A dynamic load regulation method and system based on real-time power monitoring

Abstract

The application belongs to the technical field of power system regulation, and particularly relates to a dynamic load regulation method and system based on real-time power monitoring. By deploying edge collection terminals to synchronously collect electrical parameters of each node of a power grid at a high frequency, an adaptive dynamic threshold regulation model is constructed. Multidimensional influence factors such as meteorological environment, power consumption period, and equipment operating state are included in dynamic threshold calculation to realize adaptive iterative updating of the threshold. Meanwhile, terminal loads are finely classified, a three-class differentiated regulation strategy library of core guarantee loads, flexibly adjustable loads, and short-time excisable loads is established, and regulation parameters are continuously optimized in combination with a closed-loop feedback correction mechanism. By adopting the above technical solution, the application can realize real-time, accuracy and flexibility of distribution network load regulation, and significantly improve the operation stability and power efficiency of the distribution network.

Description

Technical Field

[0001] This invention belongs to the field of power system control technology, and particularly relates to a dynamic load control method and system based on real-time power monitoring. Background Technology

[0002] With the deepening of the global energy transition, distributed photovoltaic and distributed energy storage devices are rapidly developing and being connected to the grid on a large scale. At the same time, diversified load types, such as industrial flexible loads, commercial building loads, and residential electricity loads, continue to grow. Modern power distribution networks are increasingly exhibiting complex operating characteristics, including strong random fluctuations, significant intermittency, and widening peak-to-valley differences. Against this backdrop, traditional static and fixed load control models are no longer effectively adapted to the current complex power grid operating conditions. As a core component for the safe and stable operation and energy efficiency optimization of the power distribution network, the real-time performance, accuracy, and flexibility of load control directly determine the power supply reliability, renewable energy absorption capacity, and user electricity economy of the power distribution network.

[0003] However, existing power distribution network load control technologies generally suffer from numerous deep-seated technical defects in practical applications, severely restricting the further improvement of the power grid's refined dispatching level. Specifically, firstly, existing power monitoring systems mostly adopt a cloud-based periodic batch acquisition mode, which has a low data sampling frequency, high transmission delay, and generally suffers from technical bottlenecks such as data redundancy and the inability to effectively identify interference data. This makes it difficult to achieve millisecond-level, high-precision real-time power status perception, and thus cannot promptly capture instantaneous changes in power grid conditions such as sudden load increases and sharp drops in renewable energy output. This results in a serious time lag in control decisions, which can easily lead to a series of safety hazards such as power grid imbalance, line overload, and voltage fluctuations in actual operation.

[0004] Secondly, traditional load control schemes all use fixed power thresholds and preset static control rules for scheduling. These thresholds cannot be adaptively updated based on real-time grid conditions, electricity consumption periods, environmental factors, and equipment operating status. Faced with complex scenarios of intermittent fluctuations in renewable energy output and random load changes, static control rules are prone to over-control, under-control, and control failure, exhibiting extremely poor adaptability and making it difficult to achieve dynamic balance control of grid power.

[0005] Furthermore, most existing load control methods adopt a rigid, "one-size-fits-all" control strategy, failing to finely classify and categorize terminal loads. This makes it impossible to effectively distinguish between core guaranteed loads, flexibly adjustable loads, and non-core, switchable loads. Under emergency conditions of grid peak load overload and power exceeding limits, this can easily lead to the indiscriminate disconnection of important loads and unwarranted interference with core production and residential power supply, failing to balance grid operational safety with user power reliability. At the same time, traditional control models often focus on single-load-side regulation, lacking a multi-dimensional coordinated dispatch mechanism involving power sources, grid, load, and storage. This prevents the use of energy storage charging and discharging characteristics to offset power fluctuations, resulting in low local absorption rates of new energy sources, ineffective mitigation of grid peak-valley differences, and overall poor energy efficiency.

[0006] Therefore, existing load control technologies suffer from several technical defects, including poor real-time monitoring, static and rigid control rules, low precision in load control, insufficient coordination between power sources, grids, loads, and storage, lack of closed-loop feedback correction mechanisms, weak fault tolerance and anti-interference capabilities, and low overall energy efficiency. This invention aims to address these technical defects by providing a dynamic load control method and system based on real-time power monitoring. Summary of the Invention

[0007] The purpose of this invention is to provide a dynamic load control method and system based on real-time power monitoring, so as to solve the technical problems existing in the prior art.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: Firstly, a dynamic load control method based on real-time power monitoring is provided, comprising the following steps: S1: By deploying edge acquisition terminals on each branch circuit and terminal load device of the target power supply network, high-frequency synchronous acquisition of multi-dimensional raw electrical parameters such as voltage, current, active power, reactive power and power factor of each terminal load is performed; preprocessing operations including local noise reduction filtering, abnormal data identification and compliance calibration are performed on the raw electrical parameters, and standardized real-time power data is output. S2: Based on the preset historical operating sequence data of the equipment, coupled with multi-dimensional influencing factors including real-time meteorological environment parameters, electricity consumption time type parameters, and real-time operating status parameters of the equipment, the power grid power safety threshold and load controllability margin are dynamically updated in real time, and an adaptive dynamic threshold control model adapted to the current power grid operating conditions is constructed and updated online. S3: Establish standardized load classification rules in advance based on load power consumption rigidity coefficient, power supply importance level, power adjustment range, and operation delay tolerance parameters. Classify all terminal loads in the target power supply network according to the load classification rules, and classify each terminal load into any one of the following: core guaranteed load, flexibly adjustable load, and short-term disconnectable load. Configure corresponding control logic for the operation characteristics of the three types of loads respectively, and build and store a differentiated load control strategy library that can be called in real time. S4: Import standardized real-time power data into the adaptive dynamic threshold control model, determine the current operating status of the power grid through model comparison, and accurately identify abnormal power imbalance conditions; combine the identified power grid operating condition type, match and call the pre-set differentiated load control strategy library in S3, and generate preliminary load control instructions; S5: Execute the initial load control command and synchronously collect the actual operating data of the entire network after the command is executed in real time; compare the collected actual operating data with the preset optimal control target parameters to obtain the real-time control deviation value; based on the control deviation value, reverse correct the decision parameters of the adaptive dynamic threshold control model and the load control execution intensity, iteratively optimize to obtain the precise control command, and complete the single control closed-loop correction; S6: Using the corrected and precise control results as the benchmark for the new round of control, the entire process steps from S1 to S5 are executed iteratively in a loop, continuously updating real-time power data, adaptive dynamic threshold control model, power grid condition identification results and control instructions, so as to achieve continuous dynamic and flexible control of terminal load.

[0009] Preferably, the specific process of step S2 is as follows: S21: Calculation of the power grid baseline safety threshold (time-series baseline threshold): Based on historical operating data of the distribution network during the same period, a baseline safety threshold under undisturbed steady-state conditions is constructed as a benchmark value for dynamic correction. The calculation formula is as follows: ; in, The basic power safety threshold of the power grid is the reference threshold without real-time operating condition correction. The average active power for the same historical period is taken from the equipment's historical operating sequence database and serves as the baseline power for steady-state operation of the distribution network. The rated allowable active power of the distribution network branch / transformer is a fixed parameter from the equipment factory and the power grid ledger. These are the historical time series weighting coefficients and the equipment rated constraint weighting coefficients, respectively, satisfying... The value range is [0,1], and it is calibrated according to the distribution network operation procedure. S22: Calculate the quantitative correction coefficients for multi-dimensional influencing factors: Construct quantitative correction coefficients for three core influencing factors: real-time meteorological environment, electricity consumption period type, and equipment operating status, including a meteorological environment correction coefficient. Electricity usage period correction factor and equipment operating status correction coefficient This enables dynamic adaptation and correction of the basic threshold, with all factors being quantified and calculated based on real-time collected data. S23: Calculate the adaptive dynamic power safety threshold: By integrating multi-dimensional correction factors, the basic power safety threshold of the power grid is dynamically weighted and corrected to obtain a dynamic power safety threshold that adapts to the current real-time operating conditions, providing a dynamic benchmark for power limit exceeding judgment of the power grid: ; in, The real-time dynamic power safety threshold is the core output parameter of this model and is updated iteratively in real time. S24: Calculation of controllable load margin: Based on the difference between real-time operating power and dynamic power safety threshold, the current dispatchable and adjustable power margin of the power grid is quantitatively calculated to guide the allocation of subsequent load tiered regulation efforts. ; in, Real-time load controllable margin, unit: kW; This indicates that the power grid has a power surplus and can accommodate the increased load. This indicates that the power grid has exceeded its limit, and load regulation needs to be initiated. The standardized real-time active power output of S1; Real-time dynamic power safety threshold; S25: Model Temporal Iterative Update Mechanism: To ensure the model continuously adapts to changes in operating conditions, a temporal iterative refresh mechanism is adopted, which updates the model parameters in real time based on the latest collected data. The formula is as follows: ; in, The latest dynamic power safety threshold at time t; The historical dynamic power safety threshold at time t-1 retains the temporal memory characteristics; : Temporal smoothing iteration coefficient, with values ​​[0,1], to avoid threshold abrupt oscillations and improve model stability; The base threshold and correction coefficients are calculated in real time at time t.

[0010] Preferably, the specific process of step S22 is as follows: Calculate meteorological and environmental correction factors The impact of temperature, sunlight, wind, rain, and weather on the output of new energy sources and the power consumption of the load: ; in, This is a comprehensive correction factor for meteorological environment; Real-time ambient temperature and reference steady-state temperature; These are real-time illumination intensity and baseline standard illumination intensity, respectively. These are the temperature correction factor and the illumination correction factor, respectively, which are empirical calibration parameters for the meteorological-power response of the power system. Calculate the correction factor for electricity consumption period Distinguish the differences in load electricity consumption characteristics during peak, flat, and valley periods; ; in, Correction factor for electricity usage period; These are fixed correction coefficients for peak, normal, and off-peak periods, calibrated based on the time-series characteristics of electricity consumption in the regional distribution network, to meet... Peak periods are defined as the continuous range in which the total active power of the entire network in real time is greater than 120% of the average active power of the entire day; normal periods are defined as the continuous range in which the total active power of the entire network in real time is between 80% and 120% of the average active power of the entire day; and low periods are defined as the continuous range in which the total active power of the entire network in real time is lower than 80% of the average active power of the entire day.

[0011] Equipment operating status correction factor : Adapt to the impact of fluctuations in equipment load rate and operational health status; ; in, This is a correction factor for the equipment's operating status. The real-time load rate of the equipment is calculated from the real-time power collected by S1. ; The standardized real-time active power output by S1; This is the load attenuation correction factor, with a value range of (0,1).

[0012] Preferably, the specific process of step S3 is as follows: S31: Normalize the parameters of power consumption rigidity coefficient, power supply importance level, power adjustment range, and operating delay tolerance to obtain the load. Normalized power consumption rigidity coefficient ,load Normalized importance weights ,load Normalized power adjustable amplitude margin coefficient and load Normalized runtime delay tolerance coefficient ; Original electrical rigidity coefficient The value range is 1 (extremely flexible load) to 10 (fully rigid load, power cannot be interrupted), representing the electrical rigidity coefficient. The normalization formula is as follows: ; in, :load Normalized electricity rigidity coefficient The larger the value, the stronger the electrical rigidity, and it cannot be adjusted or disconnected. :load Original power consumption rigidity coefficient; The minimum rigidity of the total network load electricity consumption can be set to 1. The maximum rigid load power consumption of the entire network can be set to 10. Original load importance level The power supply importance levels are categorized as follows: 1 (Ordinary Load), 2 (Secondary Important Load), 3 (Primary Important Load), and 4 (Special Grade Guarantee Load). The normalization formula for these importance levels is as follows: ; in, :load Normalized importance weights The higher the value, the higher the priority of power supply protection; :load Original power supply importance level; The minimum load importance level is fixed at 1. The maximum value for load importance level is fixed at 4; load The formula for calculating the normalized power adjustability margin coefficient is as follows: ; in, :load Normalized power adjustable amplitude margin coefficient The larger the value, the greater the range of adjustable load power. For load The original maximum adjustable active power corresponding to the power adjustment range For load Rated active power; load The formula for calculating the normalized runtime delay tolerance coefficient is as follows: ; in, :load Normalized runtime delay tolerance coefficient The higher the value, the more prohibited short-term power outages are for the load. :load Original permissible short-term power outage duration; The maximum tolerable power outage time for the entire network load can be set to 2 hours. S32: Construct a standardized load comprehensive assessment model to perform batch load classification across the entire network, and automatically label each load category. A weighted comprehensive scoring model is used. S33: Execute differentiated load control logic modeling: for Based on the operational characteristics, establish the active power regulation objective function and constraints respectively, which serve as the underlying logic of the regulation strategy: S34: Differentiated load balancing strategy for storage based on a hierarchical storage structure with a strategy library. The hierarchical storage structure of the strategy library includes: Bottom layer: Standardized hierarchical rule base: stores normalization formulas for four types of parameters and comprehensive load scores. S i The calculation formula and grading threshold standards are used to achieve real-time automatic classification of the entire network load based on the original parameters of S2. Middle layer: Differentiated load control model library: stores separately L 1. L 2. L 3. Objective functions, constraints, and parameter boundaries specific to the three types of loads; Top-level: Scenario-based scheduling strategy instance library: Based on the bottom-level hierarchy + middle-level model, it generates strategies that can be directly called in multiple scenarios, including scenario strategies for steady-state normal operating conditions, mild power imbalance operating conditions, and severe power imbalance operating conditions.

[0013] Preferably, the specific process of step S32 is as follows: First, the comprehensive load assurance score is calculated based on the standardized comprehensive load assessment model, using the following formula: ; in, :load The higher the comprehensive protection score, the higher the priority of power supply protection and the stronger the constraint against adjustment or disconnection. The rules for setting the tiered thresholds are as follows: like :load Classified as core backup load ; like :load Classified as flexibly adjustable load ; like :load Classified as load that can be cut off in a short time ; in, : Core protection load classification threshold; The threshold for short-term load shedding is set according to the power grid emergency dispatch standards.

[0014] Finally, perform batch classification of the entire network load: Traverse all terminal loads of the target power supply network ; Normalization calculations are performed sequentially on a single load to obtain standardized derived parameters. ; Substitute into the comprehensive scoring formula to calculate ; Automatically label load categories by comparing grading thresholds. ; Output load classification ledger: load number, rated power, four-dimensional original parameters, normalized parameters, comprehensive score, load category; Preferably, the specific process of step S33 is as follows: Core protection load Control logic (prioritize power supply, prohibit load reduction / disconnection): Control objectives: Maintain the rated operating status of core guaranteed loads, minimize active power fluctuations, and strictly prohibit load reduction, power outages, and disconnections; Objective function: ; Constraints: ; in, The objective function is to minimize the deviation of the core guaranteed load power. :load Real-time active power; Core load power upper and lower limits For load The rated active power; The allowable power outage duration for a single load interruption is set to 0 for core loads, and any short-term disconnection operations are prohibited. Flexible load adjustment Control logic (power reduction allowed, load shedding strictly prohibited): Control objective: Under power deficit conditions, maximize the reduction of flexible load power within the safe operating range of equipment, ensuring uninterrupted power supply and no load shedding throughout the process; Objective function: ; Constraints:

[0015] in, The objective function for maximizing the total adjustable power of a flexibly adjustable load; :load Actual reduction in active power; The minimum safe operating power for flexible loads; operating below this value will cause equipment failure. Flexible load adjustment prohibits short-term power outages; The load can be removed in a short time. Control logic (orderly short-term disconnection when power deficit exceeds the limit): Control objective: When the power deficit of the power grid exceeds the maximum adjustment capacity of the flexible load, the loads that can be disconnected are disconnected in an orderly short-term manner according to the load priority, and the actual power outage duration is strictly controlled to not exceed the allowable threshold.

[0016] Objective function: ; Constraints: ; in, The objective function is to maximize the total power cut-off for loads that can be cut off in a short time. : Load removal of 0-1 decision variables ; :load Actual single power outage duration (h, derived variable, distinguishing from original tolerance duration) ); Real-time total active power deficit of the power grid (kW, global dynamic parameter of the power grid). The total maximum adjustable power of the entire network's flexible loads is calculated from the preceding original parameters.

[0017] Preferably, the specific process of step S4 is as follows: S41: Obtain standardized real-time power data, including power grid data. Real-time total active power of the entire network Total rated installed active power of the power grid power grid Real-time active power deficit and load exist Real-time operation of active power ; S42: Construction and updating of the adaptive dynamic threshold model: The adaptive dynamic threshold model uses a power imbalance baseline threshold. +Time-series fluctuation correction +Adaptive dynamic correction coefficient A three-layer progressive correction structure is used to obtain the final adaptive dynamic threshold. ; S43: Perform grid operation status comparison and power imbalance anomaly identification: identify the real-time power deficit of the grid. With adaptive dynamic threshold By comparing and classifying operating conditions into three categories according to power grid dispatching industry standards, the operating status of the power grid can be accurately identified: Rules for determining the operating status of power grids: Steady-state normal operating conditions: This means that the real-time power deviation of the power grid is within the dynamic allowable range, the power supply and demand are balanced, there is no need for regulation, and the load of the entire grid maintains the current rated operating state. Mild power imbalance condition: This means that the power grid has a small power deficit and a slight imbalance between supply and demand. There is no need to cut off the load; only flexible power regulation should be activated, that is, to regulate the load that can be flexibly regulated. L 2; Severe power imbalance conditions: This means that the power grid is experiencing a power deficit exceeding the limit, resulting in a severe imbalance between supply and demand. It is necessary to coordinate the regulation of flexible load peak shaving and orderly load shedding that can be cut off in a short time. In other words, it is necessary to coordinate the regulation of flexible load L2 and load shedding that can be cut off in a short time. For power surplus conditions, this step focuses on load-side peak shaving and valley filling regulation, and only executes the corresponding regulation logic for power deficit and imbalance conditions. Based on the power grid operation status determination result, a differentiated load control strategy is matched. If it is a mild power imbalance condition or a severe power imbalance condition, step S44 is executed. S44: For a mild power imbalance condition: continue using the objective function: ; An upper limit constraint is added to the constraints to achieve precise balancing (a 5% fine-tuning redundancy range is added to balance grid control stability and completely avoid power over-adjustment problems): ; For severe power imbalance conditions: first exhaustively search... Maximum adjustable capacity, remaining deficit is determined by Ordered short-term excision and repair, using S3. Regulation Model and Safety Constraints: ; ; Constraints: ; in, : The remaining power deficit in the power grid at any given time refers to the power deficit that still needs to be made up after the flexible load has been fully adjusted. Load shedding is a 0-1 decision variable, where 0 = power supply preservation and 1 = load shedding. :load Actual duration of a single power outage; :load Original runtime latency tolerance; The optimal control target parameter, i.e., the load, is obtained by solving the problem based on the corresponding objective function and constraints. Optimal target power adjustment This generates corresponding preliminary control instructions.

[0018] Preferably, the specific process of step S42 is as follows: Power imbalance benchmark threshold calculation: Based on the grid's rated installed capacity, a benchmark value for the maximum allowable power deviation during steady-state operation is set as the fundamental constant standard for determining operating conditions. ; in, The power imbalance reference threshold is the maximum allowable power deviation under steady-state conditions. The allowable coefficient for steady-state power deviation of the power grid is 0.02~0.05, and 0.03 is taken under normal operating conditions. : Rated total active power of the power grid; Calculation of time-series fluctuation correction: Quantify the real-time fluctuation amplitude of the power grid at the current moment relative to the steady-state mean, and use it as a real-time correction component for the model to accurately characterize the degree of operating condition fluctuation. ; in, : Time-series fluctuation correction amount, and model-independent real-time fluctuation correction component; The time series fluctuation weighting coefficient can be set to 0.8 to balance the model's sensitivity and stability. Power Grid Real-time total active power; : The time-series average active power of the power grid over the past hour (sliding window statistical steady-state value); Adaptive dynamic correction coefficient calculation: The time-series fluctuation correction amount is normalized, and the dimensionless correction coefficient is solved to achieve adaptive scaling and adaptation to the benchmark threshold. ; in, : Real-time adaptive dynamic correction coefficient ; : Time-series fluctuation correction amount; The time-series average active power of the power grid over the past hour; Final adaptive dynamic threshold calculation: By fusing the steady-state benchmark threshold and the real-time fluctuation correction coefficient, a power imbalance judgment threshold that is dynamically updated according to operating conditions is generated, providing an accurate dynamic standard for grid condition identification. ; in, : Dynamic power safety threshold of the power grid at all times, and core judgment indicators for operating condition identification; : Real-time adaptive dynamic correction coefficient; Power imbalance reference threshold : Time-series fluctuation correction amount.

[0019] Preferably, the specific process of step S5 is as follows: S51: After data acquisition and control are completed Real-time total active power of the entire power grid After the adjustment is completed Real active power deficit of the power grid at all times Load after regulation Actual operating active power and load Actual power regulation ; S52: Obtain the optimal control target parameters, including The optimal steady-state total active power of the power grid at any given time (S4 control target value) Optimal power deficit of the power grid at any time (Ideal steady-state value, defaults to close to 0) and load Optimal target power adjustment (Solving the S4 model) optimal value ); S53: Single Load Control Deviation Calculation: Quantifies the deviation between the actual control amount and the optimal control amount for a single load, reflecting the accuracy of single load control. ; in, : Time load The single-load real-time control deviation value; :load Actual power regulation amount; :load Preset the optimal target power adjustment amount; Deviation Judgment Rules: In order to regulate, Due to insufficient regulation, For precise regulation; S54: Calculation of Overall Grid Control Deviation: Based on the overall grid power deficit deviation, an overall control deviation index is constructed to reflect the overall grid control balance effect and provide a basis for global model correction. ; in, : Real-time comprehensive control deviation value across the entire network; The actual active power deficit in the power grid after regulation; : The optimal power deficit of the power grid at any given moment (ideally, the steady-state value is 0); S55: Perform reverse correction of adaptive dynamic threshold model parameters: the correction targets are the time-series fluctuation weight coefficient and the dynamic correction coefficient, which have the highest sensitivity to the model. Inverse correction of time series fluctuation weighting coefficient: Based on comprehensive regulation deviation correction Optimize the model's sensitivity to operating condition fluctuations and avoid over- and under-regulation: ; in, : Corrected Time-series fluctuation weighting coefficient; The initial fixed time-series fluctuation weighting coefficient can be set to 0.8; The model iteration correction step size can be set to 0.1 to ensure that fine-tuning does not oscillate. : Real-time comprehensive control deviation value across the entire network; : Rated total active power of the power grid; Constraints: Limit the parameters to a reasonable range to avoid model failure; Iterative optimization of dynamic correction coefficients: By using the corrected time-series fluctuation weighting coefficients, the time-series fluctuation correction amount is recalculated, and the dynamic correction coefficients are iteratively updated to achieve closed-loop optimization of the model. ; ; in, Corrected time series fluctuation amount; : Adaptive dynamic correction coefficient after correction; The time-series average active power of the power grid over the past hour; Dynamic threshold update: Based on the iteratively optimized correction coefficients, the dynamic threshold for determining grid power imbalance is updated to complete the closed-loop correction of the model's core parameters. ; in, The threshold for determining dynamic power imbalance after iterative optimization; Power imbalance reference threshold (fixed steady-state parameters, no iterative correction); S56: Optimization of Load Control Execution and Precise Command Generation: Based on single load control deviations, the execution intensity of various types of load control is corrected, and targeted optimization is achieved. Flexible adjustment range and By eliminating the proportion of components and avoiding deviations in monomer load regulation, precise regulation commands are generated after iteration. Adjustment of flexible load regulation intensity: targeting The load can be flexibly adjusted, and the actual adjustable power is corrected based on the individual unit deviation to optimize the control accuracy. ; in, Post-iteration load Optimal power adjustment; The load regulation intensity correction coefficient can be taken as 0.5; : Single load real-time control deviation value; Constraints: The load regulation constraints S3 and S4 are retained to ensure... The load power adjustment range shall not be exceeded. The iteratively optimized dynamic threshold parameters and load control parameters are synchronously updated to the adaptive dynamic threshold control model and the S3 differentiated load control strategy library to generate the final precise control command, completing a single complete closed-loop correction of "command execution - data feedback - deviation correction - model iteration - command optimization". The optimized model parameters will participate in the power grid condition identification and control decision-making in the next time period, achieving continuous adaptive optimization.

[0020] Secondly, a dynamic load control system based on real-time power monitoring is provided to implement any one of the dynamic load control methods based on real-time power monitoring, including an edge real-time sensing module, which is deployed in each branch circuit and terminal load equipment to synchronously collect multi-dimensional electrical parameters and perform preprocessing. The adaptive decision-making module is communicatively connected to the edge real-time perception module and is used to construct an adaptive dynamic threshold control model based on standardized real-time power data, historical time-series data and multiple influencing factors, and dynamically update the power safety threshold and control margin. The load classification and control module is communicatively connected to the adaptive decision module. It is used to classify and classify terminal loads, establish a differentiated control strategy library, and generate preliminary load control instructions in combination with the real-time operating conditions of the power grid. The collaborative execution module is connected to the adaptive decision-making module and the load hierarchical control module respectively, and is used to coordinate the power grid, load, distributed photovoltaic and energy storage equipment, execute the initial load control command, and complete the source-grid-load-storage collaborative control. The closed-loop feedback correction module is bidirectionally connected to the collaborative execution module. It is used to collect the power grid and load operation data after regulation, compare the regulation deviation, correct the regulation parameters, and at the same time run the fault tolerance mechanism to ensure the stable operation of the system.

[0021] The beneficial effects of this invention include: By constructing an adaptive dynamic threshold control model, multi-dimensional influencing factors such as meteorological environment, electricity consumption period, and equipment operating status are incorporated into the dynamic threshold calculation, realizing adaptive iterative updates of the control threshold. This effectively solves the technical problems of static and fixed control rules and poor adaptability to operating conditions in existing technologies, enabling the control threshold to be dynamically adjusted according to real-time grid operating conditions, adapting to complex scenarios such as intermittent fluctuations in new energy output and random changes in load.

[0022] By classifying and categorizing terminal loads in a refined manner, three differentiated control strategies have been established: core guaranteed loads, flexibly adjustable loads, and loads that can be cut off in a short time. This effectively solves the technical problems of low precision in load control and inability to balance grid security and user reliability in existing technologies. Under emergency conditions such as grid peak load overload and power exceeding limits, non-core loads can be controlled in an orderly manner, avoiding the blind cutting off of important loads and ensuring the continuity of power supply for core production and people's livelihood.

[0023] By constructing a multi-dimensional collaborative scheduling mechanism involving sources, grids, loads, and storage, the technical problems of single load-side regulation and insufficient coordination between sources, grids, loads, and storage in existing technologies are effectively solved. It can offset power fluctuations through the charging and discharging characteristics of energy storage, improve the local consumption rate of new energy, smooth out the peak-valley difference of the power grid, and improve the overall energy efficiency of electricity consumption. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the adaptive dynamic threshold control process of the present invention; Figure 2 This is a schematic diagram illustrating the construction process of the load classification and differentiated control strategy library of the present invention; Figure 3 This is a schematic diagram of the power grid operation status determination and closed-loop feedback correction process of the present invention. Detailed Implementation

[0025] The following is in conjunction with the appendix Figures 1-3 The present invention will be further described in detail below: See appendix Figure 1 As shown, this invention provides a dynamic load control method and system based on real-time power monitoring. This method collects electrical parameters of each branch circuit and terminal load equipment of the power grid at high frequency and synchronously, constructs an adaptive dynamic threshold control model, realizes refined classification and differentiated control of terminal loads, and forms a closed-loop optimization by combining a real-time feedback correction mechanism to achieve the goal of dynamic, accurate and flexible load control of the distribution network.

[0026] The dynamic load control system based on real-time power monitoring provided in this embodiment of the invention includes an edge real-time sensing module, an adaptive decision-making module, a load grading control module, a collaborative execution module, and a closed-loop feedback correction module. The modules interact with each other and link functions through a communication network.

[0027] The edge real-time sensing module is deployed in each branch circuit and terminal load device of the target power supply network to synchronously collect multi-dimensional raw electrical parameters such as voltage, current, active power, reactive power, and power factor corresponding to each terminal load at high frequency. The module integrates a noise reduction and filtering circuit, an abnormal data identification module, and a compliance calibration module. It performs local noise reduction filtering, abnormal data identification, and compliance calibration preprocessing operations on the raw electrical parameters and outputs standardized real-time power data to the adaptive decision module.

[0028] The adaptive decision-making module communicates with the edge real-time sensing module to construct an adaptive dynamic threshold control model based on standardized real-time power data, historical time-series data, and multiple influencing factors, dynamically updating the power safety threshold and control margin. This module incorporates a multi-dimensional influencing factor quantification correction unit, a time-series iterative update unit, and an adaptive dynamic threshold calculation unit. It can perform calculations of the grid's basic power safety threshold, multi-dimensional influencing factor quantification correction coefficients, and iterative updates of the adaptive dynamic power safety threshold based on real-time acquired data.

[0029] The load classification and control module communicates with the adaptive decision-making module to classify and categorize terminal loads, establish a differentiated control strategy library, and generate preliminary load control instructions based on real-time grid conditions. This module includes a load parameter normalization processing unit, a comprehensive scoring calculation unit, a load classification determination unit, and a differentiated control strategy generation unit. It can classify loads based on four dimensions: load rigidity coefficient, power supply importance level, power adjustable range, and operating delay tolerance.

[0030] The collaborative execution module is connected to the adaptive decision-making module and the load hierarchical control module, respectively, to coordinate the power grid, load, distributed photovoltaic and energy storage devices, execute preliminary load control commands, and complete the coordinated control of the power source, grid, load and storage. This module includes a multi-energy collaborative scheduling unit, a command decomposition and execution unit and an equipment status monitoring unit, which can decompose the control commands into specific equipment control commands and send them to each terminal for execution.

[0031] The closed-loop feedback correction module and the collaborative execution module are bidirectionally connected to collect power grid and load operation data after regulation, compare regulation deviations, correct regulation parameters, and simultaneously operate an anomaly tolerance mechanism to ensure stable system operation. This module includes a real-time data acquisition unit, a deviation calculation unit, a parameter correction unit, and an anomaly handling unit. It can complete a single regulation closed-loop correction and synchronously update the optimized parameters to the adaptive decision-making module and the load-level regulation module.

[0032] In its implementation, the edge real-time sensing module is deployed in a distributed architecture. Each terminal load device is configured with an edge acquisition terminal. Each terminal collects data and aggregates it to the regional concentrator through the Internet of Things communication protocol. The regional concentrator then uploads the standardized real-time power data to the adaptive decision module through a dedicated communication channel.

[0033] The preprocessing operation includes the following three steps. The first step is local noise reduction filtering, which uses an adaptive Kalman filter algorithm to reduce noise in the raw sampled data. The filter parameters are automatically adjusted based on the signal-noise characteristics to effectively filter out noise components caused by electromagnetic interference. The second step is outlier detection, which uses a statistically based outlier detection method to fit a normal distribution to the continuous sampled data, identifying and removing outlier data points that significantly deviate from the normal range. The third step is compliance calibration, which performs timestamp calibration, range conversion, and unit normalization on the detected valid data, outputting standardized real-time power data.

[0034] Reference Figure 1 The construction and updating process of the adaptive dynamic threshold control model includes four core steps: calculation of the power grid base power safety threshold, calculation of the multi-dimensional influence factor quantification correction coefficient, calculation of the adaptive dynamic power safety threshold, and model time-series iterative update.

[0035] Calculation of the power grid baseline power safety threshold: This baseline power safety threshold is constructed based on historical operating data of the distribution network during the same period, serving as a benchmark value under undisturbed steady-state conditions. The calculation formula is as follows: ; in, The basic power safety threshold of the power grid is the reference threshold without real-time operating condition correction. The average active power for the same historical period is taken from the equipment's historical operating sequence database and serves as the baseline power for steady-state operation of the distribution network. The rated allowable active power of the distribution network branch / transformer is a fixed parameter from the equipment factory and the power grid ledger. These are the historical time series weighting coefficients and the equipment rated constraint weighting coefficients, respectively, satisfying... The value range is [0,1], and it is calibrated according to the distribution network operation procedure. Calculation of quantitative correction coefficients for multi-dimensional influencing factors: Quantitative correction coefficients are constructed for three core influencing factors: real-time meteorological environment, electricity consumption period type, and equipment operating status, including a meteorological environment correction coefficient. Electricity usage period correction factor and equipment operating status correction coefficient This enables dynamic adaptation and correction of the base threshold.

[0036] The calculation of the meteorological environment correction factor needs to adapt to the impact of temperature, sunlight, and wind and rain on the output of new energy sources and the power consumption of the load. In the specific calculation, the deviations between the real-time ambient temperature and the reference steady-state temperature, the deviations between the real-time sunlight intensity and the reference standard sunlight intensity are comprehensively considered, and quantitative calculations are performed in conjunction with empirical calibration parameters of the power system's meteorological-power response. The specific formula is as follows: ; in, This is a comprehensive correction factor for meteorological environment; Real-time ambient temperature and reference steady-state temperature; These are real-time illumination intensity and baseline standard illumination intensity, respectively. These are the temperature correction factor and the illumination correction factor, respectively, which are empirical calibration parameters for the meteorological-power response of the power system. In this embodiment, they can be set to 0.3 and 0.2, respectively.

[0037] The calculation of the electricity consumption time correction factor needs to distinguish the differences in load characteristics during peak, average, and off-peak periods. This correction factor sets fixed correction factors for peak, average, and off-peak periods, and is calibrated based on the time-series characteristics of the regional distribution network. The specific formula is as follows: ; in, Correction factor for electricity usage period; These are fixed correction coefficients for peak, normal, and off-peak periods, calibrated based on the time-series characteristics of electricity consumption in the regional distribution network, to meet... In this embodiment, the correction factor for peak hours is set to 1.2, the correction factor for normal hours is set to 1.0, and the correction factor for off-peak hours is set to 0.8, satisfying the relationship that the correction factor for peak hours is greater than the correction factor for normal hours, and the correction factor for normal hours is greater than the correction factor for off-peak hours.

[0038] The calculation of the equipment operating status correction factor needs to adapt to the impact of equipment load rate and fluctuations in operating health status. This correction factor is calculated based on the equipment's real-time load rate, which is derived from the ratio of standardized real-time power data to the equipment's rated power, and then quantitatively corrected using a load attenuation correction factor. The specific formula is as follows: ; in, This is a correction factor for the equipment's operating status. The real-time load rate of the equipment is calculated from the real-time power collected by S1. ; The standardized real-time active power output by S1; This is the load attenuation correction factor, with a value range of (0,1). It is obtained by looking up a table according to the equipment type, and in this embodiment, it can be set to 0.15.

[0039] Calculation of the adaptive dynamic power safety threshold: By integrating multi-dimensional correction factors, the basic power safety threshold of the power grid is dynamically weighted and corrected to obtain a dynamic power safety threshold adapted to the current real-time operating conditions. The specific calculation formula is as follows: ; in, : Real-time dynamic power safety threshold, updated iteratively in real time.

[0040] The load controllability margin is quantified based on the difference between real-time operating power and the dynamic power safety threshold, and is used to guide the allocation of subsequent load tiered control measures. The specific calculation formula is as follows: ; in, Real-time load controllable margin, unit: kW; This indicates that the power grid has a power surplus and can accommodate the increased load. This indicates that the power grid has exceeded its limit, and load regulation needs to be initiated. The standardized real-time active power output of S1; Real-time dynamic power safety threshold.

[0041] In this embodiment, the margin value is quantified and displayed in kilowatts, and two alarm thresholds are set in the system. When the margin value is lower than the first alarm threshold, the system generates a mild power warning to alert maintenance personnel; when the margin value is lower than the second alarm threshold, the system automatically triggers the load control process, and the second alarm threshold is set to 20% of the dynamic power safety threshold.

[0042] Reference Figure 2 The specific process of constructing the load classification and differentiated control strategy library includes four core steps: load classification parameter normalization processing, standardized load comprehensive assessment model construction, batch classification of the entire network load, and differentiated load control logic modeling.

[0043] Load classification parameter normalization: It is necessary to normalize the parameters of power consumption rigidity coefficient, power supply importance level, power adjustment range, and operation delay tolerance to obtain normalized derived indices for each parameter.

[0044] The original power rigidity coefficient ranges from extremely flexible loads to fully rigid loads. After normalization, a normalized power rigidity coefficient is obtained. A higher coefficient indicates stronger power rigidity, meaning the load is neither adjustable nor subject to power outages. In practice, operating data for each load over the past thirty days is collected, and its power variation characteristics and interruption tolerance are analyzed. The system automatically determines the original rigidity coefficient and normalizes it to the range of 0 to 1. The normalized rigidity coefficient for fully rigid loads is 1, and for extremely flexible loads, it is 0.

[0045] Original load importance level The power supply importance levels are categorized as follows: 1 (Ordinary Load), 2 (Secondary Important Load), 3 (Primary Important Load), and 4 (Special Grade Guarantee Load). The normalization formula for these importance levels is as follows: ; in, :load Normalized importance weights The higher the value, the higher the priority of power supply protection; :load Original power supply importance level; The minimum load importance level is fixed at 1. The maximum value for the load importance level is fixed at 4.

[0046] The ratio of the adjustable power range to the rated power is normalized to obtain the normalized adjustable power range margin coefficient, and the specific formula is as follows: ; in, :load Normalized power adjustable amplitude margin coefficient The larger the value, the greater the range of adjustable load power. For load The original maximum adjustable active power corresponding to the power adjustment range For load Rated active power; in specific implementation, the maximum allowable power reduction ratio is calculated according to the equipment type and operating procedures of each load, with a typical range of 0 to 50%.

[0047] The normalized runtime latency tolerance coefficient is obtained after normalization, and the specific formula is as follows: ; in, :load Normalized runtime delay tolerance coefficient The higher the value, the more prohibited short-term power outages are for the load. :load Original permissible short-term power outage duration; The maximum tolerable power outage time for the entire network load can be set to 2 hours. This parameter is automatically determined by the system based on the load type. The normalized tolerance coefficient is 0 for extremely sensitive loads such as hospital intensive care units, 0.5 for regular production loads, and 1 for completely interruptible loads.

[0048] Construction of a standardized load comprehensive assessment model: This comprehensive assessment model adopts a weighted comprehensive scoring method, and the weight parameters are set according to the guidelines for flexible load dispatching technology of power grids. In this specific embodiment, the weights are set as follows: power rigidity weight 0.35, load importance weight 0.35, power adjustability weight 0.15, and operating delay tolerance weight 0.15.

[0049] The comprehensive load assurance score is calculated as follows: Each normalized derived index is substituted into the weighted comprehensive scoring formula to obtain the load comprehensive assurance score. A higher score indicates a higher priority for power supply assurance and stronger constraints prohibiting regulation and disconnection. The specific calculation formula is as follows: ; in, :load The comprehensive protection score indicates a higher priority for power supply protection and stronger constraints prohibiting adjustment or disconnection. In this specific embodiment, it can be set as a rigid weight for power consumption. Load importance weight Adjustable power amplitude weight Runtime delay tolerance weight ,satisfy ; Batch load grading across the entire network: The load grading threshold rules are set as follows: When the comprehensive guarantee score is greater than or equal to the first preset threshold... At that time, the load was classified as core backup load. When the comprehensive protection score is within the first preset threshold With the second preset threshold During this period, the load is divided into flexibly adjustable loads. When the comprehensive protection score is less than the second preset threshold At that time, the load is divided into loads that can be cut off for short periods of time. In this embodiment, the first preset threshold Set to 0.75, second preset threshold Set to 0.4. The thresholds for each level are set according to the power grid emergency dispatch standards.

[0050] The specific process for completing the batch classification of the entire network load is as follows: Traverse all terminal loads of the target power supply network, perform normalization calculations on each individual load sequentially to obtain standardized derived parameters, substitute these parameters into the comprehensive scoring formula to calculate the comprehensive guarantee score, compare with the classification threshold to automatically label the load category, and output a load classification ledger. This ledger includes key information such as load number, rated power, four-dimensional original parameters, normalized parameters, comprehensive score, and load category.

[0051] Differentiated load control logic modeling: For core guaranteed loads, priority is given to ensuring power supply, and load reduction and disconnection are prohibited. The control objective is to maintain the rated operating state of core guaranteed loads, minimize active power fluctuations, and strictly prohibit load reduction, power outages, and disconnections. Constraints include: real-time active power must be maintained near the rated power value, both upper and lower power limits are equal to the rated power value, the permissible power outage duration is forced to zero, and any short-term disconnection operations are prohibited.

[0052] For the control logic of flexibly adjustable loads, power reduction is permitted, but load shedding is strictly prohibited. The control objective is to maximize the power reduction of flexible loads within the safe operating range of equipment under grid power deficit conditions, ensuring uninterrupted power supply and no load shedding. The objective function is as follows: ; The constraints are: .

[0053] in, The objective function is to minimize the deviation of the core guaranteed load power. :load Real-time active power; Core load power upper and lower limits For load The rated active power; The allowable single power outage duration for the load is set to 0 for the core load, prohibiting any short-term disconnection operations. In this embodiment, the minimum safe operating power for the flexible load is set to 60% of its rated power.

[0054] For the control logic of loads that can be temporarily shelved, these loads are shelved in an orderly and short-term manner when the power deficit exceeds the limit. The control objective is to shelve shelvable loads in an orderly and short-term manner according to load priority when the power deficit of the grid exceeds the maximum adjustment capacity of the flexible load, and control the actual power outage duration to not exceed the allowable threshold.

[0055] The objective function is as follows: ; The constraints are: .

[0056] in, The objective function for maximizing the total adjustable power of a flexibly adjustable load; :load Actual reduction in active power; The minimum safe operating power for flexible loads; operating below this value will cause equipment failure. : The load can be flexibly adjusted and short-term power outages are prohibited.

[0057] The load can be removed in a short time. Control logic (orderly short-term disconnection when power deficit exceeds the limit): Control objective: When the power deficit of the power grid exceeds the maximum adjustment capacity of the flexible load, the loads that can be disconnected are disconnected in an orderly short-term manner according to the load priority, and the actual power outage duration is strictly controlled to not exceed the allowable threshold.

[0058] Objective function: ; Constraints: ; in, The objective function is to maximize the total power cut-off for loads that can be cut off in a short time. : Load removal of 0-1 decision variables ; :load Actual duration of a single power outage; Real-time total active power deficit in the power grid; The total maximum adjustable power of the entire network's flexible loads is calculated from the preceding original parameters.

[0059] The differentiated load control strategy library adopts a hierarchical storage structure, comprising the following three layers: The bottom layer is a standardized hierarchical rule library, storing normalization formulas for four types of parameters, load comprehensive score calculation formulas, and hierarchical threshold standards, achieving real-time automatic classification of the entire network load based on historical runtime sequence data. The middle layer is a differentiated load control model library, storing the specific objective functions, constraints, and parameter boundaries for three types of loads: core guaranteed loads, flexibly adjustable loads, and loads that can be temporarily cut off. The top layer is a scenario-based scheduling strategy instance library, generating strategies that can be directly invoked in multiple scenarios based on the bottom-layer hierarchical rules and the middle-layer control models, including scenario strategies for steady-state normal operating conditions, mild power imbalance operating conditions, and severe power imbalance operating conditions.

[0060] Reference Figure 3 The specific process of determining the power grid operating status and generating preliminary control instructions includes four core steps: standardized real-time power data acquisition, adaptive dynamic threshold model construction and threshold update, power grid operating status comparison and power imbalance abnormal operating condition identification, and preliminary control instruction generation.

[0061] Standardized real-time power data acquisition. This step acquires the total active power of the entire grid at any given time, the total active power of the grid's rated installed capacity, the real-time active power deficit of the grid at any given time, and the real-time active power of each load at any given time. The formula for calculating the real-time active power deficit is: Real-time active power deficit equals the total active power of rated installed capacity minus the total real-time active power.

[0062] Construction and updating of adaptive dynamic threshold model: The adaptive dynamic threshold model obtains the final adaptive dynamic threshold through a three-layer progressive correction structure of baseline threshold, time-series fluctuation correction amount, and adaptive dynamic correction coefficient.

[0063] The benchmark threshold is calculated based on the maximum allowable power deviation benchmark value for steady-state operation set by the rated installed capacity of the power grid. This benchmark serves as the basic constant standard for determining the operating condition. The specific formula is as follows: ; in, The power imbalance reference threshold is the maximum allowable power deviation under steady-state conditions. The allowable coefficient for steady-state power deviation of the power grid is 0.02~0.05, and 0.03 is taken under normal operating conditions. Rated installed total active power of the power grid.

[0064] The calculation of the time-series fluctuation correction quantifies the real-time fluctuation amplitude of the power grid's current power relative to the steady-state mean, serving as a real-time correction component for the model and accurately characterizing the degree of operating condition fluctuation. Specifically, the average active power of the power grid over the past hour is first calculated as the steady-state mean, and then the deviation between the current active power and the steady-state mean is calculated as the time-series fluctuation correction.

[0065] The specific formula is as follows: ; in, : Time-series fluctuation correction amount, and model-independent real-time fluctuation correction component; The time series fluctuation weighting coefficient can be set to 0.8 to balance the model's sensitivity and stability. Power Grid Real-time total active power; The time-series average active power of the power grid over the past hour is the statistical steady-state value obtained through a sliding window.

[0066] The calculation of the adaptive dynamic correction coefficient normalizes the time-series fluctuation correction amount and solves for the dimensionless correction coefficient, achieving adaptive scaling to the baseline threshold. The specific formula is as follows: ; in, : Real-time adaptive dynamic correction coefficient ; : Time-series fluctuation correction amount; The average active power of the power grid over the past hour.

[0067] The final adaptive dynamic threshold is calculated by combining the steady-state baseline threshold and the real-time fluctuation correction coefficient to generate a power imbalance judgment threshold that is dynamically updated according to operating conditions. The specific calculation formula is as follows: Final adaptive dynamic threshold calculation: By fusing the steady-state benchmark threshold and the real-time fluctuation correction coefficient, a power imbalance judgment threshold that is dynamically updated according to operating conditions is generated, providing an accurate dynamic standard for grid condition identification. ; in, : Dynamic power safety threshold of the power grid at all times, and core judgment indicators for operating condition identification; : Real-time adaptive dynamic correction coefficient; Power imbalance reference threshold : Time-series fluctuation correction amount.

[0068] Power grid operation status comparison and power imbalance abnormal condition identification: The real-time power deficit of the power grid is compared with the adaptive dynamic threshold, and three types of operation conditions are divided according to the power grid dispatching industry standard to accurately identify the power grid operation status.

[0069] The rule for determining steady-state normal operating conditions is: the absolute value of the real-time power deviation of the power grid is less than or equal to the adaptive dynamic threshold, that is, the real-time power deviation of the power grid is within the dynamic allowable range, the power supply and demand are balanced, there is no need for regulation, and the load of the entire network maintains the current rated operating state.

[0070] The criteria for determining a mild power imbalance are: if the real-time power deficit of the power grid is greater than the adaptive dynamic threshold but less than or equal to twice the adaptive dynamic threshold, then the power grid has a small power deficit and a slight imbalance between supply and demand. There is no need to cut off the load; only flexible power regulation should be initiated to regulate the load that can be flexibly regulated.

[0071] The criteria for determining severe power imbalance conditions are as follows: when the real-time power deficit of the power grid is greater than twice the adaptive dynamic threshold, that is, the power deficit of the power grid exceeds the limit, the supply and demand are severely imbalanced, and it is necessary to coordinate the flexible load peak shaving and the orderly disconnection of loads that can be cut off in a short time for coordinated regulation.

[0072] The rule for determining the power surplus condition is: when the real-time power deficit of the power grid is less than zero, this condition focuses on load-side peak shaving and valley filling regulation, and only executes the corresponding regulation logic for the power deficit imbalance condition.

[0073] Preliminary control instructions are generated. Based on the power grid operating status determination results, a differentiated load control strategy is matched. If the condition is a mild or severe power imbalance, subsequent control instruction generation steps are executed.

[0074] The control command generation under mild power imbalance conditions adopts the objective function of maximizing the total adjustable power of flexibly adjustable loads. An upper limit constraint is added to the constraints to achieve precise leveling, and a moderately fine-tuned redundancy range is added to balance grid control stability while completely avoiding power over-adjustment. Specifically, the objective function is to maximize the total power reduction of flexibly adjustable loads, with constraints including that the power reduction of each flexible load does not exceed its adjustable range, the adjusted power is not lower than the minimum safe operating power, and the total power reduction of the entire network is greater than or equal to the real-time power deficit.

[0075] The generation of control commands under severe power imbalance conditions first exhaustively considers the maximum adjustable capacity of flexible loads. The remaining deficit is then filled by the orderly, short-term shelving of loads that can be temporarily cut off. An objective function maximizing the total power cut off by loads that can be temporarily cut off is adopted, ensuring, under constraints, that the total power of the cut-off loads is greater than or equal to the power deficit that still needs to be filled after the flexible loads are fully adjusted. Specifically, the maximum adjustable total capacity of flexible loads is first calculated. The real-time power deficit is compared with this total capacity. If the deficit is less than or equal to the maximum adjustable total capacity, only the flexible loads are adjusted; if the deficit is greater than the maximum adjustable total capacity, the adjustment of the flexible loads is maximized first, and the remaining deficit is satisfied by the sequential shelving of loads that can be temporarily cut off according to priority.

[0076] Based on the corresponding objective function and constraints, an optimization solution is performed to obtain the optimal control target parameters, i.e., the optimal target power adjustment amount of the load, and to generate the corresponding preliminary control command. In this embodiment, the optimization solution adopts a linear programming algorithm, which is deployed in the execution unit of the adaptive decision module.

[0077] The specific process of closed-loop feedback correction includes five core steps: data acquisition after regulation, calculation of single load regulation deviation, calculation of comprehensive regulation deviation of the whole network, reverse correction of adaptive dynamic threshold model parameters, and optimization of load regulation execution intensity and generation of precise instructions.

[0078] Post-control data acquisition: This step collects the actual total active power of the entire power grid at the time of control completion, the actual active power deficit of the power grid at the time of control completion, the actual operating active power of each load after control completion, and the actual power adjustment amount executed by each load. The data acquisition delay is set to five minutes after the control command is executed to ensure that the equipment response is complete.

[0079] Single-load control deviation calculation: This quantifies the deviation between the actual control amount and the optimal control amount for a single load, reflecting the accuracy of single-load control. In the specific calculation, the actual power adjustment amount of each load is compared with the optimal target power adjustment amount to calculate the deviation value. A result greater than zero indicates over-control, a result less than zero indicates under-control, and a result equal to zero indicates precise control.

[0080] Calculation of Overall Network Control Deviation: A comprehensive control deviation index is constructed based on the overall network power deficit deviation to reflect the overall grid control balance effect and provide a basis for global model correction. Specifically, the deviation between the actual active power deficit of the grid at the time of control completion and the optimal power deficit of zero is used as the comprehensive deviation index.

[0081] Reverse correction of adaptive dynamic threshold model parameters: The correction targets are the time-series fluctuation weight coefficient and the dynamic correction coefficient, which have the highest sensitivity to the model.

[0082] The reverse correction of the time-series fluctuation weight coefficient is based on the comprehensive control deviation correction of the time-series fluctuation weight coefficient, optimizing the model's sensitivity to operating condition fluctuations and avoiding over-control and under-control. This correction sets constraints to limit the parameters within a reasonable range to prevent model failure.

[0083] The iterative optimization of the dynamic correction coefficient uses the corrected time-series fluctuation weight coefficient to recalculate the time-series fluctuation correction amount, and iteratively updates the dynamic correction coefficient to achieve closed-loop optimization of the model. In this embodiment, the gradient descent method is used for this iterative process, and the iteration termination condition is that the comprehensive control deviation is less than the preset convergence threshold or the number of iterations reaches ten.

[0084] The dynamic threshold is updated based on the iteratively optimized correction coefficient, and the dynamic judgment threshold of power grid imbalance is updated to complete the closed-loop correction of the core parameters of the model.

[0085] Optimization of load regulation execution intensity and generation of precise instructions: Based on the deviation of single load regulation, the execution intensity of various types of load regulation is corrected, the flexible adjustment range and cut-off ratio are optimized in a targeted manner, the deviation of single load regulation is avoided, and precise regulation instructions are generated after iteration.

[0086] The flexible load regulation correction is designed for flexibly adjustable loads. It corrects the actual adjustable power based on individual unit deviations to optimize regulation accuracy. This correction sets constraints, using the same load regulation constraints, to ensure that the optimal load regulation power after iteration does not exceed the upper limit of the load power's adjustable range. In this embodiment, the corrected actual adjustable power equals the original adjustable power plus the deviation correction amount, which is set to half the actual deviation value.

[0087] The iteratively optimized dynamic threshold parameters and load control parameters are synchronously updated to the adaptive dynamic threshold control model and the differentiated load control strategy library to generate the final precise control command. This completes a single closed-loop correction of command execution, data feedback, deviation correction, model iteration, and command optimization. The optimized model parameters will participate in the power grid condition identification and control decision-making for the next time period, achieving continuous adaptive optimization.

[0088] In its implementation, the collaborative execution module includes the following functional units: The multi-energy collaborative dispatch unit is responsible for coordinating the output matching of the main power source, distributed photovoltaic power, energy storage systems, and controllable loads. Under power surplus conditions, this unit prioritizes charging energy storage systems and increasing the supply of adjustable loads; under power deficit conditions, this unit prioritizes discharging energy storage systems and reducing the supply of controllable loads. This unit maintains communication with the superior power grid dispatch system, receives dispatch instructions, and reports the system's operating status upwards.

[0089] The instruction decomposition and execution unit is responsible for decomposing load control instructions into control instructions for individual terminal devices. Based on the load classification ledger and control strategy library, this unit generates corresponding device control instructions and distributes them to each terminal device via the communication network. The format of the control instructions includes parameters such as device identifier, target power value, adjustment rate, and duration.

[0090] The equipment status monitoring unit is responsible for real-time monitoring of the operating status and control command execution of each terminal device. This unit receives real-time operating data from the devices, including actual power, operating status, and alarm information, and generates alarms for abnormal situations. This unit also has a device fault isolation function, automatically isolating devices when a fault occurs to prevent impact on the normal operation of other devices.

Claims

1. A dynamic load control method based on real-time power monitoring, characterized in that, Includes the following steps: S1: Collect multi-dimensional raw electrical parameters such as voltage, current, active power, reactive power and power factor corresponding to each terminal load, and output standardized real-time power data after preprocessing; S2: Based on the historical operating sequence data of the equipment, it couples multiple influencing factors including real-time meteorological environmental parameters, electricity consumption time type parameters, and real-time operating status parameters of the equipment, and dynamically updates the power grid power safety threshold and load controllability margin in real time. It also builds and updates an adaptive dynamic threshold control model that adapts to the current power grid operating conditions online. S3: Establish standardized load classification rules in advance based on load power consumption rigidity coefficient, power supply importance level, power adjustment range, and operation delay tolerance parameters. According to the load classification rules, classify all terminal loads in the target power grid into any one of the following: core guaranteed load, flexible adjustable load, and short-term cut-off load, and construct a differentiated load control strategy library. S4: Import standardized real-time power data into the adaptive dynamic threshold control model, determine the current operating status of the power grid and identify the power grid operating condition type through the model; based on the identified power grid operating condition type, match and call the differentiated load control strategy library to generate preliminary load control instructions; S5: Executes preliminary load control commands and synchronously collects real-time data on the actual operation of the entire network after the commands are executed; The actual operating data collected is compared with the preset optimal control target parameters to obtain the real-time control deviation value. Based on the real-time control deviation value, the decision parameters of the adaptive dynamic threshold control model and the load control execution intensity are corrected in reverse. The precise control command is obtained through iterative optimization, and the single control closed-loop correction is completed. S6: Using the corrected and precise control results as the benchmark for the new round of control, S1 to S5 are executed iteratively in a loop, continuously updating real-time power data, adaptive dynamic threshold control model, power grid condition identification results and control instructions, so as to achieve continuous dynamic flexible control of terminal load. The original power rigidity coefficient ranges from extremely flexible loads to fully rigid loads. After normalization, the normalized power rigidity coefficient is obtained. The larger the coefficient value, the stronger the power rigidity, meaning it is not adjustable and cannot be disconnected. The comprehensive load assurance score is calculated as follows: Each normalized derived index is substituted into the weighted comprehensive scoring formula to obtain the load comprehensive assurance score. A higher score indicates a higher priority for power supply assurance and stronger constraints prohibiting regulation and disconnection. The specific calculation formula is as follows: ； in, :load The higher the comprehensive protection score, the higher the priority of power supply protection and the stronger the constraint against adjustment or disconnection. For load Normalized electricity consumption rigidity coefficient; For load Normalized importance weights; For load Normalized power adjustable margin coefficient; For load Normalized runtime delay tolerance coefficient; For the rigid weight of electricity consumption, For the importance weight of the load, For power adjustable amplitude weight, As the runtime latency tolerance weight, satisfying .

2. The dynamic load control method based on real-time power monitoring according to claim 1, characterized in that, The specific process of step S2 is as follows: S21: Calculation of the power grid base power safety threshold: Based on historical operating data of the distribution network during the same period, construct the power grid base power safety threshold under undisturbed steady-state conditions. , serving as a benchmark value for dynamic correction; S22: Calculate the quantitative correction coefficients for multi-dimensional influencing factors: Construct quantitative correction coefficients for three core influencing factors: real-time meteorological environment, electricity consumption period type, and equipment operating status, including a meteorological environment correction coefficient. Electricity usage period correction factor and equipment operating status correction coefficient ; S23: Calculate the adaptive dynamic power safety threshold: Integrate multi-dimensional correction factors to dynamically weight and correct the grid's basic power safety threshold, obtaining a dynamic power safety threshold adapted to the current real-time operating conditions. ; S24: Load controllability margin calculation: Based on the difference between real-time operating power and dynamic power safety threshold, the current dispatchable and adjustable power margin of the power grid is quantitatively calculated, and the real-time load controllability margin is calculated. ; S25: Model time-series iterative update mechanism: The model parameters are updated in real time based on the latest acquired data to obtain the latest dynamic power safety threshold at time t. .

3. The dynamic load control method based on real-time power monitoring according to claim 2, characterized in that, The specific process of step S22 is as follows: Calculate meteorological and environmental correction factors To adapt to the impact of temperature, sunlight, wind, rain and wind on the output of new energy sources and the power consumption of the load; Calculate the correction factor for electricity consumption period To differentiate the differences in load electricity consumption characteristics during peak, off-peak, and valley periods; Calculate the equipment operating status correction factor To adapt to fluctuations in equipment load rate and operational health status.

4. The dynamic load control method based on real-time power monitoring according to claim 1, characterized in that, The specific process of step S3 is as follows: S31: Normalize the parameters of power consumption rigidity coefficient, power supply importance level, power adjustment range, and operating delay tolerance to obtain the load. Normalized power consumption rigidity coefficient ,load Normalized importance weights ,load Normalized power adjustable amplitude margin coefficient and load Normalized runtime delay tolerance coefficient ; S32: Construct a standardized load comprehensive assessment model to perform batch load classification across the entire network, and automatically label each load category. , These correspond to core protection loads, flexibly adjustable loads, and loads that can be temporarily cut off, respectively. S33: Targeting Based on the operating characteristics, establish the active power regulation objective function and constraints respectively, and perform differentiated load regulation logic modeling: S34: Differentiated load control strategies are stored based on a hierarchical storage structure using a strategy library, where the hierarchical storage structure includes: The underlying layer is a standardized hierarchical rule base: storing normalization formulas for four types of parameters and comprehensive load scores. S i Calculation formulas and grading threshold standards; The middle layer is a library of differentiated load control models: storing them separately. L 1. L 2. L 3. Objective functions, constraints, and parameter boundaries for the three types of loads; The top layer is a scenario-based scheduling strategy instance library: based on the bottom-level hierarchy + middle-level model, it generates multiple scenarios that can be directly invoked, including scenario strategies for steady-state normal operating conditions, mild power imbalance operating conditions, and severe power imbalance operating conditions.

5. The dynamic load control method based on real-time power monitoring according to claim 4, characterized in that, The specific process of step S32 is as follows: The load comprehensive guarantee score is calculated based on the standardized load comprehensive assessment model to obtain the load. Comprehensive protection score and based on load The comprehensive protection score is used to classify the network load in batches, and the classification threshold rules are as follows: like :load Classified as core backup load ; like :load Classified as flexibly adjustable load ; like :load Classified as load that can be cut off in a short time ; in, The core guarantee is the load classification threshold; The threshold for short-term load grading can be used; Perform batch classification of network-wide load: Traverse all terminal loads of the target power supply network ; Normalization calculations are performed sequentially on a single load to obtain standardized derived parameters. ; Substitute into the comprehensive scoring formula to calculate By comparing the classification thresholds, the load category is automatically labeled. .

6. The dynamic load control method based on real-time power monitoring according to claim 4, characterized in that, The specific process of step S33 is as follows: Core protection load The control logic prioritizes ensuring power supply and prohibits load reduction or power disconnection. Objective function: ; Constraints: ; in, The objective function is to minimize the deviation of the core guaranteed load power. :load Real-time active power; Core load power upper and lower limits For load The rated active power; Permissible power outage duration for a single load operation; Flexible load adjustment The control logic allows for power reduction but strictly prohibits power cut-off. Objective function: ; Constraints: ; in, The objective function for maximizing the total adjustable power of a flexibly adjustable load; :load Actual reduction in active power; Minimum safe operating power for flexible loads; Flexible load adjustment prohibits short-term power outages; Can remove load for short time The control logic is to perform orderly short-term cutoff when the power deficit exceeds the limit: Objective function: ; Constraints: ; in, The objective function is to maximize the total power cut-off for loads that can be cut off in a short time. : Load removal of 0-1 decision variables; :load Actual duration of a single power outage; Real-time total active power deficit in the power grid; The total maximum adjustable power of the entire network's flexible loads; :load Original allowable short-term power outage duration.

7. The dynamic load control method based on real-time power monitoring according to claim 6, characterized in that, The specific process of step S4 is as follows: S41: Obtain standardized real-time power data, including power grid data. Real-time total active power of the entire network Total rated installed active power of the power grid power grid Real-time active power deficit and load exist Real-time operation of active power ; S42: Construct a power imbalance baseline threshold +Time-series fluctuation correction +Adaptive dynamic correction coefficient A three-layer progressive adaptive dynamic threshold model is used to obtain the final adaptive dynamic threshold. ; S43: Real-time power deficit in the power grid With adaptive dynamic threshold The comparison is performed to classify the three operating conditions and identify the power grid operating status, including steady-state normal operating condition, mild power imbalance operating condition and severe power imbalance operating condition; based on the power grid operating status determination result, a differentiated load control strategy is matched. If it is a mild power imbalance operating condition or a severe power imbalance operating condition, step S44 is executed. S44: For a mild power imbalance condition: continue using the objective function: Furthermore, an upper limit constraint was added to the constraints to achieve precise balancing; For severe power imbalance conditions: first exhaustively search... Maximum adjustable capacity, remaining deficit is determined by Ordered short-term excision and repair, following the S3 method. Regulation models and safety constraints; The optimal control target parameter, i.e., the load, is obtained by solving the problem based on the corresponding objective function and constraints. Optimal target power adjustment This generates corresponding preliminary control instructions.

8. The dynamic load control method based on real-time power monitoring according to claim 7, characterized in that, The specific process of step S42 is as follows: Power imbalance baseline threshold calculation: ; in, This serves as the baseline threshold for power imbalance. : Permissible coefficient for steady-state power deviation of power grid; : Rated total active power of the power grid; Calculation of time series fluctuation correction: ; in, : Time-series fluctuation correction amount; Power Grid Real-time total active power; The time-series average active power of the power grid over the past hour; Time-series fluctuation weighting coefficient; Adaptive dynamic correction coefficient calculation: ; in, : Real-time adaptive dynamic correction coefficient; : Time-series fluctuation correction amount; The time-series average active power of the power grid over the past hour; Final adaptive dynamic threshold calculation: ; in, : Adaptive dynamic threshold for determining grid power imbalance at any given time; : Real-time adaptive dynamic correction coefficient; Power imbalance reference threshold; : Time-series fluctuation correction amount.

9. A dynamic load control method based on real-time power monitoring according to claim 7, characterized in that, The specific process of step S5 is as follows: S51: After data acquisition and control are completed Real-time total active power of the entire power grid After the adjustment is completed Real active power deficit of the power grid at all times Load after regulation Actual operating active power and load Actual power regulation ; S52: Obtain the optimal control target parameters, including The optimal steady-state total active power of the power grid at any given time , Optimal power deficit of the power grid at any time and load Optimal target power adjustment ; S53: Single Load Control Deviation Calculation: ; in, : Time load The single-load real-time control deviation value; :load Actual power regulation amount; :load Preset the optimal target power adjustment amount; S54: Calculation of Overall Network Control Deviation: ; in, : Real-time comprehensive control deviation value across the entire network; The actual active power deficit in the power grid after regulation; : The optimal power deficit of the power grid at any given time; S55: Perform reverse correction of adaptive dynamic threshold model parameters: Inverse correction of time-series fluctuation weighting coefficient: Based on comprehensive control deviation correction After being corrected Time-series fluctuation weighting coefficient ; Dynamic correction coefficient iterative optimization: Using the corrected time series fluctuation weighting coefficient, the time series fluctuation correction amount is recalculated, and the dynamic correction coefficient is iteratively updated to obtain the corrected time series fluctuation correction amount. and the corrected adaptive dynamic correction coefficient ; Dynamic threshold update: Based on the iteratively optimized correction coefficients, the dynamic power imbalance judgment threshold is updated to obtain the iteratively optimized dynamic power imbalance judgment threshold. : S56: Optimization of Load Control Execution and Precise Command Generation: Based on single load control deviations, the execution intensity of various types of load control is corrected, and targeted optimization is achieved. Flexible adjustment range and By removing the ratio of components, deviations in individual load regulation are avoided, and precise regulation commands are generated after iteration.

10. A dynamic load control system based on real-time power monitoring, used to implement the dynamic load control method based on real-time power monitoring as described in any one of claims 1-9, characterized in that, It includes an edge real-time sensing module, which is deployed in each branch circuit and terminal load equipment to synchronously collect multi-dimensional electrical parameters and perform preprocessing. The adaptive decision-making module is communicatively connected to the edge real-time perception module and is used to construct an adaptive dynamic threshold control model based on standardized real-time power data, historical time-series data and multiple influencing factors, and dynamically update the power safety threshold and control margin. The load classification and control module is communicatively connected to the adaptive decision module. It is used to classify and classify terminal loads, establish a differentiated control strategy library, and generate preliminary load control instructions in combination with the real-time operating conditions of the power grid. The collaborative execution module is connected to the adaptive decision-making module and the load hierarchical control module respectively, and is used to coordinate the power grid, load, distributed photovoltaic and energy storage equipment, execute the initial load control command, and complete the source-grid-load-storage collaborative control. The closed-loop feedback correction module is bidirectionally connected to the collaborative execution module. It is used to collect the power grid and load operation data after regulation, compare the regulation deviation, correct the regulation parameters, and at the same time run the fault tolerance mechanism to ensure the stable operation of the system.

Images