Micro-grid dynamic scheduling method and system based on edge calculation and electric fingerprint identification

Abstract

The invention relates to the technical field of micro-grid scheduling, and discloses a micro-grid dynamic scheduling method and system based on edge calculation and electric fingerprint identification. The method comprises the following steps: collecting current and voltage waveforms of a user total inlet, capturing a transient waveform data segment, and identifying the type and real-time power of an electric appliance; electric appliance types and real-time power are received and aggregated, and a load detail table is constructed; a schedulable load resource pool is constructed based on the load detail list, and the micro-grid central controller calculates a load regulation and control instruction according to the schedulable load resource pool; and the edge computing node selects an electric appliance from the load detail list to execute control and feeds back the cut-down power to the micro-grid central controller. According to the method, transient waveforms are collected in a non-intrusive mode at a user total entrance through the electric fingerprint identification technology, V-I track features, harmonic components and starting transient features are extracted for template matching, accurate identification of each electric appliance type and real-time power is achieved, and the technical problem that a traditional dispatching system can only obtain the total power is solved.

Description

Technical Field

[0001] This invention relates to the field of microgrid scheduling technology, and in particular to a dynamic scheduling method and system for microgrids based on edge computing and electronic fingerprint recognition. Background Technology

[0002] As an important carrier for integrating distributed photovoltaic, wind power and other renewable energy sources, the stable and economical operation of microgrids depends on accurate load forecasting and flexible dispatch strategies. However, traditional microgrid dispatch systems can only obtain power data from the user's total inlet and cannot identify the operating status and power characteristics of specific internal appliances, resulting in a lack of fine-grained load information to support dispatch decisions.

[0003] Existing technologies employ a centralized architecture, uploading all user terminal data to the cloud for processing. This not only creates significant communication bandwidth pressure and computational latency but also fails to meet the real-time scheduling requirements of microgrids at the second to minute level. Furthermore, due to a lack of accurate identification of user-side appliance types and dispatchability, the scheduling system cannot effectively manage interruptible or shiftable loads such as air conditioners, electric water heaters, and electric vehicle charging stations. This results in low demand-side response resource utilization, poor peak shaving and valley filling effects, and consequently impacts the economic viability and renewable energy absorption capacity of the microgrid. Summary of the Invention

[0004] The main objective of this invention is to provide a dynamic scheduling method and system for microgrids based on edge computing and electronic fingerprint recognition. This invention uses electronic fingerprint recognition technology to non-intrusively collect transient waveforms at the user's main inlet and extract VI trajectory features, harmonic components, and start-up transient features for template matching, thereby achieving accurate identification of each appliance type and real-time power, and solving the technical problem that traditional scheduling systems can only obtain total power.

[0005] To achieve the above objectives, this invention provides a dynamic scheduling method for microgrids based on edge computing and electronic fingerprinting, comprising the following steps: Collect the current and voltage waveforms of the user's main inlet and capture transient waveform data segments. Identify the appliance type and real-time power based on the transient waveform data segments. The edge computing node receives and aggregates the appliance type and the real-time power, and constructs a load detail table based on the appliance type; Based on the load details table, a schedulable load resource pool is constructed, and the microgrid central controller calculates load control instructions according to the schedulable load resource pool. The edge computing node selects electrical appliances from the load details table according to the load control command and performs control, and feeds back the power reduction to the microgrid central controller.

[0006] Optionally, in a first implementation of the first aspect of the present invention, acquiring the current and voltage waveforms of the user's main inlet and capturing transient waveform data segments, and identifying the appliance type and real-time power based on the transient waveform data segments, includes: The electronic fingerprint smart terminal collects the current and voltage waveforms at the user's main entrance and calculates the instantaneous active power of the current and voltage waveforms. The instantaneous active power is smoothed to obtain a smoothed power sequence, and the power difference between adjacent sliding windows in the smoothed power sequence is calculated. When the power difference exceeds the power threshold, it is determined as a load switching event, and the current and voltage waveforms before and after the load switching event are captured as transient waveform data segments. The VI trajectory features, harmonic components, and start-up transient features of the transient waveform data segment are extracted and matched with the electrical appliance feature template library to identify the electrical appliance type and real-time power.

[0007] Optionally, in a second implementation of the first aspect of the present invention, extracting the VI trajectory features, harmonic components, and startup transient features of the transient waveform data segment and matching them with an electrical appliance feature template library to identify the appliance type and real-time power includes: Extract VI trajectory features and harmonic components from the steady-state interval of the transient waveform data segment, and extract startup transient features from the transient interval of the transient waveform data segment; A comprehensive feature vector is constructed based on the VI trajectory features, the harmonic components, and the start-up transient features, and the cosine similarity between the comprehensive feature vector and each first electrical appliance template in the electrical appliance feature template library is calculated. The corresponding second electrical appliance template is determined based on the cosine similarity, and the appliance type and real-time power of the second electrical appliance template are obtained.

[0008] Optionally, in a third implementation of the first aspect of the present invention, the edge computing node receives and aggregates the appliance type and the real-time power, and constructs a load detail table based on the appliance type, including: The edge computing node receives the appliance type and real-time power uploaded by multiple electronic fingerprint smart terminals, parses and extracts the user identifier, the appliance type, the operating status and the real-time power and stores them in the load details table; The system queries the schedulability rule base corresponding to each type of appliance in the load details table, returns the query results, and labels air conditioners and electric water heaters with interruptible tags and electric vehicle charging piles with slewable tags based on the query results, and updates the load details table accordingly.

[0009] Optionally, in a fourth implementation of the first aspect of the present invention, a schedulable load resource pool is constructed based on the load detail table, and the microgrid central controller calculates load control instructions according to the schedulable load resource pool, including: Calculate the baseline load of appliances in operation as shown in the load details table, and generate a predicted load curve based on the baseline load; Iterate through the load details table, filter appliances with the dispatchability tag as interruptible and the operating status as running, and calculate the total interruptible load; The load details table is traversed to filter appliances with the schedulability tag as movable, a movable load list is generated, and a schedulable load resource pool is constructed based on the total interruptible load and the movable load list. The microgrid central controller calculates load control commands based on the dispatchable load resource pool.

[0010] Optionally, in a fifth implementation of the first aspect of the present invention, calculating the reference load of appliances in operation in the load details table, and generating a predicted load curve based on the reference load, includes: The load details table is traversed to filter the appliances that are in operation and the corresponding real-time power is accumulated to obtain the baseline load. The historical start probability of appliances that are not running is queried. The historical start probability is multiplied by the expected power of the corresponding appliance and then added to the baseline load to generate a predicted load curve.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, the microgrid central controller calculates load control instructions based on the schedulable load resource pool, including: The microgrid central controller receives the dispatchable load resource pool and the predicted load curve, collects distributed photovoltaic power output, energy storage system state of charge and upstream grid electricity price, and establishes an optimization model with the minimum microgrid operating cost as the objective function; The optimization model is set with power balance constraints, energy storage charging and discharging constraints and demand-side regulation constraints. The demand-side regulation constraints set the upper limit of the demand-side regulation amount to the total amount of interruptible load in the schedulable load resource pool, and set the total charging demand of the shiftable load to be completed within the parking time window. Solve the optimization model to obtain the optimal scheduling scheme, and extract the demand-side control quantity from the optimal scheduling scheme to generate load control instructions.

[0012] Optionally, in the seventh implementation of the first aspect of the present invention, solving the optimization model to obtain the optimal scheduling scheme, and extracting the demand-side control quantity from the optimal scheduling scheme to generate a load control instruction includes: The optimization model is solved using a mixed-integer linear programming solver to obtain the optimal scheduling scheme that includes energy storage charging and discharging power, power purchased from the upper-level grid, and demand-side regulation. Extract the demand-side control amount for each time period from the optimal scheduling scheme, and allocate the power reduction of each edge computing node according to the proportion of the demand-side control amount and the total interruptible load of each edge computing node. The power reduction and control period corresponding to each edge computing node are combined to generate a load control command containing the edge computing node identifier, control period, and power reduction, which is then sent to the corresponding edge computing node.

[0013] Optionally, in an eighth implementation of the first aspect of the present invention, the edge computing node selects electrical appliances from the load detail table according to the load control instruction and feeds back the power reduction to the microgrid central controller, including: The edge computing node receives load control instructions, parses the power reduction and control period, filters the appliances in the load details table whose schedulability tag is interruptible and whose running status is running, and sequentially selects the corresponding appliances to accumulate real-time power until the power reduction is reached. The system generates corresponding control commands for each appliance and sends them through the communication module, such as sending a command to increase the set temperature to the air conditioner, a command to cut off the power supply to the electric water heater, and a command to reduce the charging power to the charging pile. Record the first total load power at the start of the control period, collect the second total load power after the control command is executed, calculate the difference between the first total load power and the second total load power, and feed it back to the microgrid central controller.

[0014] This invention also provides a microgrid dynamic scheduling system based on edge computing and electronic fingerprint recognition, comprising: The acquisition module is used to acquire the current and voltage waveforms of the user's main inlet and capture transient waveform data segments, and identify the type of electrical appliance and real-time power based on the transient waveform data segments; A construction module is used for edge computing nodes to receive and aggregate the appliance type and the real-time power, and to construct a load detail table based on the appliance type; The calculation module is used to construct a schedulable load resource pool based on the load details table, and the microgrid central controller calculates load control instructions according to the schedulable load resource pool. The feedback module is used by the edge computing node to select electrical appliances from the load details table according to the load control instructions, perform control, and feed back the power reduction to the microgrid central controller.

[0015] In summary, this invention utilizes electronic fingerprinting technology to non-intrusively acquire transient waveforms at the user's main entry point and extract VI trajectory features, harmonic components, and startup transient features for template matching. This achieves accurate identification of each appliance type and real-time power, solving the technical problem of traditional dispatching systems that can only obtain total power. Edge computing nodes aggregate the identification results and label them with dispatchability tags according to appliance type to construct a load detail table. Localized processing significantly reduces data transmission volume and decision latency, enabling the system to respond to load fluctuations in real time. Based on the load detail table, ultra-short-term load forecasting is performed, and interruptible and shiftable loads are statistically analyzed to construct a dispatchable load resource pool, transforming previously invisible user-side resources into quantifiable dispatch variables. The microgrid central controller incorporates the dispatchable load resource pool into the demand-side control constraints of the optimization model, achieving coordinated optimization of source, grid, load, and storage, transforming demand-side resources from passive response to flexible resources actively participating in dispatch. Edge computing nodes select specific appliances from the load detail table according to priority based on load control commands and provide feedback on actual power reduction, ensuring the accessibility and effectiveness of dispatch commands and improving the microgrid's economy, reliability, and renewable energy absorption capacity. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the steps of a microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition in one embodiment of the present invention; Figure 2 This is a block diagram of a microgrid dynamic scheduling system based on edge computing and electronic fingerprint recognition in one embodiment of the present invention.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Reference Figure 1 This embodiment provides a dynamic scheduling method for microgrids based on edge computing and electronic fingerprint recognition, including the following steps: S1, collect the current and voltage waveforms of the user's main inlet and capture transient waveform data segments, and identify the type of electrical appliance and real-time power based on the transient waveform data segments; S2, the edge computing node receives and aggregates appliance types and real-time power, and builds a load detail table based on appliance types; S3, construct a dispatchable load resource pool based on the load details table, and the microgrid central controller calculates load control instructions according to the dispatchable load resource pool; S4, the edge computing node selects electrical appliances from the load details table according to the load control instructions, performs control, and feeds back the power reduction to the microgrid central controller.

[0020] In one example, the current and voltage waveforms at the user's main inlet are acquired and transient waveform data segments are captured. Based on these transient waveform data segments, the appliance type and real-time power are identified, including: The electronic fingerprint smart terminal collects the current and voltage waveforms at the user's main entrance and calculates the instantaneous active power of the current and voltage waveforms. The instantaneous active power is smoothed to obtain a smoothed power sequence, and the power difference between adjacent sliding windows in the smoothed power sequence is calculated. When the power difference exceeds the power threshold, it is identified as a load switching event, and the current and voltage waveforms before and after the load switching event are captured as transient waveform data segments. Extract the VI trajectory features, harmonic components, and start-up transient features of transient waveform data segments and match them with the electrical appliance feature template library to identify the appliance type and real-time power.

[0021] In this example, the electronic fingerprint smart terminal continuously monitors the current and voltage waveforms at a high sampling rate at the user's main power inlet. Based on each sampling period, it calculates the instantaneous active power sequence between current and voltage in real time by multiplying point by point, obtaining the original energy change signal. A moving average algorithm is applied to smooth the instantaneous active power sequence, and a smoothed power sequence is generated by sliding within a preset time window. The average power value of two adjacent sliding windows is selected in the power sequence, and the difference between them is calculated to construct the power change curve in the smoothed power sequence. When the power difference exceeds a set power threshold (e.g., 50W) at a certain moment, it is determined that a potential load switching event has occurred, and a transient waveform capture mechanism is immediately triggered to retain key feature information. In the triggering mechanism, the electronic fingerprint smart terminal automatically records the original current and voltage waveform data for a fixed time period before and after the power change point, covering the range from 200 milliseconds before the event to 800 milliseconds after the event, forming a one-second transient waveform data segment, which covers the dynamic response characteristics and steady-state operating characteristics during the appliance's start-up or shutdown process. Based on transient waveform data segments, a stable operating condition interval between 400 and 800 milliseconds after the event is selected. The correspondence between voltage and current waveforms within this interval is extracted to construct a VI trajectory curve. The VI trajectory curve is then normalized and sampled at equal intervals to form a stable VI trajectory feature vector, characterizing the linear or nonlinear load characteristics of the electrical appliance. A Fast Fourier Transform is performed on the current waveform within a complete power frequency cycle after the event to extract the amplitude and phase characteristics of the fundamental and higher-order harmonics, forming a harmonic component feature vector that reflects the power quality and power electronic characteristics of the electrical appliance. A startup transient feature extraction is performed on the waveform segment spanning a total of 400 milliseconds before and after the event, calculating indicators such as the ratio between the peak current and the mean steady-state current, the impact duration, and the current rise slope to describe the typical dynamic behavior of motor-type loads during switching. The terminal concatenates the VI trajectory, harmonic features, and transient features to form a multi-dimensional feature vector, and performs similarity calculations with various templates in the terminal's pre-built standard electrical appliance feature template library. By calculating the cosine similarity or other matching index between each template and the current feature vector, the template with the highest similarity score and exceeding the recognition threshold (e.g., 0.85) is selected as the recognition result, and the electrical appliance type, operating status (on or off) and real-time operating power corresponding to the load switching event are output.

[0022] In one example, the VI trajectory features, harmonic components, and startup transient features of transient waveform data segments are extracted and matched with an electrical appliance feature template library to identify the appliance type and real-time power, including: Extract VI trajectory features and harmonic components from the steady-state interval of the transient waveform data segment, and extract start-up transient features from the transient interval of the transient waveform data segment; A comprehensive feature vector is constructed based on the VI trajectory features, harmonic components, and start-up transient features, and the cosine similarity between the comprehensive feature vector and each first electrical appliance template in the electrical appliance feature template library is calculated. The corresponding second electrical appliance template is determined based on cosine similarity, and the appliance type and real-time power of the second electrical appliance template are obtained.

[0023] In this example, the transient waveform data segment after the load switching event is structurally divided. The latter half of the stable operating range is designated as the steady-state segment, and the transition phase before and after the sudden event is designated as the transient segment. Based on this, feature information of different dimensions is extracted to construct an electrical fingerprint description. In the steady-state range, a VI trajectory is established with the voltage waveform as the horizontal axis and the current waveform as the vertical axis. By normalizing and equidistantly sampling the voltage and current sampling points within a complete power frequency cycle, the VI curve is transformed into a set of structurally stable and measurable trajectory coordinates, effectively reflecting the electrical properties of the load, such as linearity, phase characteristics, and nonlinear response. Simultaneously, a Fast Fourier Transform (FFT) operation is performed on the current signal in the steady-state range to extract the amplitude and phase angle of the fundamental component and the 2nd to 9th harmonics, forming a harmonic component vector to characterize the frequency domain characteristics and potential power factor characteristics of the electrical appliance during operation. The transient region, approximately 200 milliseconds before and after the load jump in the transient waveform data segment, is located. Within this region, indicators such as the peak amplitude of the starting current, the duration of the impact, and the rate of current rise per unit time are identified to extract dynamic response features reflecting the starting behavior of the appliance. The VI trajectory features, harmonic features, and starting transient features are sequentially concatenated to construct a comprehensive feature vector. This comprehensive feature vector is then compared with each first appliance template in a pre-stored appliance feature template library on the local terminal using cosine similarity calculations. The degree of similarity is determined by measuring the cosine of the angle between the two vectors. After calculating the similarity scores of all templates, the set of templates with the highest similarity is selected as the second appliance template. If its similarity score exceeds a set recognition threshold, a successful match is determined, and the corresponding appliance type and standard real-time power output value are extracted from the second appliance template.

[0024] In one example, the edge computing node receives and aggregates appliance type and real-time power, and constructs a load detail table based on appliance type, including: Edge computing nodes receive appliance types and real-time power data uploaded by multiple electronic fingerprint smart terminals, parse and extract user identifiers, appliance types, operating statuses, and real-time power data, and store them in a load detail table; Query the schedulability rule base corresponding to each type of appliance in the load details table, return the query results, and label air conditioners and electric water heaters with interruptible tags and electric vehicle charging piles with slewable tags according to the query results, and update the load details table.

[0025] In this example, edge computing nodes continuously receive periodic identification result data packets from multiple electronic fingerprint smart terminals within their jurisdiction via wired or wireless communication links. Each data packet contains a unique timestamp of the appliance, the corresponding user identifier, the appliance type, the current operating status, and real-time power information. The edge computing node parses the structure of the identification result data packets, extracting key fields, including the user identifier, appliance type, operating status (e.g., on or off), and the current real-time power value. A unique key value is generated based on the combination of "user identifier + appliance type" to retrieve and update the corresponding entry in the load details table. If the appliance entry does not yet exist in the load details table, a new record is added using the key value as an index, and the currently parsed attribute data is filled in. If the corresponding entry already exists, its operating status and real-time power fields are updated to maintain dynamic consistency of the load data. Simultaneously, the edge node executes a schedulable identification process. The node has a built-in schedulability rule base, which is indexed by appliance type and pre-sets schedulability tags and priority labels for each type of appliance. For example, "air conditioner" and "electric water heater" are identified as "interruptible" loads because their short-term power interruption has a controllable impact on user experience; "electric vehicle charging pile" is identified as a "movable" load because its operation has a certain degree of delay and power adjustment; while devices such as "lighting" and "refrigerator" are identified as "uninterruptible" loads because their shutdown would directly affect user safety or basic daily needs. After parsing each identification result data, the edge node submits the appliance type field to the schedulability rule base for querying. If a match is found in the rule base, the corresponding schedulability tag and its scheduling priority are returned. Based on the query results, the node appends and updates the "schedulability tag" and "priority" fields to the corresponding record in the load details table, setting the tag value to "interruptible," "movable," or "uninterruptible."

[0026] After labeling air conditioners and electric water heaters with interruptible tags and electric vehicle charging piles with movable tags based on the query results and updating the load details table, the process also includes: edge computing nodes calculating the continuous running time of appliances labeled as interruptible in the load details table from startup to the current moment; when the continuous running time exceeds a preset running threshold, the interruptible priority of the corresponding appliance is reduced; edge computing nodes determining whether the continuous running time of electric water heaters labeled as interruptible in the load details table is less than the heating time threshold; when it is less than the heating time threshold, the schedulability tag of the corresponding electric water heater is modified to temporarily uninterruptible; edge computing nodes recording the last controlled timestamp of each interruptible appliance in the load details table, calculating the time interval between the current moment and the last controlled timestamp as the control interval duration and storing it in the load details table; and edge computing nodes generating a comprehensive control priority for interruptible appliances in the load details table based on the interruptible priority and the control interval duration, with the comprehensive control priority increasing as the control interval duration increases.

[0027] In one example, a dispatchable load resource pool is constructed based on a load detail table. The microgrid central controller calculates load control instructions based on the dispatchable load resource pool, including: Calculate the baseline load of appliances in operation in the load details table, and generate a predicted load curve based on the baseline load; Iterate through the load details table, filter appliances with the dispatchability tag as interruptible and the running status as running, and calculate the total interruptible load. Iterate through the load details table, filter appliances with the dispatchability tag "movable", generate a list of movable loads, and construct a dispatchable load resource pool based on the total interruptible load and the list of movable loads. The microgrid central controller calculates load control commands based on the dispatchable load resource pool.

[0028] In this example, the edge computing node analyzes the current operating status of various electrical appliances based on a real-time maintained load detail table, selects all appliances with the operating status marked as "running," extracts their corresponding real-time power values, and sums these real-time power values ​​over time to form the baseline load value for the current moment. Assuming that the operating status remains unchanged for a short period—that is, the behavior of each appliance maintaining its current power level without abrupt changes in the next 5 to 30 minutes—a baseline load curve is constructed based on this assumption, forming the first-layer deterministic load prediction model output. To construct a load resource description for regulation optimization, after the baseline load is constructed, two types of filtering operations are performed on the same load detail table to extract schedulable load information. Iterate through all records and filter out the set of appliances whose schedulability tag is marked "interruptible" and whose operating status is still "running". Accumulate the real-time power values ​​of all appliances in this set to calculate the total interruptible load in the current area. Iterate through the load details table and filter out appliances whose schedulability tag is "movable". Regardless of whether they are currently running, extract their rated power, average single usage duration, most recent usage time, and typical operating periods in a structured manner to construct a movable load list. Each list entry contains a load power value and a corresponding adjustable time window, forming [(P j , Δt j A structured list set is used to describe flexible load resources with time adjustability. The total amount of interruptible load and the list of shiftable loads are integrated into a scheduleable load resource pool for the current time window. When the central controller receives scheduleable load resource pool data uploaded from multiple edge nodes, it inputs the scheduleable load resource pool and synchronously collected global operating parameters such as distributed power output, energy storage charge status, and upstream grid electricity price information into a multi-objective optimization model. Under the constraints of power balance, cost minimization, and demand response feasibility, a centralized optimization solution is performed, and specific load control instructions are generated for each edge node. The load control instructions include power reduction targets, control types (interruption or shift), and control time periods.

[0029] In one example, the baseline load of appliances in the operating status in the load details table is calculated, and a predicted load curve is generated based on the baseline load, including: Iterate through the load details table, filter out appliances that are in operation, and accumulate the corresponding real-time power to obtain the baseline load; The historical start probability of appliances that are not running is queried. The historical start probability is multiplied by the expected power of the corresponding appliance and then summed and added to the baseline load to generate a predicted load curve.

[0030] In this example, the current operating status information of each appliance is extracted from the real-time maintained load details table. All entries are traversed to filter out the set of appliances with an operating status of "Running". The real-time power values ​​of each appliance in this set are accumulated item by item to calculate the baseline load value at the current moment. The baseline load reflects the continuous distribution of the total power of appliances without considering new switching events and serves as the deterministic basis curve for the prediction model. To introduce probabilistic corrections for possible new load startup behaviors in the future, a historical behavior query operation is performed on all appliances with an operating status of "Not Running" in the load details table. A database of appliance startup probabilities based on users' daily usage habits is maintained locally in advance, recording the historical probability distribution of startup events of various appliances in different time periods. The edge node determines the prediction start time based on the current system clock and queries the historical startup probability P of each "Not Running" appliance within the current time period (e.g., 18:00 to 18:15). k Then, extract the expected starting power value P from the electrical characteristic data. ek Multiplying the two results in the electrical appliance's contribution to future load during that period, P. k × P ek After performing probability product calculations on all "non-operational" appliances and summing them up one by one, the total probability-corrected load value is obtained. The probability-corrected load value is then superimposed on the baseline load to construct a predicted load curve covering future time periods.

[0031] After multiplying the historical start-up probabilities by the expected power of the corresponding appliances and summing them to the baseline load to generate the predicted load curve, the process also includes: edge computing nodes recording the predicted load curve and the actual load curve for each load prediction period, and calculating the prediction error between the predicted load curve and the actual load curve; edge computing nodes statistically analyzing the prediction error distribution of each appliance type within a preset time period, identifying target appliance types whose prediction errors exceed the error threshold, and extracting the actual start-up time and runtime of the target appliance type within the preset time period; recalculating the start-up probability distribution of the target appliance type in different time periods based on the actual start-up time and runtime, and updating the historical appliance usage data with the recalculated start-up probability distribution to replace the original start-up probability; edge computing nodes using the updated historical appliance usage data to perform subsequent ultra-short-term load prediction, and continuously monitoring the prediction error to achieve adaptive iterative optimization of the prediction model.

[0032] In one example, the microgrid central controller calculates load control instructions based on the dispatchable load resource pool, including: The microgrid central controller receives the dispatchable load resource pool and the predicted load curve, collects the output of distributed photovoltaic power, the state of charge of the energy storage system and the electricity price of the upstream grid, and establishes an optimization model with the minimum operating cost of the microgrid as the objective function. To optimize the model, power balance constraints, energy storage charging and discharging constraints, and demand-side regulation constraints are set. The upper limit of the demand-side regulation amount is set to the total amount of interruptible load in the schedulable load resource pool, and the total charging demand of the shiftable load is set to be completed within the parking time window. Solve the optimization model to obtain the optimal scheduling scheme, and extract the demand-side control quantity from the optimal scheduling scheme to generate load control instructions.

[0033] In this example, the microgrid central controller periodically receives dispatchable load resource pool information and predicted load curve data uploaded by each edge computing node. The dispatchable load resource pool includes the total interruptible load and the list of loads that can be moved reported by each node, while the predicted load curve reflects the trend of total load changes in the short term. Simultaneously, the central controller collects current output data of the distributed photovoltaic power generation system, real-time state of charge (SOC) of the energy storage system, maximum charge / discharge power constraints, and time-period electricity price curves from the upstream grid. The central controller constructs an optimization model with the objective function of minimizing the microgrid operating cost. The objective function consists of three parts: the integral term of the product of the upstream grid's purchased power and electricity price, the product of the diesel generator output and fuel price, and the negative term of the product of the demand-side response and the incentive subsidy price, reflecting the economic contribution of each regulatory action. To ensure the model is solvable and the scheduling results are operable, the central controller sets multiple operational constraints on the optimization model. Among these, the power balance constraint ensures that, at any given scheduling time, the predicted total load should be met by distributed power generation output, energy storage system discharge, grid power purchase, and demand-side reduction. The energy storage system constraint specifies that its charging and discharging power limits are restricted by the equipment's rated parameters, and requires that the State of Charge (SOC) at the end of the scheduling cycle not be lower than the minimum operating threshold. Regarding demand response constraints, the central controller uses the total interruptible load as the maximum upper limit of the demand-side response for that period, ensuring that the regulated electrical load does not exceed the controllable capacity reported by the edge nodes. Furthermore, for electric vehicle loads belonging to the movable category, a time-coupled constraint is introduced, requiring all total charging demand to be completed within their physical parking time window to prevent peak-shifting delays from causing users to fail to meet charging standards. After completing the optimization modeling, the central controller calls a mixed-integer linear programming solver to numerically solve the model, with the solution process controlled within 30 seconds to meet the real-time requirements of scheduling. After the solution is completed, the controller obtains a globally optimal set of scheduling strategies, which includes the charging and discharging power time series of the energy storage system during the scheduling cycle, the planned power curve of electricity purchased from the grid, and the demand-side control amount allocated to each edge node. In the output results, the central controller extracts the optimal control command amount for each region for the demand side, and combines it with the originally reported list of interruptible and shiftable loads to generate load control commands. The command format includes the target node number, the start and end time of the control period, the required power reduction value, and the control method (interruption or shift). It then sends these commands to the corresponding edge nodes through the communication network to drive the actual execution at the appliance level.

[0034] To optimize the model, power balance constraints, energy storage charging and discharging constraints, and demand-side regulation constraints are set. The demand-side regulation constraint sets the upper limit of the demand-side regulation amount to the total interruptible load in the dispatchable load resource pool, and sets the total charging demand of the transferable load to be completed within the parking time window. This includes: the microgrid central controller extracting the total charging demand, parking time window, and rated charging power of electric vehicle charging piles from the transferable load list in the dispatchable load resource pool; dividing the parking time window into multiple scheduling periods, setting a charging power decision variable for each scheduling period, with the value of the charging power decision variable ranging from zero to the rated charging power; setting a total charging demand constraint, requiring that the sum of the products of the charging power decision variables of all scheduling periods and the corresponding period length equals the total charging demand; introducing a charging cost term into the objective function of the optimization model, which is the sum of the products of the charging power decision variables of each scheduling period and the corresponding period's upstream grid electricity price; and obtaining the optimal charging power for each scheduling period by solving the optimization model, generating a charging scheduling plan that includes the charging period and the corresponding charging power.

[0035] In one example, the optimization model is solved to obtain the optimal scheduling scheme, and demand-side control quantities are extracted from the optimal scheduling scheme to generate load control instructions, including: The optimization model is solved by a mixed-integer linear programming solver, and the optimal scheduling scheme including energy storage charging and discharging power, power purchased from the upper-level grid and demand-side regulation is obtained. Extract the demand-side control amount for each time period from the optimal scheduling scheme, and allocate the power reduction of each edge computing node according to the proportion of the demand-side control amount and the total interruptible load of each edge computing node. The power reduction and control period corresponding to each edge computing node are combined to generate a load control command containing the edge computing node identifier, control period, and power reduction, which is then sent to the corresponding edge computing node.

[0036] In this example, a mixed-integer linear programming solver is used to perform periodic global optimization calculations on the model. The solver supports the joint participation of continuous variables (such as energy storage power) and integer variables (such as the switching states of loads that can be shifted) in the calculation, and seeks to minimize the operating cost of the microgrid under multiple constraints and parallel conditions. After completing the model solution within the set time limit, the solver outputs a set of time-series optimal scheduling variables, including: the charging and discharging power curve of the energy storage system, the power purchase plan curve interacting with the upper-level grid, and the total demand-side control amount in each control cycle. The total demand-side control amount represents the total power reduction that can be flexibly adjusted by the user load side while maintaining system balance and economy. The total control volume is allocated to each subordinate edge computing node. To this end, the total interruptible load reported by each edge node during the specified time period is extracted. A control weighting factor is constructed based on the proportion of each node's interruptible resources in the global scheduleable load. The total central control volume is then proportionally divided, and the power reduction allocated to each node is calculated as: the total demand-side control power multiplied by the ratio of the current node's interruptible load to the total interruptible load of the entire network. This yields the specific power reduction target for each edge computing node. The central controller combines the power allocation results with the control time window information to construct a load control instruction data packet. Each instruction includes a unique identifier for the edge computing node, the corresponding control start and end time periods, and the allocated power reduction value. After data packet encapsulation, the central controller synchronously sends all instructions to each edge computing node via the microgrid communication link. Upon receiving the control task, the node executes the control strategy according to the allocated reduction target, calling its local load details table. During control execution, the node continuously monitors the load response and feeds it back to the central controller.

[0037] After combining the power reduction and control period corresponding to each edge computing node to generate a load control instruction containing the edge computing node identifier, control period, and power reduction, and issuing it to the corresponding edge computing node, the process also includes: after receiving the load control instruction, the edge computing node queries the total interruptible load in its local load details table to determine whether the total interruptible load is less than the power reduction in the load control instruction; when the total interruptible load is less than the power reduction, it calculates the resource gap, which is equal to the difference between the power reduction and the total interruptible load, and the edge computing node reports the resource gap to the microgrid central controller; after receiving the resource gap, the microgrid central controller queries the schedulable load resource pool of other edge computing nodes, selects at least one edge computing node from other edge computing nodes whose total interruptible load is greater than zero, and allocates the resource gap according to the proportion of the total interruptible load of each edge computing node; the microgrid central controller generates a coordinated control instruction containing the allocated power reduction and issues it to the other selected edge computing nodes, and the edge computing nodes and other edge computing nodes synchronously execute the control instruction to complete the coordinated reduction.

[0038] In one example, edge computing nodes select appliances from a load detail table based on load regulation instructions, perform control, and feed back the power reduction to the microgrid central controller, including: The edge computing node receives load control instructions, parses the power reduction and control period, filters the appliances in the load details table whose schedulability tag is interruptible and whose running status is running, and selects the corresponding appliances in turn to accumulate the real-time power until the power reduction is achieved. The system generates corresponding control commands for each appliance and sends them through the communication module, such as sending a command to increase the set temperature to the air conditioner, a command to cut off the power supply to the electric water heater, and a command to reduce the charging power to the charging pile. Record the first total load power at the start of the control period, collect the second total load power after the control command is executed, calculate the difference between the first total load power and the second total load power, and feed it back to the microgrid central controller.

[0039] In this example, before the scheduling cycle begins, the edge computing node receives a load control command from the microgrid central controller and performs structured parsing of the command content, extracting key fields, including the target power reduction value and the corresponding control start and end time periods. The edge computing node accesses the locally maintained load detail table, filters out the set of appliances whose current operating status is "running" and whose scheduleability tag is marked as "interruptible," forming a candidate appliance set. Based on a preset control priority strategy (such as the appliance's continuous operating time, the time interval of the last control, or the control level set by the user), the candidate appliances are sorted, and appliances are selected sequentially from the sorted results, accumulating their current real-time power values ​​until the sum first reaches or slightly exceeds the target power reduction value. After selecting candidate appliances, the edge nodes match corresponding control strategies based on appliance type and automatically generate appliance-level control commands. For air conditioners, a "raise set temperature" control command is constructed, increasing the set temperature by 1 to 2 degrees Celsius via communication protocol to reduce the compressor's duty cycle. For electric water heaters, a "power off" command is generated, directly cutting off the device's power supply via a smart socket control loop for complete shutdown. For electric vehicle charging stations, the system sends a "reduce charging power" command, lowering the current charging current to 50% or below the rated power to slow down charging and release instantaneous power demand. These control commands are sent to each target appliance via the communication module integrated into the edge nodes, using various protocol stacks such as Zigbee, WiFi, PLC, or 485 bus. Simultaneously, the edge computing nodes immediately read the current total load power at the start of the control period as the first total load value, and after a set sampling delay following the execution of all control commands, read the real-time total load of the system again as the second total load value. The difference between the first and second total load powers is calculated to obtain the actual load reduction effect. The actual power reduction value, along with indicators such as control response delay and equipment response success rate, is encapsulated into a feedback data structure and uploaded to the microgrid central controller through a dedicated data channel. After receiving the feedback data, the central controller compares the deviation between the control target value and the actual response result. If the deviation exceeds a set threshold, the central controller dynamically calls upon backup control resources or activates the energy storage system to compensate for the difference within the control cycle.

[0040] The process involves: selecting appliances in the load detail table with the schedulability tag "interruptible" and an "operating" status; sequentially selecting the corresponding appliances and accumulating their real-time power until the power reduction target is reached; the edge computing node selecting appliances in the load detail table with the schedulability tag "interruptible" and an "operating" status to form a candidate appliance set; extracting the comprehensive control priority and control interval duration of each appliance in the candidate appliance set; sorting the candidate appliance set in descending order of comprehensive control priority to obtain a first sorting result; traversing the first sorting result and sorting appliances with the same comprehensive control priority in ascending order of control interval duration to obtain a second sorting result; sequentially selecting appliances from the second sorting result and accumulating the real-time power of the corresponding appliances to obtain the cumulative power reduction target; determining whether the cumulative power reduction target has been reached; stopping the selection when it has been reached and forming a control object list of the selected appliances; the edge computing node recording the current control timestamp for each appliance in the control object list and updating the previous control timestamp of the corresponding appliance in the load detail table to the current control timestamp.

[0041] After calculating the difference between the first total load power and the second total load power and feeding it back to the microgrid central controller, the process also includes: the microgrid central controller receiving the difference fed back by the edge computing nodes as the actual power reduction, calculating the reduction deviation between the power reduction in the load control command and the actual power reduction; determining whether the absolute value of the reduction deviation exceeds the deviation threshold, and if the absolute value of the reduction deviation exceeds the deviation threshold, determining that the control effect is not up to standard and calculating the compensation power, which is equal to the difference between the power reduction and the actual power reduction; the microgrid central controller determining the sign of the compensation power, and if the compensation power is positive, generating a secondary control command containing the compensation power and sending it to the edge computing nodes or calling the energy storage system to discharge and compensate for the compensation power; after receiving the secondary control command, the edge computing nodes select electrical appliances from the interruptible electrical appliances that were not initially controlled in the load details table to perform compensation control, or the microgrid central controller sends a discharge command to the energy storage system to make the energy storage system output compensation power.

[0042] Reference Figure 2 This embodiment provides a microgrid dynamic scheduling system based on edge computing and electronic fingerprint recognition, including: Acquisition module 1 is used to acquire the current and voltage waveforms of the user's main inlet and capture transient waveform data segments, and identify the type of electrical appliance and real-time power based on the transient waveform data segments; Module 2 is used by edge computing nodes to receive and aggregate appliance types and real-time power, and to build a load detail table based on appliance types; Calculation module 3 is used to construct a schedulable load resource pool based on the load details table, and the microgrid central controller calculates load control instructions according to the schedulable load resource pool; Feedback module 4 is used by edge computing nodes to select electrical appliances from the load details table according to load control instructions, execute control, and feed back the power reduction to the microgrid central controller.

[0043] In this embodiment, the specific implementation of each unit in the above system embodiment is described in the above method embodiment, and will not be repeated here.

[0044] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, system, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, system, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, system, article, or method that includes that element.

[0045] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A micro-grid dynamic scheduling method based on edge computing and electrical fingerprinting, characterized in that, include: Collect the current and voltage waveforms of the user's main inlet and capture transient waveform data segments. Identify the appliance type and real-time power based on the transient waveform data segments. The edge computing node receives and aggregates the appliance type and the real-time power, and constructs a load detail table based on the appliance type; Based on the load details table, a schedulable load resource pool is constructed, and the microgrid central controller calculates load control instructions according to the schedulable load resource pool. The edge computing node selects electrical appliances from the load details table according to the load control command and performs control, and feeds back the power reduction to the microgrid central controller. 2.The microgrid dynamic scheduling method based on edge computing and electric fingerprint identification according to claim 1, wherein, Acquire the current and voltage waveforms at the user's main inlet and capture transient waveform data segments. Based on the transient waveform data segments, identify the appliance type and real-time power, including: The electronic fingerprint smart terminal collects the current and voltage waveforms at the user's main entrance and calculates the instantaneous active power of the current and voltage waveforms. The instantaneous active power is smoothed to obtain a smoothed power sequence, and the power difference between adjacent sliding windows in the smoothed power sequence is calculated. When the power difference exceeds the power threshold, it is determined as a load switching event, and the current and voltage waveforms before and after the load switching event are captured as transient waveform data segments. The VI trajectory features, harmonic components, and start-up transient features of the transient waveform data segment are extracted and matched with the electrical appliance feature template library to identify the electrical appliance type and real-time power.

3. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 2, characterized in that, Extract the VI trajectory features, harmonic components, and startup transient features of the transient waveform data segment and match them with the electrical appliance feature template library to identify the appliance type and real-time power, including: Extract VI trajectory features and harmonic components from the steady-state interval of the transient waveform data segment, and extract startup transient features from the transient interval of the transient waveform data segment; A comprehensive feature vector is constructed based on the VI trajectory features, the harmonic components, and the start-up transient features, and the cosine similarity between the comprehensive feature vector and each first electrical appliance template in the electrical appliance feature template library is calculated. The corresponding second electrical appliance template is determined based on the cosine similarity, and the appliance type and real-time power of the second electrical appliance template are obtained.

4. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 1, characterized in that, The edge computing node receives and aggregates the appliance type and the real-time power, and constructs a load detail table based on the appliance type, including: The edge computing node receives the appliance type and real-time power uploaded by multiple electronic fingerprint smart terminals, parses and extracts the user identifier, the appliance type, the operating status and the real-time power and stores them in the load details table; The system queries the schedulability rule base corresponding to each type of appliance in the load details table, returns the query results, and labels air conditioners and electric water heaters with interruptible tags and electric vehicle charging piles with movable tags based on the query results, and updates the load details table accordingly.

5. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 4, characterized in that, Based on the load details table, a dispatchable load resource pool is constructed. The microgrid central controller calculates load control instructions according to the dispatchable load resource pool, including: Calculate the baseline load of appliances in operation as shown in the load details table, and generate a predicted load curve based on the baseline load; Iterate through the load details table, filter appliances with the dispatchability tag as interruptible and the operating status as running, and calculate the total interruptible load; The load details table is traversed to filter appliances with the schedulability tag as movable, a movable load list is generated, and a schedulable load resource pool is constructed based on the total interruptible load and the movable load list. The microgrid central controller calculates load control commands based on the dispatchable load resource pool.

6. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 5, characterized in that, Calculate the baseline load of appliances in operation as shown in the load details table, and generate a predicted load curve based on the baseline load, including: The load details table is traversed to filter the appliances that are in operation and the corresponding real-time power is accumulated to obtain the baseline load. The historical start probability of appliances that are not running is queried. The historical start probability is multiplied by the expected power of the corresponding appliance and then added to the baseline load to generate a predicted load curve.

7. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 5, characterized in that, The microgrid central controller calculates load control instructions based on the dispatchable load resource pool, including: The microgrid central controller receives the dispatchable load resource pool and the predicted load curve, collects distributed photovoltaic power output, energy storage system state of charge and upstream grid electricity price, and establishes an optimization model with the minimum microgrid operating cost as the objective function; The optimization model is set with power balance constraints, energy storage charging and discharging constraints and demand-side regulation constraints. The demand-side regulation constraints set the upper limit of the demand-side regulation amount to the total amount of interruptible load in the schedulable load resource pool, and set the total charging demand of the shiftable load to be completed within the parking time window. Solve the optimization model to obtain the optimal scheduling scheme, and extract the demand-side control quantity from the optimal scheduling scheme to generate load control instructions.

8. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 7, characterized in that, Solving the optimization model yields the optimal scheduling scheme, and demand-side control parameters are extracted from the optimal scheduling scheme to generate load control instructions, including: The optimization model is solved using a mixed-integer linear programming solver to obtain the optimal scheduling scheme that includes energy storage charging and discharging power, power purchased from the upper-level grid, and demand-side regulation. Extract the demand-side control amount for each time period from the optimal scheduling scheme, and allocate the power reduction of each edge computing node according to the proportion of the demand-side control amount and the total interruptible load of each edge computing node. The power reduction and control period corresponding to each edge computing node are combined to generate a load control command containing the edge computing node identifier, control period, and power reduction, which is then sent to the corresponding edge computing node.

9. The microgrid dynamic scheduling method based on edge computing and electronic fingerprint recognition according to claim 8, characterized in that, The edge computing node selects electrical appliances from the load details table according to the load control command, performs control, and feeds back the power reduction to the microgrid central controller, including: The edge computing node receives load control instructions, parses the power reduction and control period, filters the appliances in the load details table whose schedulability tag is interruptible and whose running status is running, and sequentially selects the corresponding appliances to accumulate real-time power until the power reduction is reached. The system generates corresponding control commands for each appliance and sends them through the communication module, such as sending a command to increase the set temperature to the air conditioner, a command to cut off the power supply to the electric water heater, and a command to reduce the charging power to the charging pile. Record the first total load power at the start of the control period, collect the second total load power after the control command is executed, calculate the difference between the first total load power and the second total load power, and feed it back to the microgrid central controller.

10. A microgrid dynamic scheduling system based on edge computing and electronic fingerprint recognition, characterized in that, The steps for implementing the microgrid dynamic scheduling method based on edge computing and electrical fingerprinting as described in any one of claims 1 to 9 include: The acquisition module is used to acquire the current and voltage waveforms of the user's main inlet and capture transient waveform data segments, and identify the type of electrical appliance and real-time power based on the transient waveform data segments; A construction module is used for edge computing nodes to receive and aggregate the appliance type and the real-time power, and to construct a load detail table based on the appliance type; The calculation module is used to construct a schedulable load resource pool based on the load details table, and the microgrid central controller calculates load control instructions according to the schedulable load resource pool. The feedback module is used by the edge computing node to select electrical appliances from the load details table according to the load control instructions, perform control, and feed back the power reduction to the microgrid central controller.

Images