid control parameter space. In the current implementation, the microgrid control parameter space is a three-dimensional space composed of output fluctuation index, demand fluctuation index, and frequency fluctuation index. Therefore, the position vector of each particle also contains three components. Each component of the position vector represents an optimization decision variable, that is, a certain aspect of the control interval parameter or a mapping relationship related to the fluctuation index. The random assignment process must be carried out within the preset upper and lower bounds of each decision variable. The upper and lower bounds define the feasible region of the optimization problem. These boundary values ​​are set based on the analysis of the physical constraints and operating experience of the microgrid control system. For example, the lower bound of the control interval is limited by the response speed of the controller actuator, while the upper bound of the control interval is constrained by the system stability. The uniform random initialization strategy can ensure that the initial particle swarm can be uniformly distributed throughout the entire feasible microgrid control parameter space. The particle velocity range is set to a preset minimum and maximum value. The particle velocity vector has the same dimension as the position vector, and the velocity vector determines the step size and direction of the particle's position update in each iteration. The velocity range is achieved by setting the maximum velocity limit value in each dimension. The maximum velocity limit value is a critical parameter, and its size needs to be carefully selected. An excessively large maximum velocity limit value may cause the particle's flight step size to be too large, resulting in oscillations near the optimal solution or even flying over the optimal solution region. An excessively small maximum velocity limit value will weaken the particle's exploration ability, slow the convergence speed, and make it prone to stagnation in local optima. The maximum velocity limit value is usually proportional to the search space range of the corresponding decision variable. A common setting method is to set the maximum velocity limit value to a fixed proportion of the search space range of the variable (upper bound minus lower bound), such as 10% to 20%. The setting of the velocity range provides basic dynamic constraints for the movement of particles in the microgrid control parameter space.

[0055] Defining the swarm size and upper limit of the number of iterations are essential steps in configuring the running parameters of the particle swarm optimization algorithm. The swarm size and upper limit of the number of iterations together determine the upper limit of the algorithm's computational cost. The upper limit of the number of iterations specifies the maximum number of generations the algorithm can run, which is crucial to preventing the algorithm from looping infinitely. When the algorithm reaches the upper limit of the number of iterations, the search process terminates regardless of whether other convergence conditions are met, and the currently found optimal solution is output. The product of the swarm size and the upper limit of the number of iterations is approximately equal to the total number of times the algorithm needs to evaluate fitness values, which directly affects the computation time. In practical applications, the specific values ​​of the swarm size and the upper limit of the number of iterations need to be weighed based on the complexity of the problem and the quality requirements for the solution. A typical parameter combination might include a swarm size of 20 to 50 particles and an upper limit of 100 to 500 generations. The specific configuration of these parameters can refer to empirical values ​​for similar optimization problems, and parameter sensitivity analysis should be performed in an offline environment to determine the most suitable settings for the current microgrid control optimization problem. Table 1 shows a set of parameter configuration examples that can be used to initialize the particle swarm optimization algorithm; these parameters are based on debugging experience in a typical microgrid control scenario.

[0056] Table 1: Particle Swarm Optimization Algorithm Parameter Configuration

[0057]

[0058] Calculating the fitness value of each particle serves as a bridge between particle swarm optimization (PSO) and microgrid control optimization problems. The fitness value is a quantitative evaluation of the quality of the solution represented by the particle. The microgrid operation under the control intervals corresponding to the particles is simulated using a high-fidelity dynamic simulation model. This model needs to accurately reflect the characteristics of distributed generation units, the dynamic response of loads, and the distribution of network power flow. The particle position vectors are decoded into specific control interval parameters, and these parameters are configured into the controller of the simulation model. A representative operating time is simulated in the simulation environment; this time should be long enough to cover various operating conditions to comprehensively evaluate the performance of the control intervals. The simulation requires input of real or typical historical data sequences, such as solar irradiance variation curves, wind speed variation curves, and load fluctuation curves, to simulate uncertainties under real-world conditions. The calculation of energy supply and demand balance deviation and frequency stability index is performed. Energy supply and demand balance deviation can be quantified by the statistical characteristics of the system's net power (generated power minus load power) during the simulation period, such as root mean square error or mean absolute error. Frequency stability index focuses on the dynamic quality of the system frequency; commonly used indicators include the maximum value of frequency deviation, the root mean square value of frequency deviation, and the maximum value of the rate of change of frequency. These indicators reflect the control effect from different perspectives and need to be comprehensively considered. The fitness value is obtained by weighted summation of the deviation and index. Weighted summation integrates multiple indicators with potentially different physical meanings and dimensions into a single scalar value, facilitating comparison and selection by the particle swarm optimization algorithm. The weight allocation reflects the different priorities of microgrid operation; for example, microgrids with extremely high power supply reliability requirements will assign higher weight to energy supply and demand balance deviation, while microgrids sensitive to power quality will place greater emphasis on frequency stability index. The calculated fitness value is the direct driving force for the iterative optimization of the particle swarm optimization algorithm.

[0059] See Figure 5 This study presents the distribution characteristics of the particle swarm optimization algorithm in the control parameter space during the initialization phase. The three dimensions correspond to key fluctuation indicators of the microgrid system, forming a complete state representation space. The uniform spatial distribution of particles ensures that the algorithm possesses sufficient global exploration capability, avoiding premature entrapment in local optima. The color mapping displaying the fitness value gradient reveals the differences in control effects between different parameter combinations, providing an intuitive reference for understanding the characteristics of the parameter space. This initialization strategy lays a solid foundation for subsequent optimization searches, ensuring that the algorithm can find high-quality solutions to complex multimodal problems.

[0060] Example 5: Periodically adjusting distributed generation units and load units in a microgrid according to the optimal control interval is a key stage in translating optimization decisions into actual control actions. The optimal control interval is the best control period explored by the particle swarm optimization algorithm based on the current system fluctuation state. Taking a microgrid containing photovoltaic generation units, energy storage converters, and adjustable industrial loads as an example, assume that the optimal control interval obtained through optimization calculation is five minutes. The optimal control interval is divided into continuous control periods, each with a fixed length of five minutes. This means that the control system will periodically generate and issue operating instructions for all controlled units in the next period, using five minutes as the basic time unit. A power generation setpoint and a load adjustment value are allocated to each period. The power generation setpoint clearly specifies the active power output target that the photovoltaic generation unit needs to maintain in the next five minutes. This target value may not be its maximum output capacity, but rather an instruction given after comprehensively considering system balance. The setpoint of the energy storage converter may include power instructions for charging or discharging and their duration. The load adjustment value is for those industrial loads participating in demand response, such as instructing a certain interruptible production equipment to reduce its power consumption by a predetermined value in the next five minutes.

[0061] A rolling optimization strategy is employed to calculate the power allocation scheme for each time period. This strategy is an advanced control method based on model predictive control. At the beginning of each control period, it re-optimizes the control commands for one or more future time periods using the latest system state measurement information. At the start of the five-minute control interval, the central controller collects the real-time output of all distributed generation units in the microgrid, the real-time power of all loads, the current state of charge of energy storage units, and the power exchange value of the point of common coupling. The controller combines this real-time data with the ultra-short-term photovoltaic power prediction data and load prediction data for the next five minutes to construct a refined finite-time domain optimization problem. The goal of the optimization problem is to minimize the total operating cost or specific performance indicators within the next five minutes, while satisfying all equipment operating constraints. Examples of optimization methods include minimizing the cost of purchasing electricity from the main grid, minimizing the cycle losses of energy storage units, and maintaining the voltage frequency within the allowable range. The decision variables of the optimization problem include the active / reactive setpoints of the photovoltaic generation units, the charge / discharge power curves of the energy storage units, and the switching status and power level of the adjustable loads.

[0062] Ensuring that system constraints are met is the core of the rolling optimization model. System constraints constitute the feasible region boundary of the optimization problem. Equipment operation constraints include the maximum allowable output limit of photovoltaic inverters, the maximum charging and discharging power limit of energy storage converters and their upper and lower limits of state of charge, and the power regulation range and minimum start-up and shutdown time of adjustable loads. Network security constraints include the thermal stability limit of line transmission power, the upper and lower limits of critical node voltage, and the allowable deviation range of system frequency. These constraints are embedded in the optimization model in the form of inequalities or equations. The power allocation scheme solved by the optimization algorithm must strictly satisfy all these constraints to ensure the feasibility of control commands and the safe and stable operation of the system. Solving this constrained optimization problem may employ mathematical programming methods such as linear programming, quadratic programming, or mixed integer programming, depending on the linearity of the model and whether discrete decision variables are included. The set values ​​for all time periods are aggregated to form a control instruction set. For a rolling optimization using a five-minute control interval, each solution typically generates a detailed power instruction sequence for the next five-minute period. It may also cover multiple future periods but only execute the instructions for the first period. The control instruction set is a structured data collection that contains the specific control commands that need to be issued to each controlled distributed generation unit and each controlled load unit in the next control period. Each instruction is marked with a clear timestamp, execution object identifier, and control parameter value. For example, it may instruct photovoltaic inverter A to maintain its active power output at 50 kW during time T0 to T0+300 seconds, instruct energy storage converter B to charge at 20 kW during time T0 to T0+150 seconds and stop working during time T0+150 seconds to T0+300 seconds, and instruct industrial load C to reduce its power consumption from 100 kW to 70 kW during time T0 to T0+300 seconds.

[0063] The generation and execution of control command sets completes the closed loop from decision-making to action. These commands are then distributed to distributed generation controllers and load controllers via a control network. This control network requires high reliability and low latency, typically employing industrial Ethernet or dedicated fiber optic networks to ensure timely and accurate command transmission. Upon receiving commands, the distributed generation controllers and load controllers drive their local power electronics or switching devices to perform corresponding operations. For example, photovoltaic inverters adjust their maximum power point tracking algorithm reference values, energy storage converters switch their operating modes and power levels, and load controllers disconnect or connect to the power circuit. The execution status of the commands is monitored in real time, and actual operating data is collected and compared with expected values. During command execution, sensors deployed in each unit continuously measure operating parameters such as actual power generation, actual load power, and actual system frequency. The central controller compares these actual measured values ​​with the expected setpoints in the control command set in real time and calculates the execution deviation. If the deviation exceeds the tolerance range, the process of acquiring data from multiple monitoring nodes of the microgrid within the historical operating range is restarted for a new round of control calculations. The tolerance range is preset according to the equipment control accuracy and system stability requirements. For example, the power deviation tolerance is ±2% of the command value, and the frequency deviation tolerance is ±0.05 Hz. When the actual operating value of one or more units is detected to continuously deviate from the command value and the deviation exceeds the tolerance range, it indicates that the current control interval or power allocation scheme may no longer be suitable for the actual dynamic changes of the system, or an unexpected disturbance has occurred. At this time, the control system will immediately interrupt the current execution cycle and trigger a re-optimization process, that is, reacquire the latest historical operating data, perform multi-resolution fluctuation analysis, and start the particle swarm optimization algorithm to recalculate the optimal control interval and the new power allocation scheme, thereby forming an adaptive control closed loop that can quickly respond to system mutations.

[0064] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic coordinated control method for microgrids of distributed energy resources, characterized in that, The method includes: The system acquires energy output datasets, energy demand datasets, and system frequency datasets from multiple monitoring nodes of the microgrid within its historical operating range; performs multi-resolution fluctuation analysis on the energy output datasets, energy demand datasets, and system frequency datasets to generate three fluctuation indices: output fluctuation index, demand fluctuation index, and frequency fluctuation index. Using the fluctuation index as a guiding variable, a swarm optimization algorithm is used to explore and locate the optimal control interval in the microgrid control parameter space. Based on the optimal control interval, the distributed generation units and load units in the microgrid are periodically adjusted to generate a set of control instructions and execute them to achieve dynamic coordinated control. Acquiring energy output datasets, energy demand datasets, and system frequency datasets from multiple monitoring nodes of a microgrid within a historical operating period includes: defining the historical operating period as a sliding time window, with the window size adaptively adjusted based on load change patterns and external environmental factors in the microgrid's operating history; continuously recording energy output datasets and energy demand datasets using power sensing devices deployed at distributed energy interfaces and consumption sensing devices deployed at load connection points; and collecting system frequency datasets through a system frequency monitor. Performing multi-resolution fluctuation analysis on the energy output dataset, energy demand dataset, and system frequency dataset includes: using dynamic mode decomposition technology to decompose each dataset into a series of dynamic mode components with different oscillation frequencies; calculating the energy distribution characteristics of each dynamic mode component; determining the contribution weight of each dynamic mode component based on the energy distribution characteristics; and weighting and fusing the fluctuation characteristics of all dynamic mode components according to the contribution weights to generate an output fluctuation index, a demand fluctuation index, and a frequency fluctuation index. Using the aforementioned fluctuation index as a guiding variable, the exploration in the microgrid control parameter space employs a swarm optimization algorithm, including: constructing the microgrid control parameter space as a multi-dimensional space, with dimensions corresponding to the output fluctuation index, demand fluctuation index, and frequency fluctuation index; initializing a particle swarm, where each particle represents a candidate solution for the control interval; calculating the fitness value of each particle, with the fitness function based on the control error evaluation of the control interval in the simulation environment; updating the particle position and velocity, iteratively searching until convergence, and outputting the optimal control interval. The periodic adjustment of distributed generation units and load units in the microgrid according to the optimal control interval includes: dividing the optimal control interval into continuous control periods, allocating power generation setpoints and load adjustment values ​​to each period; using a rolling optimization strategy to calculate the power allocation scheme for each period to ensure that system constraints are met; and summarizing the setpoints of all periods to form a control instruction set. The generation and execution of the control instruction set includes: distributing the control instruction set to the distributed generation controller and load controller through the control network; monitoring the execution status of the instructions in real time and comparing the actual operating data with the expected values; if the deviation exceeds the tolerance range, restarting the process of acquiring data from multiple monitoring nodes of the microgrid within the historical operating range and performing a new round of control calculations.

2. The microgrid dynamic coordination control method for distributed energy resources as described in claim 1, characterized in that, The dynamic mode decomposition technique is used to decompose each dataset into a series of dynamic mode components with different oscillation frequencies. This includes: constructing a time-shift matrix of the data sequence; performing singular value decomposition on the time-shift matrix to extract the main feature patterns; solving for the oscillation frequency and decay rate of the dynamic modes through eigenvalue decomposition; retaining the main dynamic mode components according to a preset mode selection threshold; reconstructing the retained dynamic mode components and verifying their fit with the original data.

3. The microgrid dynamic coordination control method for distributed energy resources as described in claim 2, characterized in that, Constructing the microgrid control parameter space into a multi-dimensional space includes: collecting multiple sets of sample output fluctuation indices, sample demand fluctuation indices, and sample frequency fluctuation indices from historical operation, as well as the corresponding multiple sets of sample control intervals as training samples; establishing a three-dimensional coordinate framework, where the X-axis corresponds to the output fluctuation index, the Y-axis corresponds to the demand fluctuation index, and the Z-axis corresponds to the frequency fluctuation index; mapping the training samples onto the three-dimensional coordinate framework to form a labeled point cloud structure, thus defining the microgrid control parameter space.

4. The microgrid dynamic coordination control method for distributed energy resources as described in claim 3, characterized in that, Initializing the particle swarm includes: randomly generating a certain number of particles, with the position vector of each particle randomly assigned within the microgrid control parameter space; setting the particle velocity range to a preset minimum and maximum value; and defining the swarm size and upper limit of the number of iterations.

5. The microgrid dynamic coordination control method for distributed energy resources as described in claim 4, characterized in that, The fitness value of each particle is calculated by: simulating the operation of the microgrid under the control interval corresponding to the particle, calculating the energy supply and demand balance deviation and frequency stability index; and combining the deviation and index with a weighted sum to obtain the fitness value, with the weights set according to the microgrid priority.

Images