Distributed power supply grid-connected control method based on edge computing

Abstract

The present application relates to the field of distributed power grid connection technology, and discloses a distributed power grid connection control method based on edge computing. The method comprises collecting the output current, voltage and frequency parameters of the distributed power supply and the grid voltage, frequency and power factor parameters of the grid connection point in real time through the edge computing node to form two types of data sets of power supply operation and grid state; performing preliminary abnormality analysis on the grid connection operation state based on the two types of data sets to generate a preliminary abnormality indication signal; starting a detailed evaluation process according to the signal, which includes power supply internal state and grid interaction state evaluation. The edge computing node performs two types of evaluation respectively, calculates the corresponding evaluation values, and determines the grid connection stability after fusion to generate grid connection maintenance or adjustment instructions; the active power and reactive power adjustment parameters are finally determined according to the instruction analysis control parameters to realize accurate control of the distributed power grid connection.

Description

Technical Field

[0001] This invention relates to the field of distributed power grid connection technology, specifically to a distributed power grid connection control method based on edge computing. Background Technology

[0002] With the rapid development of the new energy industry, distributed power sources such as photovoltaic power plants and small wind power plants are being used more and more widely. Distributed power sources are characterized by flexible installation and high energy efficiency, which can effectively supplement the shortcomings of traditional centralized power generation and help transform the energy structure. However, the large-scale integration of distributed power sources also brings new challenges to grid operation. Their output power is easily affected by factors such as natural conditions and fluctuates. If grid connection control is not proper, it may impact the stable operation of the grid and even cause problems such as deterioration of power quality and equipment damage.

[0003] Currently, grid-connected control of distributed power sources largely relies on traditional centralized control architectures. Under this architecture, data acquisition and control command transmission must traverse long-distance communication networks, resulting in significant delays in data processing. When the output parameters of distributed power sources or the grid state change rapidly, centralized control struggles to respond in real time and cannot adjust control strategies promptly, thus affecting grid stability. Furthermore, traditional control methods often have limitations in data acquisition, focusing primarily on collecting partial output parameters of distributed power sources or single state parameters of the grid. This results in incomplete data dimensions, making it difficult to comprehensively reflect the actual operation of the power source and the state of the grid.

[0004] In anomaly analysis and assessment, existing technologies often employ a single-dimensional judgment approach, lacking a comprehensive consideration of the internal state of the power source and its interaction with the grid. Some control methods focus only on whether the power source output parameters exceed thresholds, ignoring the impact of the operating status of internal power source components such as inverters and energy storage units on grid connection. Other methods, while considering the grid state, fail to deeply analyze the interaction between distributed power sources and the grid, such as harmonic interference and voltage fluctuation propagation during power exchange. This singular assessment model easily leads to inaccurate anomaly judgment, failing to identify potential grid connection risks in a timely manner, potentially causing misjudgments or omissions, and thus affecting the effectiveness of control commands.

[0005] In determining control parameters, traditional methods often rely on empirical values ​​or fixed models for parameter calculation, making it difficult to flexibly adjust parameters according to dynamic changes in power source operation and grid conditions. When the number of distributed power sources connected increases or the grid load fluctuates significantly, fixed control parameters cannot adapt to complex and changing operating scenarios. This may lead to insufficient accuracy in active and reactive power regulation, hindering the coordinated operation of distributed power sources and the grid. This not only affects the power generation efficiency of distributed power sources but also poses a threat to the safe and stable operation of the grid. Summary of the Invention

[0006] The purpose of this invention is to provide a distributed power grid-connected control method based on edge computing to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a distributed power grid-connected control method based on edge computing, the method comprising:

[0008] The distributed power source's output current, output voltage, and output frequency parameters are collected in real time through edge computing nodes. Simultaneously, the grid voltage, grid frequency, and grid power factor parameters at the grid connection points are collected to form a power source operation data set and a grid status data set. Based on the power source operation data set and the grid status data set, a preliminary anomaly analysis is performed on the grid-connected operation status of the distributed power source to generate a preliminary anomaly indication signal. A detailed evaluation process is initiated based on the preliminary anomaly indication signal.

[0009] The detailed evaluation process includes power source internal state evaluation and grid interaction state evaluation; the power source internal state evaluation is performed through edge computing nodes to obtain the operating status data of the internal components of the distributed power source and calculate the power source internal state evaluation value; the grid interaction state evaluation is performed through edge computing nodes to obtain the interaction impact data between the distributed power source and the grid and calculate the grid interaction state evaluation value; the power source internal state evaluation value and the grid interaction state evaluation value are fused to determine grid connection stability and generate grid connection maintenance instructions or grid connection adjustment instructions; control parameter analysis is performed based on the grid connection maintenance instructions or grid connection adjustment instructions to determine active power adjustment parameters and reactive power adjustment parameters.

[0010] Preferably, the specific process of the preliminary anomaly analysis is as follows:

[0011] Output voltage and frequency parameters are extracted from the power supply operation data set to obtain output voltage and frequency data sequences. The fluctuation amplitude values ​​of the output voltage and frequency data sequences are calculated, and the larger value is taken as the power supply output fluctuation assessment value. Grid voltage and frequency parameters are extracted from the grid status data set to obtain grid voltage and frequency data sequences. The absolute values ​​of the deviations between the grid voltage and frequency data sequences and the rated voltage and frequency data sequences are calculated, and the larger value is taken as the grid parameter deviation assessment value. Power supply output fluctuation thresholds and grid parameter deviation thresholds are set. The power supply output fluctuation assessment value is compared with the power supply output fluctuation threshold. When the power supply output fluctuation assessment value is greater than or equal to the power supply output fluctuation threshold, it is marked as a power supply output anomaly. The grid parameter deviation assessment value is compared with the grid parameter deviation threshold. When the grid parameter deviation assessment value is greater than or equal to the grid parameter deviation threshold, it is marked as a grid parameter anomaly. When a power supply output anomaly or a grid parameter anomaly occurs, a preliminary anomaly indication signal is generated.

[0012] Preferably, the specific process for assessing the internal state of the power supply is as follows:

[0013] The inverter and battery pack components of the distributed power source are monitored for status, acquiring temperature, efficiency, and switching frequency data of the inverter components, and voltage, current, and temperature data of the battery pack components. An inverter health index is calculated based on the temperature, efficiency, and switching frequency data of the inverter components; a battery health index is calculated based on the voltage, current, and temperature data of the battery pack components; and the inverter and battery health indices are weighted and fused to obtain an internal state assessment value for the power source.

[0014] Preferably, the specific process of the power grid interaction status assessment is as follows:

[0015] Harmonic current parameters and voltage flicker parameters injected into the grid by distributed generation are monitored to obtain harmonic current data sequences and voltage flicker data sequences. The total harmonic distortion rate (THR) value is extracted from the harmonic current data sequence, and the short-time flicker severity value is extracted from the voltage flicker data sequence. The THR value is compared with a preset harmonic limit to obtain the harmonic impact value. The short-time flicker severity value is compared with a preset flicker limit to obtain the flicker impact value. The harmonic impact value and flicker impact value are normalized to obtain the grid interaction state assessment value.

[0016] Preferably, the specific process for determining grid connection stability is as follows:

[0017] Set the weighting coefficients for the power source internal state assessment value and the grid interaction state assessment value. Multiply the power source internal state assessment value by the corresponding weighting coefficient, and multiply the grid interaction state assessment value by the corresponding weighting coefficient. Add the two products to obtain the grid connection stability index. Compare the grid connection stability index with the preset grid connection stability threshold. When the grid connection stability index is greater than or equal to the preset grid connection stability threshold, a grid connection maintenance command is generated. When the grid connection stability index is less than the preset grid connection stability threshold, a grid connection adjustment command is generated.

[0018] Preferably, the specific process of the control parameter analysis is as follows:

[0019] When a grid connection maintenance command is received, the current active power output parameters and reactive power output parameters of the distributed power source are kept unchanged. When a grid connection adjustment command is received, the active power adjustment value and reactive power adjustment value are calculated based on the power source internal state assessment value and the grid interaction state assessment value. The active power adjustment parameters are determined based on the active power adjustment value, and the reactive power adjustment parameters are determined based on the reactive power adjustment value.

[0020] Preferably, in the control parameter analysis, when a grid connection adjustment command is received, the specific process for calculating the active power adjustment value and the reactive power adjustment value is as follows:

[0021] Obtain the internal state assessment value of the power source and the interaction state assessment value of the power grid; query the preset power adjustment mapping table based on the internal state assessment value of the power source to determine the current allowable active power adjustment range and the current allowable reactive power adjustment range of the distributed power source; analyze the stability requirements of the power grid based on the interaction state assessment value of the power grid to determine the urgency of active power adjustment and reactive power adjustment; combine the current allowable adjustment range and urgency level, and calculate the active power adjustment value and reactive power adjustment value using a priority strategy.

[0022] Preferably, the edge computing node also implements data security caching and communication redundancy mechanisms:

[0023] During data acquisition, edge computing nodes cache power operation data sets and grid status data sets in local storage units in real time. When the communication link between the edge computing node and the distributed power source or grid monitoring equipment is interrupted, the node reads the most recently cached data from the local storage unit to continue the preliminary anomaly analysis and detailed evaluation process. At the same time, the edge computing node starts the backup communication module to attempt to restore the connection. When communication is restored, the data cached during the interruption is uploaded to the cloud data center in batches.

[0024] Preferably, the method further includes collaborative work among multiple edge computing nodes:

[0025] When multiple distributed power sources are connected to the grid to form a cluster, the edge computing nodes corresponding to each distributed power source communicate with each other through the local area network, exchanging their respective power source internal state assessment values ​​and grid interaction state assessment values; all edge computing nodes periodically elect a master node, which is responsible for summarizing all assessment values ​​and calculating the overall grid connection stability index of the cluster; based on the overall grid connection stability index of the cluster, it generates cluster-level grid connection adjustment instructions and coordinates the power adjustment parameters of each distributed power source.

[0026] Preferably, the method further includes a data preprocessing step:

[0027] After collecting power supply operation data sets and power grid status data sets, noise filtering and outlier removal are performed on the two data sets respectively to obtain purified power supply operation data sets and purified power grid status data sets; preliminary anomaly analysis is then performed using the purified data sets.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] Edge computing nodes can be deployed directly near distributed power sources or around grid connection points. They can collect the output current, output voltage, and output frequency parameters of distributed power sources in real time, as well as the grid voltage, grid frequency, and grid power factor parameters of grid connection points. This avoids the time consumption during long-distance data transmission, ensures the timeliness of data acquisition, and lays the foundation for subsequent grid connection status analysis and control command generation. This allows control strategies to respond quickly to changes in power source operation and grid status.

[0030] Regarding the comprehensiveness of data acquisition, this method simultaneously generates both a power source operation dataset and a grid status dataset, covering key parameters at both the distributed power source output end and the grid connection end. Compared to traditional methods that only collect partial parameters, this approach can more completely reflect the operational status of distributed power sources and the real-time state of the grid. Preliminary anomaly analysis based on these two datasets can initially identify abnormal signs in grid-connected operation from both power source and grid dimensions, reducing missed anomaly detections due to data gaps, improving the coverage and accuracy of preliminary anomaly analysis, and providing a reliable initial basis for subsequent detailed assessments.

[0031] In the detailed evaluation process, this method combines the assessment of the internal state of the power source with the assessment of the interaction state with the grid, achieving a multi-dimensional and in-depth analysis of the grid connection status. By acquiring the operational status data of the internal components of the distributed power source through edge computing nodes, it is possible to understand the working status of core components such as inverters and energy storage modules, and determine whether there are potential problems affecting grid connection. At the same time, by acquiring the interaction data between the distributed power source and the grid, it is possible to analyze the effects of power exchange, harmonic transmission, and other interaction processes on grid connection stability. This dual-dimensional evaluation mode breaks through the limitations of traditional single-assessment methods, enabling a more comprehensive understanding of the overall grid connection operation. The grid connection stability judgment based on the integration of the two types of evaluation values ​​can more accurately determine whether the current grid connection is stable, ensuring the rationality of grid connection maintenance or adjustment commands.

[0032] In the control parameter determination stage, this method analyzes control parameters based on grid connection maintenance or adjustment instructions, and specifically determines active power and reactive power adjustment parameters. This parameter determination method is based on real-time power source operation data and grid status data, as well as in-depth grid connection stability assessment results, and can adapt to different operating scenarios. When distributed power source output fluctuates or grid load changes, active power parameters can be adjusted to ensure that power source output matches grid load demand, and reactive power parameters can be adjusted to improve the grid power factor and optimize power quality. This process does not rely on fixed empirical values ​​or static models, and can dynamically optimize control parameters according to actual operating conditions, enabling distributed power sources to better integrate into the grid and achieve coordinated operation between the two, ensuring both normal power generation by distributed power sources and stable grid operation. Attached Figure Description

[0033] Figure 1 This is a schematic diagram illustrating the working principle of the distributed power grid-connected control method based on edge computing described in this invention.

[0034] Figure 2 A flowchart for assessing the interactive state of the power grid;

[0035] Figure 3 This is a flowchart for calculating the power adjustment amount. 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 1This invention provides a distributed power source grid-connected control method based on edge computing. The method includes: edge computing nodes collecting output current, output voltage, and output frequency parameters of the distributed power source in real time, and simultaneously collecting grid voltage, grid frequency, and grid power factor parameters at grid connection points, forming a power source operation data set and a grid status data set. Based on the power source operation data set and the grid status data set, a preliminary anomaly analysis is performed on the grid-connected operation status of the distributed power source, generating a preliminary anomaly indication signal. A detailed evaluation process is initiated based on the preliminary anomaly indication signal. The detailed evaluation process includes power source internal status evaluation and grid interaction status evaluation. The power source internal status evaluation is performed by the edge computing nodes to obtain the operating status data of the internal components of the distributed power source and calculate the power source internal status evaluation value. The grid interaction status evaluation is performed by the edge computing nodes to obtain the interaction impact data between the distributed power source and the grid and calculate the grid interaction status evaluation value. The power source internal status evaluation value and the grid interaction status evaluation value are fused to determine grid-connected stability and generate a grid-connected maintenance command or a grid-connected adjustment command. Based on the grid-connected maintenance command or the grid-connected adjustment command, control parameter analysis is performed to determine active power adjustment parameters and reactive power adjustment parameters. Edge computing nodes are deployed locally within the distributed power source, employing a microprocessor architecture to implement data acquisition, analysis, and control logic, ensuring low-latency response. Power source operation data sets and grid status data sets are acquired in real-time via sensor modules, with sampling frequencies configured according to grid standards. Preliminary anomaly analysis quickly triggers detailed assessments based on threshold comparisons, avoiding unnecessary computational load. Internal power source status assessment focuses on the health of core distributed power source components, while grid interaction status assessment emphasizes grid-connection compatibility. Grid-connection stability determination employs a weighted fusion strategy to balance internal and external factors. Control parameter analysis dynamically adjusts power output to maintain grid stability.

[0038] Example 1: Edge computing nodes extract output voltage and frequency parameters from the real-time power supply operation data set to construct output voltage and frequency data sequences. The fluctuation characteristics of the output voltage and frequency data sequences are quantified by calculating the fluctuation amplitude. The fluctuation amplitude is calculated using a sliding window variance analysis method, with the window size configured based on the type and dynamic response characteristics of the distributed power source. The power supply output fluctuation assessment value is selected as the larger of the output voltage and frequency fluctuation amplitude values; this principle aims to capture the most significant abnormal fluctuation characteristics. The processing of the grid status data set follows similar logic. Edge computing nodes extract grid voltage and frequency parameters to form grid voltage and frequency data sequences. The grid parameter deviation assessment value is obtained by calculating the absolute value of the deviation between the grid voltage data sequence and the rated voltage, and the absolute value of the deviation between the grid frequency data sequence and the rated frequency, again selecting the larger of the two. The power supply output fluctuation threshold and grid parameter deviation threshold are set with reference to relevant International Electrotechnical Commission (IEC) standards and grid operation specifications; the threshold values ​​are pre-stored in the non-volatile memory of the edge computing nodes. The edge computing node compares the real-time calculated power output fluctuation assessment value with the power output fluctuation threshold. When the power output fluctuation assessment value is greater than or equal to the power output fluctuation threshold, the logic processing unit of the edge computing node marks the power output as abnormal. The comparison between the grid parameter deviation assessment value and the grid parameter deviation threshold is performed simultaneously. If the grid parameter deviation assessment value is greater than or equal to the grid parameter deviation threshold, a grid parameter anomaly flag is triggered. The generation of either the power output anomaly flag or the grid parameter anomaly flag directly leads to the generation of a preliminary anomaly indication signal. This preliminary anomaly indication signal is a digital flag, and its state change initiates the subsequent detailed assessment process. The entire preliminary anomaly analysis process is completed in a closed loop within the edge computing node, with the analysis cycle synchronized with the data acquisition cycle to ensure real-time response.

[0039] Monitoring the internal components of a distributed power source is fundamental to assessing its internal condition. Edge computing nodes acquire temperature, efficiency, and switching frequency data from inverter components via a sensor network integrated within the distributed power source. Temperature data is obtained from thermocouples mounted on the inverter's power devices; efficiency data is calculated by the ratio of input power to output power; and switching frequency data is captured by a high-frequency sampling circuit. Battery pack component monitoring data includes voltage, current, and temperature data. Voltage and current data are acquired via high-precision analog-to-digital converters, while temperature data is provided by temperature sensors distributed within the battery modules. The inverter health index calculation integrates a temperature deviation coefficient, efficiency degradation rate, and switching frequency stability index. The temperature deviation coefficient is a normalized value of the difference between the real-time temperature and the rated operating temperature; the efficiency degradation rate is the ratio of current efficiency to initial efficiency; and switching frequency stability is measured by the standard deviation of frequency fluctuations. The calculation of the battery health index involves voltage equalization, internal resistance change rate, and temperature distribution uniformity. Voltage equalization is the maximum deviation of the voltage of individual cells within the battery pack. The internal resistance change rate is indirectly calculated using a DC internal resistance test method. Temperature distribution uniformity is represented by the variance of readings from multiple temperature sensors. Edge computing nodes perform weighted fusion calculations on the inverter health index and battery health index to obtain the power supply internal state assessment value. The weighting coefficients are pre-set based on the relative importance of the inverter and battery pack components in the distributed power system. The inverter health index typically has a higher weight than the battery health index because the inverter directly determines the reliability of power conversion. The weighted fusion calculation uses a linear weighted sum model, and the output range of the power supply internal state assessment value is standardized between 0 and 1. The power supply internal state assessment value reflects the comprehensive health status of the internal components of the distributed power supply in real time; a higher value indicates a more reliable internal state. The power supply internal state assessment value serves as a key input parameter in subsequent grid-connected stability determination, and the accuracy and real-time performance of its calculation results directly affect the generation of control commands. The data pathways within the edge computing nodes have been optimized to minimize pipeline processing latency from monitoring data collection and health index calculation to evaluation value fusion.

[0040] The preliminary anomaly analysis module and the power supply internal state assessment module work collaboratively within the edge computing node. The preliminary anomaly analysis module acts as a rapid screening step, while the power supply internal state assessment module provides in-depth state insights. Both modules share the same power supply operation data set and some hardware resources, such as the arithmetic logic unit and cache memory. The firmware of the edge computing node schedules the execution order of the two modules, with the preliminary anomaly analysis module having a higher execution priority. The power supply internal state assessment module starts after the preliminary anomaly indication signal is activated. This design achieves a reasonable allocation of computing resources, ensuring timely anomaly response while avoiding the power consumption burden of continuous deep computation. The local storage unit of the edge computing node caches the power supply operation data set and intermediate calculation results from the most recent several cycles for historical data tracing and diagnostic analysis.

[0041] Example 2: See Figure 2 Edge computing nodes continuously monitor harmonic current and voltage flicker parameters injected into the public power grid by distributed power sources. Harmonic current monitoring is achieved using high-frequency current transformers and a spectrum analysis unit, while voltage flicker monitoring utilizes high-precision voltage sensors and flicker calculation methods. The raw data obtained from the monitoring is preprocessed to form harmonic current and voltage flicker data sequences. These data sequences are stored in the memory of the edge computing nodes using a circular buffer, with the sequence length covering multiple fundamental frequency cycles of the power grid to ensure the completeness of the analysis. The total harmonic distortion (THD) value is extracted from the harmonic current data sequence. The calculation of the THD value strictly follows the measurement method defined in the international standard IEC61000-4-7. A discrete Fourier transform is performed on the acquired current signal to analyze harmonic components up to the 50th order, and the content of each harmonic and the THD are calculated according to the formula. Short-duration flicker severity values ​​are extracted from the voltage flicker data sequence. The calculation of short-duration flicker severity values ​​is based on the flicker meter model specified in the IEC61000-4-15 standard. This model simulates the human visual response to changes in incandescent lamp illuminance. The short-duration flicker severity values ​​are output through a series of signal processing steps, including square demodulation, bandpass filtering, weighting, and statistical evaluation.

[0042] Preset harmonic and flicker limits are pre-defined according to the grid guidelines and power quality standards at the grid connection point. These limits are stored as configuration parameters in the non-volatile memory of the edge computing nodes. The edge computing nodes compare the real-time calculated total harmonic distortion (THD) value with the preset harmonic limits. The harmonic impact value is obtained by dividing the THD value by the preset harmonic limits. When the THD value exceeds the preset harmonic limits, the harmonic impact value will be greater than 1. Similarly, the edge computing nodes compare the short-term flicker severity value with the preset flicker limit. The flicker impact value is obtained by dividing the short-term flicker severity value by the preset flicker limit. The harmonic and flicker impact values ​​are normalized using a linear scaling method to map the two impact values ​​to the range of 0 to 1. The final grid interaction state assessment value is the larger of these two normalized values. This approach aims to highlight the most severe grid interaction problems. The grid interaction status assessment value quantitatively describes the degree of negative impact of distributed generation on the power quality of the grid, and its value directly reflects the quality of grid connection compatibility.

[0043] The foundation of grid-connected stability assessment lies in the effective fusion of the power source's internal state assessment value and the grid interaction state assessment value. The power source's internal state assessment value, generated by an independent assessment module, reflects the health status of the distributed power source itself. An edge computing node maintains a weighting coefficient configuration table. The weighting coefficients of the power source's internal state assessment value and the grid interaction state assessment value are dynamically adjusted according to different phases and priority strategies of grid operation. During periods of grid vulnerability, the weighting coefficient of the grid interaction state assessment value is assigned a higher value. The fusion calculation process involves multiplying the power source's internal state assessment value by its corresponding weighting coefficient and the grid interaction state assessment value by its corresponding weighting coefficient, then algebraically summing these two products to obtain a scalarized grid-connected stability index. The numerical range of the grid-connected stability index is designed, typically set between 0 and 100, for easy intuitive understanding and threshold comparison.

[0044] The preset grid-connected stability threshold is a crucial threshold value, whose setting comprehensively considers the requirements for the safe and stable operation of the power grid, the regulation capability of distributed power sources, and historical operating data. The comparison logic unit embedded in the edge computing node continuously compares the calculated grid-connected stability index with the preset grid-connected stability threshold. When the grid-connected stability index is greater than or equal to the preset grid-connected stability threshold, the comparison logic unit outputs a high-level signal, which is interpreted as a grid-connected maintenance command, meaning that the current grid-connected state is within a stable and acceptable range, and no power adjustment is required. When the grid-connected stability index is less than the preset grid-connected stability threshold, the comparison logic unit outputs a low-level signal, which is interpreted as a grid-connected adjustment command. This grid-connected adjustment command triggers the subsequent control parameter analysis module to start calculating the necessary power adjustment parameters. The entire determination process has strict real-time requirements; from data acquisition and evaluation value calculation to the output of the stability determination result, it must be completed within milliseconds to adapt to the dynamic changes in the power grid. The hardware acceleration function of the edge computing node is used to ensure this high-speed computing requirement, for example, by using the built-in floating-point unit to accelerate weighted fusion calculations. The judgment results, along with relevant evaluation data, are recorded in the local log for subsequent event tracing and operational analysis, providing a data foundation for system optimization. The close collaboration between grid interaction status assessment and grid-connected stability determination constitutes the core decision-making mechanism for intelligent grid-connected control of distributed power sources, enabling edge computing nodes to autonomously respond to grid conditions in a timely and accurate manner.

[0045] Example 3: See Figure 3The control logic unit of the edge computing node receives grid-connection maintenance or grid-connection adjustment commands from the front-end processing module. These commands are transmitted via the internal bus in the form of digital signals. When the control logic unit parses a grid-connection maintenance command, it sends a hold signal to the inverter controller of the distributed power source. The current active and reactive power output parameters of the distributed power source remain unchanged, and the inverter controller continues to operate using the existing power setpoints, keeping the entire system in a steady-state operating mode. When the control logic unit parses a grid-connection adjustment command, it triggers a complex calculation process based on two important state variables: the power source's internal state assessment value and the grid interaction state assessment value. The power source's internal state assessment value reflects the health of the power source itself, while the grid interaction state assessment value characterizes the compatibility level between the power source and the grid connection. The calculation of the active power adjustment value requires comprehensive consideration of the distributed power source's own output capacity limitations and the grid's power regulation requirements, while the calculation of the reactive power adjustment value focuses on voltage support and reactive power balance. The calculation process calls the preset adjustment algorithm within the edge computing node. The algorithm determines the basic adjustment range by querying a preset power adjustment mapping table, and then calculates the specific adjustment value by combining it with the real-time evaluation value. The active power adjustment parameters are ultimately determined based on the active power adjustment value, and the reactive power adjustment parameters are determined based on the reactive power adjustment value. The determined parameters are sent to the inverter controller for execution through the communication interface, thereby completing one closed-loop control.

[0046] In specific scenarios where a grid connection adjustment command is received, the detailed process of calculating the active power adjustment value and the reactive power adjustment value is activated. Edge computing nodes retrieve the latest calculated power source internal state assessment value and grid interaction state assessment value from the data buffer. These assessment values ​​are rigorously calculated by the assessment module in the aforementioned embodiments. The power source internal state assessment value is used as an index to query a preset power adjustment mapping table. The power adjustment mapping table is a two-dimensional data table stored in non-volatile memory. Its row index corresponds to the discretized interval of the power source internal state assessment value, and the list items define the current allowable active power adjustment range and the current allowable reactive power adjustment range of the distributed power source under the corresponding state. The current allowable active power adjustment range is a closed interval, defining the limit value at which active power can be safely increased or decreased. The current allowable reactive power adjustment range defines the adjustment boundary of reactive power. These ranges are set based on the equipment nameplate parameters, real-time operating temperature, and historical performance degradation data of the distributed power source.

[0047] The grid interaction state assessment value is used to analyze the grid-side stability requirements. The assessment logic embedded in the edge computing node maps the grid interaction state assessment value to an urgency level for active power adjustment and a urgency level for reactive power adjustment. The urgency level is typically divided into three levels: low, medium, and high, each represented by a different numerical code. A high urgency level indicates that the grid urgently needs power support. The priority strategy is the core decision logic in the calculation process. This strategy stipulates that when both active and reactive power adjustment needs exist simultaneously, the power component with the most significant impact on grid stability is processed first. Combining the current allowable adjustment range and the urgency level, the calculation of the active power adjustment value follows the following relationship:

[0048] ;

[0049] in: This represents the active power adjustment value that needs to be calculated. It is a directional factor whose value is determined by the grid frequency deviation. A positive sign indicates that active power output needs to be increased, and a negative sign indicates that active power output needs to be reduced. It is a proportionality coefficient related to the type of distributed power source. This represents the quantified value indicating the urgency level of active power adjustment. This represents the radius of the currently permissible range of active power adjustment, which is half the width of the adjustment range. This represents the absolute value of the upper boundary of the currently permissible range for active power adjustment. This mathematical relationship ensures that the calculated active power adjustment value neither exceeds the safe operating boundaries of the equipment nor fails to respond appropriately to the urgency of grid demand.

[0050] Example 4: During data acquisition, the edge computing node writes the real-time power operation data set and grid status data set into its local storage unit. The local storage unit uses a non-volatile ferroelectric memory, organized as a circularly covered buffer. The buffer size is configured to store at least thirty consecutive minutes of operation data. The power operation data set, including output current, output voltage, and output frequency parameters, and the grid status data set, including grid voltage, grid frequency, and grid power factor parameters, are stored as timestamps and data value pairs. Each data entry is appended with a cyclic redundancy check (CRC) code to ensure data integrity. When the main communication link between the edge computing node and the sensor modules of the distributed power source or the monitoring equipment at the grid connection point is interrupted, the edge computing node's watchdog timer detects the loss of the communication heartbeat signal and triggers fault switching logic. The edge computing node immediately reads the most recently cached data block from the local storage unit's circular buffer. These data blocks have precise timestamps, allowing preliminary anomaly analysis and detailed evaluation processes to continue based on the latest valid data before the interruption. This mechanism ensures the continuity of control logic and avoids control blind spots caused by temporary communication failures. In parallel, the edge computing node activates its built-in backup communication module. This module supports multiple wireless communication protocols. Its activation process involves first scanning for available cellular network signals. If a 4G LTE signal strength exceeds a preset threshold, a connection is established preferentially. If the cellular network is unavailable, a switch to a satellite communication link is attempted. The connection restoration process employs an exponential backoff algorithm for retrying, avoiding channel congestion caused by frequent connection requests. Once the primary communication link is restored, the edge computing node adds sequence identifiers to all cached data accumulated in its local storage during the outage and packages it into data packets. These packets are then uploaded in batches to the cloud data center via the restored link for archiving and long-term analysis.

[0051] In a grid-connected cluster composed of multiple distributed power sources, each distributed power source's edge computing nodes communicate with each other through an independent industrial local area network (LAN). This LAN uses the TSN network protocol with deterministic latency. Each edge computing node periodically broadcasts its calculated internal power source state assessment and grid interaction state assessment to the LAN, while simultaneously receiving assessment values ​​broadcast by other nodes. The broadcast data frame format includes a node identifier, timestamp, assessment value, and data checksum. All edge computing nodes periodically elect a master node using a distributed election algorithm. The election algorithm is based on the node's computing capacity margin, network port status, and the time interval between its last time serving as master node. Refer to Table 1 for a detailed explanation of the evaluation dimensions and corresponding weights for master node election.

[0052] Table 1: Weighting Table of Master Node Election Evaluation Dimensions

[0053]

[0054] The elected master node is responsible for collecting the internal power supply status assessment values ​​and grid interaction status assessment values ​​reported by all edge computing nodes within the cluster. The master node performs a weighted average calculation on the collected assessment values ​​to obtain the overall grid-connected stability index of the cluster. The weight allocation in the weighted average calculation is based on the rated capacity of each distributed power source; the larger the capacity of the power source, the greater its impact on the overall cluster index. The overall grid-connected stability index of the cluster is compared with a preset cluster-level stability threshold, and a cluster-level grid-connected adjustment command is generated based on the comparison result. The cluster-level grid-connected adjustment command is a coordinated command for the entire cluster. The master node needs to decompose this macro-level command into specific active power adjustment parameters and reactive power adjustment parameters for each distributed power source. The allocation of power adjustment parameters follows the principle of equal incremental rate or the principle of allocation according to capacity ratio to achieve the optimal adjustment effect of the cluster as a whole. The master node distributes the calculated power adjustment parameters to the corresponding edge computing nodes for execution via the industrial LAN. After receiving the parameters from the master node, the edge computing nodes convert them into control commands for their local inverter controllers. The master node continuously monitors the cluster status and periodically re-executes the election process to prevent single points of failure and democratize cluster management. This multi-edge computing node collaborative working mode effectively improves the self-consistent operation capability and coordinated response level to grid fluctuations of the distributed power generation cluster.

[0055] Example 5: Taking a 50 kW rooftop photovoltaic grid-connected system as an example, its edge computing node collects power supply operation data sets and grid status data sets through voltage transformers and current transformers. The power supply operation data set includes the instantaneous values ​​of the three-phase output voltage and the instantaneous values ​​of the output current of the photovoltaic inverter, with a sampling frequency of 10 kHz. The grid status data set includes the effective value sequence of line voltage at grid connection points and the frequency measurement sequence, with a sampling frequency of 5 kHz. The raw data stream enters the data buffer of the edge computing node through an analog-to-digital converter. The data buffer is a first-in, first-out queue structure capable of temporarily storing high-speed sampling points for several consecutive seconds. The data preprocessing step is executed immediately after the data is retrieved from the buffer. Its core task is to improve data quality and provide clean input for preliminary anomaly analysis. Noise filtering is performed to handle high-frequency random interference in the power supply operation data set and the grid status data set. The edge computing node is configured with a digital filter bank to process different types of signals. For the instantaneous output voltage and current sequences, a Butterworth low-pass filter with a cutoff frequency of 2kHz is used for smoothing. This frequency is much higher than the grid fundamental frequency but effectively suppresses noise near the switching frequency. For the RMS voltage sequence, a moving average filter with a window width of 10 power frequency cycles is used to smooth out periodic minor fluctuations. The coefficients of the filtering algorithm are pre-stored in read-only memory and directly called during calculation. The filtering process is completed on a dedicated digital signal processing kernel to ensure real-time performance. After noise filtering, the data sequence exhibits effectively suppressed waveform spikes and random peaks, resulting in smoother data curves and facilitating the extraction of true characteristic trends.

[0056] Outlier removal follows noise filtering and aims to identify and remove erroneous data points that deviate significantly from the normal range due to momentary sensor malfunctions, electromagnetic pulse interference, or other accidental factors. Edge computing nodes employ the statistically based Laida criterion. For a continuous data window, such as one containing 100 sampling points, the algorithm first calculates the arithmetic mean and standard deviation of the data within the window. Any data point whose absolute deviation from the mean exceeds three times the standard deviation is marked as an outlier. Marked outliers are not directly deleted but replaced by a linear interpolation algorithm. This algorithm uses the values ​​of two normal data points before and after the outlier to calculate a reasonable interpolation value to fill the gap. This process maintains the temporal continuity of the data sequence, preventing timing errors in subsequent analysis due to missing data points. The data sequence after outlier removal exhibits more stable statistical characteristics, eliminating the misleading influence of individual extreme values ​​on the overall analysis results.

[0057] The purified power supply operation data set and the purified grid status data set are the final outputs of the data preprocessing step. These data are stored in another designated memory area on the edge computing node for the preliminary anomaly analysis module to read. The purified data set retains the temporal structure and physical meaning of the original data, but the data quality is significantly improved. The preliminary anomaly analysis module uses the purified data set for calculations, such as calculating fluctuation amplitude values ​​based on the purified output voltage data sequence and calculating absolute deviation values ​​based on the purified grid frequency data sequence. Because the input data has been purified, the thresholds set by the preliminary anomaly analysis module can be more sensitive, thus enabling earlier and more accurate detection of real anomalies and reducing the probability of false alarms.

[0058] 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 distributed power grid-connected control method based on edge computing, characterized in that, The method includes the following steps: The distributed power source's output current, output voltage, and output frequency parameters are collected in real time through edge computing nodes. Simultaneously, the grid voltage, grid frequency, and grid power factor parameters at the grid connection points are collected to form a power source operation data set and a grid status data set. Based on the power source operation data set and the grid status data set, a preliminary anomaly analysis is performed on the grid-connected operation status of the distributed power source to generate a preliminary anomaly indication signal. A detailed evaluation process is initiated based on the preliminary anomaly indication signal. The detailed evaluation process includes power source internal state evaluation and grid interaction state evaluation; the power source internal state evaluation is performed through edge computing nodes to obtain the operating status data of the internal components of the distributed power source and calculate the power source internal state evaluation value; the grid interaction state evaluation is performed through edge computing nodes to obtain the interaction impact data between the distributed power source and the grid and calculate the grid interaction state evaluation value; the power source internal state evaluation value and the grid interaction state evaluation value are fused to determine grid connection stability and generate grid connection maintenance instructions or grid connection adjustment instructions; control parameter analysis is performed based on the grid connection maintenance instructions or grid connection adjustment instructions to determine active power adjustment parameters and reactive power adjustment parameters; The specific process for assessing the internal state of the power supply is as follows: The system monitors the status of the inverter and battery pack components of the distributed power source, acquiring temperature, efficiency, and switching frequency data for the inverter components, and voltage, current, and temperature data for the battery pack components. Based on the inverter's temperature, efficiency, and switching frequency data, the system calculates the inverter health index; based on the battery pack's voltage, current, and temperature data, the system calculates the battery health index; finally, the system performs a weighted fusion calculation of the inverter and battery health indices to obtain the power source's internal status assessment value. The specific process for assessing the power grid interaction status is as follows: Harmonic current parameters and voltage flicker parameters injected into the grid by distributed generation are monitored to obtain harmonic current data sequences and voltage flicker data sequences. The total harmonic distortion rate (THR) value is extracted from the harmonic current data sequence, and the short-time flicker severity value is extracted from the voltage flicker data sequence. The THR value is compared with a preset harmonic limit to obtain the harmonic impact value. The short-time flicker severity value is compared with a preset flicker limit to obtain the flicker impact value. The harmonic impact value and flicker impact value are normalized to obtain the grid interaction state assessment value. The specific process for determining grid connection stability is as follows: Set the weighting coefficients for the power source internal state assessment value and the grid interaction state assessment value. Multiply the power source internal state assessment value by the corresponding weighting coefficients and the grid interaction state assessment value by the corresponding weighting coefficients. Add the two products to obtain the grid connection stability index. Compare the grid connection stability index with the preset grid connection stability threshold. When the grid connection stability index is greater than or equal to the preset grid connection stability threshold, a grid connection maintenance command is generated. When the grid connection stability index is less than the preset grid connection stability threshold, a grid connection adjustment command is generated. The specific process of the control parameter analysis is as follows: When a grid connection maintenance command is received, the current active power output parameters and reactive power output parameters of the distributed power source are kept unchanged. When a grid connection adjustment command is received, the active power adjustment value and reactive power adjustment value are calculated based on the power source internal state assessment value and the grid interaction state assessment value. The active power adjustment parameters are determined based on the active power adjustment value, and the reactive power adjustment parameters are determined based on the reactive power adjustment value. In the control parameter analysis, when a grid connection adjustment command is received, the specific process for calculating the active power adjustment value and the reactive power adjustment value is as follows: Obtain the internal state assessment value of the power source and the power grid interaction state assessment value; query the preset power adjustment mapping table based on the internal state assessment value of the power source to determine the current allowable active power adjustment range and the current allowable reactive power adjustment range of the distributed power source; analyze the stability requirements of the power grid based on the power grid interaction state assessment value to determine the urgency of active power adjustment and reactive power adjustment; combine the current allowable adjustment range and urgency level, and calculate the active power adjustment value and reactive power adjustment value using a priority strategy. The prioritization strategy is the core decision-making logic in the calculation process. It stipulates that when both active and reactive power adjustment needs exist simultaneously, the power component with the most significant impact on grid stability is processed first. Combining the current allowable adjustment range and the urgency level, the calculation of the active power adjustment value follows the following relationship: ； in, This represents the active power adjustment value to be calculated. It is a directional factor whose value is determined by the grid frequency deviation. A positive sign indicates that active power output needs to be increased, and a negative sign indicates that active power output needs to be decreased. It is a proportionality coefficient related to the type of distributed power source. The numerical value representing the urgency level of active power adjustment. This represents the radius of the currently permissible active power adjustment range, which is half the width of the adjustment range. This represents the absolute value of the upper boundary of the currently permissible range of active power adjustment.

2. The distributed power grid-connected control method based on edge computing according to claim 1, characterized in that, The specific process of the preliminary anomaly analysis is as follows: Output voltage and frequency parameters are extracted from the power supply operation data set to obtain output voltage and frequency data sequences. The fluctuation amplitude values ​​of the output voltage and frequency data sequences are calculated, and the larger value is taken as the power supply output fluctuation assessment value. Power grid voltage and frequency parameters are extracted from the power grid status data set to obtain power grid voltage and frequency data sequences. The absolute values ​​of the deviations between the power grid voltage and frequency data sequences and the rated voltage and frequency data sequences are calculated, and the larger value is taken as the power grid parameter deviation assessment value. Power supply output fluctuation thresholds and power grid parameter deviation thresholds are set. The power supply output fluctuation assessment value is compared with the power supply output fluctuation thresholds. When the power supply output fluctuation assessment value is greater than or equal to the power supply output fluctuation threshold, it is marked as a power supply output anomaly. The power grid parameter deviation assessment value is compared with the power grid parameter deviation threshold. When the power grid parameter deviation assessment value is greater than or equal to the power grid parameter deviation threshold, it is marked as a power grid parameter anomaly. When abnormal power output or abnormal grid parameters occur, a preliminary abnormality indication signal is generated.

3. The distributed power grid-connected control method based on edge computing according to claim 1, characterized in that, The edge computing node also implements data security caching and communication redundancy mechanisms: During data acquisition, edge computing nodes cache power operation data sets and grid status data sets in local storage units in real time. When the communication link between the edge computing node and the distributed power source or grid monitoring equipment is interrupted, the node reads the most recently cached data from the local storage unit to continue the preliminary anomaly analysis and detailed evaluation process. At the same time, the edge computing node starts the backup communication module to attempt to restore the connection. When communication is restored, the data cached during the interruption is uploaded to the cloud data center in batches.

4. The distributed power grid-connected control method based on edge computing according to claim 3, characterized in that, The method also includes collaborative work among multiple edge computing nodes: When multiple distributed power sources are connected to the grid to form a cluster, the edge computing nodes corresponding to each distributed power source communicate with each other through the local area network, exchanging their respective power source internal state assessment values ​​and grid interaction state assessment values; all edge computing nodes periodically elect a master node, which is responsible for summarizing all assessment values ​​and calculating the overall grid connection stability index of the cluster; based on the overall grid connection stability index of the cluster, it generates cluster-level grid connection adjustment instructions and coordinates the power adjustment parameters of each distributed power source.

5. The distributed power grid-connected control method based on edge computing according to claim 1, characterized in that, The method also includes a data preprocessing step: After collecting the power supply operation data set and the power grid status data set, noise filtering and outlier removal are performed on the two data sets respectively to obtain the purified power supply operation data set and the purified power grid status data set. Preliminary anomaly analysis was performed using the cleaned dataset.

Images