Load regulation and control system, method and device and computer readable storage medium

Abstract

The invention discloses a load regulation and control system, and the system comprises a cloud decision-making layer which is used for generating power regulation reference values corresponding to virtual controllable units in a future period according to each second adjustable power range and each dynamic adjustable potential curve, and transmitting the power regulation reference values to the corresponding virtual controllable units; the edge aggregation layer is used for calculating a first adjustable power range corresponding to each terminal according to the received data of each terminal; aggregating and generating a second adjustable power range and a dynamic adjustable potential curve respectively corresponding to each virtual controllable unit according to each first adjustable power range, and reporting the second adjustable power range and the dynamic adjustable potential curve to a cloud decision-making layer; generating a continuous non-inductive regulation factor corresponding to each terminal according to each power regulation reference value, and issuing the continuous non-inductive regulation factor to a terminal layer; and the terminal layer is used for performing power regulation and control on each terminal according to each continuous non-inductive regulation factor. According to the invention, the regulation and control mode is converted from rigid intervention to flexible non-inductive normal form, the power impact on the power grid is avoided, and the resource utilization rate is improved.

Description

Technical Field

[0001] This application relates to the field of power control technology, and in particular to a load regulation system, method, device and computer-readable storage medium. Background Technology

[0002] Air conditioning, electric vehicles, and smart home devices represent a massive potential source of flexible loads, considered crucial for peak shaving and valley filling in the power grid and promoting the integration of renewable energy. However, their widespread distribution, heterogeneous characteristics, random behavior, and direct impact on user electricity experience pose significant challenges to the unified regulation of the power grid.

[0003] Currently, most mainstream control methods rely on centralized aggregation scheduling or rigid control based on fixed strategies, which have obvious limitations: on the one hand, centralized decision-making mode has multiple communication layers and large response delays, making it difficult to meet the real-time and rapid adjustment needs of the power grid; on the other hand, simple one-size-fits-all start-stop control seriously sacrifices user comfort, resulting in low participation, and static aggregation models cannot accurately depict the dynamic adjustable potential of the load, resulting in insufficient resource utilization.

[0004] The control objectives are essentially centered on grid safety, and its decision-making model and utility function directly serve to correct abnormal grid indicators such as voltage exceedances and line overloads. This inevitably leads to a rigid control strategy that addresses grid "ailments," prioritizing rapid grid condition correction rather than ensuring user experience. Therefore, when controlling loads that directly affect comfort, such as air conditioners and water heaters, sudden power jumps or hard start-stop operations are easily implemented, ultimately degenerating into discrete start-stop or speed control, directly sacrificing user experience and contradicting the goal of imperceptible operation. Such abrupt changes are not only noticeable to users but may also cause power surges in the grid, creating a dilemma where solving one problem leads to new ones.

[0005] In summary, how to effectively address the problems of current load control methods, which not only cause noticeable effects on users but also lead to power surges and insufficient resource utilization in the power grid, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] The purpose of this application is to provide a load control system that realizes a paradigm shift from rigid intervention to flexible and seamless control, avoiding power surges to the power grid and improving resource utilization. Another purpose of this application is to provide a load control method, device and computer-readable storage medium.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: A load control system, comprising: The cloud-based decision layer receives load control instructions and sends load control commands to the edge aggregation layer according to the load control instructions; it generates power adjustment reference values ​​for the virtual controllable units corresponding to each edge node in the next cycle based on each second adjustable power range and each dynamic adjustable potential curve; and it sends each power adjustment reference value to the corresponding virtual controllable unit. The edge aggregation layer is used to receive terminal data uploaded by each terminal in the corresponding jurisdiction of the terminal layer through each virtual controllable unit; calculate the first adjustable power range corresponding to each terminal based on the terminal data; aggregate and generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit based on each first adjustable power range; report each second adjustable power range and each dynamic adjustable potential curve to the cloud decision layer; generate the continuous seamless adjustment factor corresponding to each terminal based on each power adjustment reference value, and send each continuous seamless adjustment factor to the terminal layer. The terminal layer is used to regulate the power of each terminal according to each continuous sensorless adjustment factor.

[0008] In one specific embodiment of this application, it is further configured to receive a comfort zone setting instruction, and set a comfort zone according to the comfort zone setting instruction; and upload the comfort zones corresponding to each terminal to the edge aggregation layer; The edge aggregation layer is specifically used to calculate the first adjustable power range corresponding to each terminal based on the data from each terminal and each comfort zone.

[0009] In one specific embodiment of this application, the edge aggregation layer is specifically used to calculate the personalized model parameters corresponding to each terminal based on the data of each terminal; and to calculate the first adjustable power range corresponding to each terminal based on the personalized model parameters and each comfort zone.

[0010] In one specific embodiment of this application, the terminal layer is further configured to collect the actual execution indicator set corresponding to each terminal and upload each actual execution indicator set to the edge aggregation layer; wherein, the actual execution indicator set includes actual execution power, actual comfort index and actual comfort over-limit flag; The edge aggregation layer is further used to obtain the predicted execution index set corresponding to each terminal, and to determine the deviation between each actual execution index set and the corresponding predicted execution index set; and to update each personalized model parameter according to each deviation; wherein, the predicted execution index set includes predicted execution power, predicted comfort index and predicted comfort over-limit flag; the predicted execution index set is an index set calculated based on the actual execution index set corresponding to the previous period and the current dynamic aggregation model parameters corresponding to the previous period.

[0011] In one specific embodiment of this application, the cloud decision layer is further configured to update the current dynamic aggregation model parameters according to each second adjustable power range to obtain updated dynamic aggregation model parameters, and determine the updated dynamic aggregation model parameters as the new current dynamic aggregation model parameters; and send the current dynamic aggregation model parameters to the edge aggregation layer.

[0012] In one specific embodiment of this application, the edge aggregation layer is specifically used to calculate the first adjustable power range corresponding to each terminal based on the current dynamic aggregation model parameters and the data of each terminal.

[0013] In one specific embodiment of this application, the terminal layer is specifically used to perform power regulation on each terminal based on each continuous seamless adjustment factor using a smooth execution algorithm.

[0014] A load control method, comprising: When a load control command is received, the virtual controllable unit corresponding to each edge node receives the terminal data uploaded by each terminal in the corresponding jurisdiction in the terminal layer. Calculate the first adjustable power range for each terminal based on the data from each terminal. Based on each first adjustable power range, the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit are generated. Each second adjustable power range and each dynamic adjustable potential curve are reported to the cloud decision layer; The system receives power adjustment reference values ​​for each virtual controllable unit within a future cycle, returned by the cloud-based decision-making layer. Each power adjustment reference value is a reference value generated by the cloud-based decision-making layer based on each second adjustable power range and each dynamic adjustable potential curve. Based on each power adjustment reference value, a continuous seamless adjustment factor is generated for each terminal, and each continuous seamless adjustment factor is sent to the terminal layer so that the terminal layer can adjust the power of each terminal according to each continuous seamless adjustment factor.

[0015] A load control device, comprising: Memory, used to store computer programs; A processor is used to implement the steps of the load control method as described above when executing the computer program.

[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the load control method described above.

[0017] The load control system provided in this application includes: a cloud decision layer, used to receive load control instructions and send load control commands to the edge aggregation layer according to the load control instructions; generate power adjustment reference values ​​for virtual controllable units corresponding to each edge node in a future cycle based on each second adjustable power range and each dynamic adjustable potential curve; and send each power adjustment reference value to the corresponding virtual controllable unit; an edge aggregation layer, used to receive terminal data uploaded by each terminal in the corresponding jurisdiction in the terminal layer through each virtual controllable unit; calculate the first adjustable power range corresponding to each terminal based on each terminal data; aggregate and generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit based on each first adjustable power range; report each second adjustable power range and each dynamic adjustable potential curve to the cloud decision layer; generate continuous seamless adjustment factors corresponding to each terminal based on each power adjustment reference value, and send each continuous seamless adjustment factor to the terminal layer; and a terminal layer, used to perform power control on each terminal based on each continuous seamless adjustment factor.

[0018] As can be seen from the above technical solution, at the edge layer, by identifying terminal characteristics in real time, the first adjustable power range corresponding to each terminal is dynamically calculated, resulting in the seamless adjustment range corresponding to each terminal. This is then aggregated to generate a precise dynamic adjustable potential curve for the virtual controllable unit. User comfort is established as a hard constraint for regulation from the source, thereby improving the accuracy of load regulation. The cloud decision layer generates power adjustment reference values ​​for each virtual controllable unit based on each second adjustable power range and each dynamic adjustable potential curve. The edge controller decomposes the power adjustment reference values ​​into continuous seamless adjustment factors for each terminal through a rigorous quadratic programming problem, ensuring that the changes in each continuous seamless adjustment factor are constrained within its own seamless adjustment range. This mechanism mathematically guarantees that macroscopic power grid commands are transformed into microscopic, comfort-strictly constrained, fine-tuning actions for massive loads. By constructing an innovative three-layer closed-loop collaborative architecture of cloud decision-making, edge aggregation, and terminal execution, a paradigm shift from rigid intervention to flexible seamless regulation is fundamentally achieved, avoiding power impact on the power grid and improving resource utilization.

[0019] Accordingly, this application also provides a load control method, device and computer-readable storage medium corresponding to the above-mentioned load control system, which have the above-mentioned technical effects, and will not be repeated here. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a structural block diagram of a load control system according to an embodiment of this application; Figure 2 This is an architectural diagram of another load control system in the embodiments of this application; Figure 3 This is a schematic diagram of the core processing flow of a local dynamic aggregation engine in an edge aggregation layer according to an embodiment of this application; Figure 4 This is a flowchart illustrating a two-level collaborative decision-making process in an embodiment of this application, involving global cycle optimization at the cloud decision layer and local real-time control at the edge aggregation layer. Figure 5 This is a flowchart illustrating the implementation of a load control method in an embodiment of this application. Figure 6 This is a structural block diagram of a load regulation device according to an embodiment of this application; Figure 7 This is a structural block diagram of a load control device according to an embodiment of this application; Figure 8 This is a schematic diagram of the specific structure of a load control device provided in an embodiment of this application.

[0022] The following labels are shown in the attached diagram: 1-Cloud decision-making layer, 2-Edge aggregation layer, 3-Terminal layer. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0025] See Figure 1 , Figure 1 This is a structural block diagram of a load control system according to an embodiment of this application. The system may include: The cloud-based decision layer 1 receives load control instructions and sends load control commands to the edge aggregation layer 2 according to the load control instructions; it generates power adjustment reference values ​​for the virtual controllable units corresponding to each edge node in the next cycle based on each second adjustable power range and each dynamic adjustable potential curve; and it sends each power adjustment reference value to the corresponding virtual controllable unit. Edge aggregation layer 2 is used to receive terminal data uploaded by each terminal (load) in the corresponding jurisdiction of terminal layer 3 through each virtual controllable unit; calculate the first adjustable power range corresponding to each terminal based on the terminal data; aggregate and generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit based on each first adjustable power range; report each second adjustable power range and each dynamic adjustable potential curve to cloud decision layer 1; generate the continuous seamless adjustment factor corresponding to each terminal based on each power adjustment reference value, and send each continuous seamless adjustment factor to terminal layer 3; Terminal layer 3 is used to regulate the power of each terminal according to each continuous sensorless adjustment factor.

[0026] The load control system provided in this application embodiment may include a cloud decision layer 1, an edge aggregation layer 2, and a cloud decision layer 3. The cloud decision layer 1 receives load control instructions and sends load control commands to the edge aggregation layer 2 according to the load control instructions. The edge aggregation layer 2 receives terminal data uploaded by each terminal in the corresponding jurisdiction of the terminal layer 3 through each Virtual Control Unit (VCU), calculates the first adjustable power range corresponding to each terminal according to the terminal data, aggregates the first adjustable power ranges to generate the second adjustable power range and dynamic adjustable potential curve corresponding to each Virtual Control Unit, and reports the second adjustable power range and the dynamic adjustable potential curve to the cloud decision layer 1.

[0027] The cloud-based decision-making layer 1 generates power adjustment reference values ​​for the virtual controllable units corresponding to each edge node within a future cycle based on the second adjustable power range and the dynamic adjustable potential curves. These reference values ​​are then distributed to the corresponding virtual controllable units. The edge aggregation layer 2 generates continuous, seamless adjustment factors for each terminal based on the power adjustment reference values ​​and distributes these factors to the terminal layer 3. The terminal layer 3 then adjusts the power of each terminal based on these continuous, seamless adjustment factors. By constructing an innovative three-layer closed-loop collaborative architecture of cloud-based decision-making, edge aggregation, and terminal execution, a fundamental shift in the control mode from rigid intervention to flexible, seamless operation is achieved, avoiding power surges to the power grid and improving resource utilization.

[0028] As can be seen from the above technical solution, at the edge layer, by identifying terminal characteristics in real time, the first adjustable power range corresponding to each terminal is dynamically calculated, resulting in the seamless adjustment range corresponding to each terminal. This is then aggregated to generate a precise dynamic adjustable potential curve for the virtual controllable unit. User comfort is established as a hard constraint for regulation from the source, thereby improving the accuracy of load regulation. The cloud decision layer generates power adjustment reference values ​​for each virtual controllable unit based on each second adjustable power range and each dynamic adjustable potential curve. The edge controller decomposes the power adjustment reference values ​​into continuous seamless adjustment factors for each terminal through a rigorous quadratic programming problem, ensuring that the changes in each continuous seamless adjustment factor are constrained within its own seamless adjustment range. This mechanism mathematically guarantees that macroscopic power grid commands are transformed into microscopic, comfort-strictly constrained, fine-tuning actions for massive loads. By constructing an innovative three-layer closed-loop collaborative architecture of cloud decision-making, edge aggregation, and terminal execution, a paradigm shift from rigid intervention to flexible seamless regulation is fundamentally achieved, avoiding power impact on the power grid and improving resource utilization.

[0029] It should be noted that, based on the above embodiments, this application also provides corresponding improvements. In subsequent embodiments, steps that are the same as or corresponding to those in the above embodiments can be referred to each other, and the corresponding beneficial effects can also be referred to each other. These improvements will not be elaborated upon in the following improved embodiments.

[0030] In one specific embodiment of this application, the terminal layer 3 is further configured to receive a comfort zone setting instruction, set a comfort zone according to the comfort zone setting instruction, and upload the comfort zone corresponding to each terminal to the edge aggregation layer 2. Edge aggregation layer 2 is specifically used to calculate the first adjustable power range corresponding to each terminal based on the data of each terminal and each comfort zone.

[0031] Terminal layer 3 is also used to receive comfort zone setting instructions, set comfort zones according to the instructions, and upload the comfort zones corresponding to each terminal to edge aggregation layer 2. Edge aggregation layer 2 calculates the first adjustable power range corresponding to each terminal based on the data of each terminal and each comfort zone. By calculating the first adjustable power range corresponding to each terminal based on the comfort zones corresponding to each terminal, user experience is guaranteed.

[0032] In one specific embodiment of this application, the edge aggregation layer 2 is specifically used to calculate the personalized model parameters corresponding to each terminal based on the data of each terminal; and to calculate the first adjustable power range corresponding to each terminal based on the personalized model parameters and each comfort zone.

[0033] The edge aggregation layer 2 is specifically used to calculate the personalized model parameters corresponding to each terminal based on the data from each terminal, and to calculate the first adjustable power range corresponding to each terminal based on the personalized model parameters and each comfort zone. By calculating the first adjustable power range corresponding to each terminal based on the personalized model parameters and each comfort zone corresponding to each terminal, the uniform adjustable power range setting for each terminal is avoided, the differences between terminals are taken into account, and the user experience is further improved.

[0034] In one specific embodiment of this application, the terminal layer 3 is further used to collect the actual execution index set corresponding to each terminal and upload each actual execution index set to the edge aggregation layer 2; wherein, the actual execution index set includes actual execution power, actual comfort index and actual comfort over-limit flag; Edge aggregation layer 2 is also used to obtain the predicted execution index set corresponding to each terminal, and to determine the deviation between each actual execution index set and the corresponding predicted execution index set; and to update each personalized model parameter according to each deviation; wherein, the predicted execution index set includes predicted execution power, predicted comfort index and predicted comfort over-limit flag; the predicted execution index set is an index set calculated based on the actual execution index set corresponding to the previous period and the current dynamic aggregation model parameter corresponding to the previous period.

[0035] A comfort protection module can be pre-set in the terminal to continuously monitor key indicators (such as indoor temperature). Once it is predicted or detected that the temperature will exceed the user's comfort range, the local protective coverage will be activated immediately, and the power will be restored to the minimum operating power required to ensure comfort. The boundary crossing indicator will be fed back to the edge aggregation layer 2.

[0036] Terminal layer 3 is also used to collect the actual execution indicator sets corresponding to each terminal. These sets include actual execution power, actual comfort index, and actual comfort exceedance flags. These actual execution indicator sets are then uploaded to edge aggregation layer 2. Edge aggregation layer 2 pre-calculates predicted execution indicator sets based on the actual execution indicator sets from the previous period and the parameters of the current dynamic aggregation model from the previous period. Edge aggregation layer 2 also obtains the predicted execution indicator sets corresponding to each terminal. These sets include predicted execution power, predicted comfort index, and predicted comfort exceedance flags. It then determines the deviation between each actual execution indicator set and its corresponding predicted execution indicator set, updating the parameters of each personalized model based on these deviations. By comparing the actual execution indicator sets with the predicted execution indicator sets and updating the personalized model parameters of the terminals based on the deviations, the model becomes increasingly accurate.

[0037] In one specific embodiment of this application, the cloud decision layer 1 is further configured to update the current dynamic aggregation model parameters according to each second adjustable power range, obtain the updated dynamic aggregation model parameters, and determine the updated dynamic aggregation model parameters as the new current dynamic aggregation model parameters; and send the current dynamic aggregation model parameters to the edge aggregation layer 2.

[0038] The cloud-based decision layer 1 is also used to update the current dynamic aggregation model parameters according to each of the second adjustable power ranges, obtain the updated dynamic aggregation model parameters, and determine the updated dynamic aggregation model parameters as the new current dynamic aggregation model parameters. These current dynamic aggregation model parameters are then distributed to the edge aggregation layer 2. By ensuring that the updated current dynamic aggregation model parameters are promptly distributed to the edge aggregation layer 2 by the cloud-based decision layer 1, the timely updating of the dynamic aggregation model parameters in the edge nodes is guaranteed, thereby ensuring the accuracy of load regulation.

[0039] In one specific embodiment of this application, the edge aggregation layer 2 is specifically used to calculate the first adjustable power range corresponding to each terminal based on the current dynamic aggregation model parameters and the data of each terminal.

[0040] Edge aggregation layer 2 is specifically used to calculate the first adjustable power range corresponding to each terminal based on the current dynamic aggregation model parameters and data from each terminal. By calculating the first adjustable power range corresponding to each terminal based on the latest current dynamic aggregation model parameters and data from each terminal, the accuracy of the calculated first adjustable power range is improved.

[0041] In one specific embodiment of this application, the terminal layer 3 is specifically used to perform power regulation on each terminal based on each continuous imperceptible adjustment factor using a smooth execution algorithm.

[0042] Terminal layer 3 is specifically used to perform power regulation on each terminal based on continuous imperceptible adjustment factors using a smooth execution algorithm. Through a continuous and smooth imperceptible adjustment mechanism, it achieves refined and user-friendly regulation of massive distributed flexible loads. By strictly constraining the changes of each continuous imperceptible adjustment factor within its imperceptible adjustment range, the actual power change is continuous and gradual, with the rate of change lower than the user-perceptible threshold, thereby ensuring that the total adjustment amount is within the local real-time aggregation potential range.

[0043] See Figure 2 , Figure 2 This is an architecture diagram of another load control system in this application embodiment. The load control system is clearly divided into three logical layers: cloud decision layer 1, edge aggregation layer 2, and terminal layer 3, with each layer forming a closed loop through bidirectional data flow.

[0044] Cloud-based decision-making layer 1, serving as the system's global intelligence center, is primarily responsible for macro-level strategy formulation and model management. Its core components include: a power grid dispatch center interface, which receives regulation demand instructions from the power grid dispatch center, such as peak-shaving power targets. Or, a valley-filling power target; a global dynamic aggregation model, which is a benchmark model trained on historical big data, used to describe the general physical characteristics and behavioral patterns of various flexible loads (such as air conditioners and electric vehicles). This model periodically updates its parameters. The data is distributed to each edge node, providing prior knowledge for accurate local modeling. The global policy optimizer employs advanced algorithms such as Deep Reinforcement Learning (DRL) to perform periodic (e.g., every 15 minutes) optimization calculations, aiming for the economical and secure operation of the entire network. Its goal is to generate power regulation reference values ​​for the virtual controllable units under the jurisdiction of each edge region within a future cycle. The objective function for optimization takes into account the power grid purchase cost, grid loss, and the smoothness of load regulation.

[0045] Edge aggregation layer 2 is the regional neural center of the system, distributed at the network edge of distribution transformer areas, intelligent buildings, etc. Its main function is to realize real-time sensing, dynamic aggregation, and rapid control of local loads. The local dynamic aggregation engine receives baseline model parameters from the cloud. And real-time data from terminal layer 3, which is responsible for dynamically identifying the real-time characteristics of each flexible load in the jurisdiction, calculating its first adjustable power range under user-unnoticed conditions, and aggregating and generating the second adjustable power range of the virtual controllable unit in real time. and dynamically adjustable potential curve This curve is a key input for subsequent real-time decision-making. The real-time collaborative controller receives power adjustment reference values ​​from the cloud. Its core task is to perform rapid optimization calculations at the second or sub-second level, accurately and smoothly decomposing the global power command into continuous, imperceptible adjustment factors for each specific load within the jurisdiction. And ensure that the decomposition process strictly follows the real-time imperceptible constraints and local total regulation potential of each load.

[0046] Terminal layer 3 consists of a massive number of flexible load terminals with communication and control functions. Examples of such devices include smart air conditioners, electric vehicle charging stations, and instant water heaters. Each terminal receives continuous, seamless adjustment factors from its associated edge controller. The system incorporates a built-in smooth power modulation algorithm and a local comfort protection module to ensure that power adjustments are continuous, gradual, and do not exceed user comfort limits. It also feeds back its real-time operating status (such as actual power and temperature) and any comfort-limit-crossing alarm signals to edge aggregation layer 2, forming a closed loop. The system data flow forms a complete loop: terminal data is sent to edge aggregation layer 2; after aggregation at the edge layer, the status and potential information of the virtual controllable unit are sent to cloud decision layer 1; cloud decision layer 1 optimizes and distributes strategies and model parameters; edge aggregation layer 2 decomposes the power adjustment reference value and sends it to the terminal for execution; the terminal's execution status is then fed back to edge aggregation layer 2. This closed loop enables the system to continuously learn and optimize.

[0047] See Figure 3 , Figure 3 This diagram illustrates the core processing flow of a local dynamic aggregation engine in an edge aggregation layer 2 according to an embodiment of this application. The goal is to dynamically extract precise and usable group control capabilities from massive, heterogeneous real-time data. The input to the local dynamic aggregation engine is multi-source and heterogeneous, including: real-time terminal data, such as data from each terminal... Instantaneous operating power Equipment status Related physical quantities (such as indoor temperature) Battery SOC User preferences (such as the set temperature of the air conditioner) and the acceptable fluctuation range for users The expected charging time of the charging station Cloud-based benchmark model parameters (Provide initial parameters for the general load model).

[0048] The core processing involves three key steps: real-time identification of terminal characteristics. The system uses real-time data streams to dynamically update the personalized model parameters of each terminal. Taking the dominant temperature control terminal (air conditioner) as an example, its simplified first-order equivalent thermal parameter model is as follows: ; in, It is a terminal exist Indoor temperature at all times for outdoor temperature at any time For the terminal Thermal resistance of the room For the terminal The heat capacity of the room For the terminal Energy efficiency ratio, Indicates terminal exist Power consumption at any moment To control the time step, continuous data is collected. , and Data, and by applying recursive least squares or Kalman filtering algorithms, can be estimated and updated online in real time. , By analyzing key parameters such as load temperature change, an accurate predictive model for load temperature changes can be obtained.

[0049] Seamless adjustment range calculation. Based on the updated load model and user preferences, calculate the upper and lower limits of power adjustment for each terminal at the current moment without affecting user experience. For air conditioners in cooling mode, to ensure room temperature... Always maintain within the user's comfort zone Within this range, the adjustable range of its power is dynamically determined by the thermodynamic equilibrium relationship: ; ; in, , For the terminal exist The power adjustable lower and upper limits at any given time. For the terminal Rated power, For users to access the terminal The set temperature, For users to access the terminal The above formula dynamically defines the seamless adjustment range for each terminal from the perspective of ensuring user comfort, allowing for acceptable temperature fluctuations.

[0050] Adjustable potential is dynamically aggregated. All within the jurisdiction... The real-time adjustable ranges of each terminal are aggregated to obtain the real-time adjustment capability of the entire virtual controllable unit: ; ; in, For virtual controllable units in The total potential for power increase at any given moment, that is, the sum of the power that all terminals can still increase. For virtual controllable units in The total potential for power reduction at any given time, i.e., the sum of power reductions that can be achieved across all terminals. For the terminal exist Actual operating power at any given time.

[0051] Furthermore, the model can extrapolate potential trends over a future period based on ultra-short-term predictions (such as outdoor temperature in the next 15 minutes and user behavior probabilities), forming a dynamically adjustable potential curve. .

[0052] The module's final output consists of two parts: the real-time total power of the virtual controllable unit and its dynamic adjustable potential curve, providing an accurate resource inventory for cloud-based global optimization and edge real-time control.

[0053] See Figure 4 , Figure 4 This is a flowchart illustrating a two-level collaborative decision-making process in this embodiment, involving global cycle optimization at the cloud-based decision layer 1 and local real-time control at the edge aggregation layer 2. The cloud-based cycle optimization process involves: the cycle reaching its end, or the power grid having new adjustment needs. Triggered upon arrival. The cloud collects reports from all edge nodes across the network regarding the second adjustable power range. and dynamically adjustable potential curve The following problem is solved using a cloud-based deep reinforcement learning agent or centralized optimization algorithm: assigning a power regulation reference value to each virtual controllable unit within a future cycle. .

[0054] The total operating cost (electricity purchase cost, grid loss) and over-limit penalty are minimized by the following formula: ; The following formula ensures that the overall regulation meets the grid demand: ; The following formula ensures that the issued instructions do not exceed the upper limit of the adjustment capacity reported from the edge: ; in, for Real-time grid electricity price for The power purchased from the upper-level power grid at all times For the corresponding line The network loss coefficient, This is the power over-limit penalty factor. The maximum allowable load power for the area. This represents the total number of virtual controllable units in the system. Virtual controllable unit exist The forecasts reported in real time can be revised upwards.

[0055] After optimization, the generated global policy Distribute to the corresponding edge nodes.

[0056] The edge real-time control process continuously receives real-time data from the terminal and commands from the cloud. In the local real-time optimization model, the edge controller solves a fast quadratic programming problem within each short control cycle (e.g., 2 seconds) to accurately track cloud commands while minimizing sudden changes in load power. The following formula achieves precise instruction tracking and smooth motion: ; Power aggregation is performed using the following formula; ; The following formula represents the core seamless constraint, ensuring that the adjustment of each load is within its comfort range, with a continuous adjustment factor. The range is dynamically constrained by the following formula: ; The following formula represents the local potential constraint, ensuring that the total adjustment remains within the real-time capacity: ; in, It is pending, targeting the terminal. The continuous regulating factor, dimensionless. It is the smoothing weighting coefficient, used to penalize drastic changes in the adjustment factor between adjacent periods. For real-time total potential for upward adjustment, It represents the real-time potential for adjustment. The optimal set of adjustment factors is obtained by performing a solution. And immediately send it to the corresponding load terminal.

[0057] The terminal receives the continuous adjustment factor sent by the edge controller. To avoid sudden power changes that might be perceived by users or cause damage to equipment, the terminal's built-in smooth execution algorithm filters the commands. A first-order inertial element is used to achieve a smooth power transition. ; in, For the terminal exist The final execution power at any given moment. This is a smoothing coefficient, the value of which is set based on the electromechanical inertia of the equipment, used to limit the rate of power change. This will cause it to fall below the perceptible threshold.

[0058] Then, comfort protection and over-limit handling are performed, and the equipment performs power smoothing. Simultaneously, key comfort indicators are continuously monitored. Once it is predicted or detected that an indicator will exceed the user's preset comfort range, local protective coverage is immediately activated to prioritize user experience and generate a boundary violation flag. For air conditioners, the protection logic is as follows: ; ; in, To provide a buffer margin for preventative control, The minimum cooling or heating power required to maintain the temperature at the boundary of the comfort range. The comfort level indicator shows whether the comfort level has exceeded the limit; 1 indicates that the limit has been exceeded, and 0 indicates that the comfort level is normal.

[0059] Subsequently, state feedback and model iterative optimization are performed, and the terminal will then calculate the actual execution power. Actual measured comfort index and boundary crossing signs Feedback is fed to edge aggregation layer 2. Edge aggregation layer 2 uses the discrepancy between the actual feedback data and the model's predictions to continuously calibrate and update the terminal characteristic parameters in the local dynamic aggregation model, making the model increasingly accurate. The aggregated comfort violation information is uploaded to the cloud. When updating its global deep reinforcement learning policy network, the cloud incorporates user experience violations as an important penalty in its reward function. In the design: ; in, These are the weighting coefficients for power grid target tracking, command tracking, and user experience, respectively. It is the set of all regulated loads in the system.

[0060] Through the above formula, the system's core intelligent deep reinforcement learning is clearly taught that while pursuing power grid regulation goals, user experience must be given high priority. Any strategy that causes user discomfort will be penalized, thereby driving the system to learn better regulation strategies that take into account the interests of all parties. This forms a complete closed loop of "data perception - strategy optimization - instruction decomposition - seamless execution - feedback learning." This closed loop enables the system to continuously optimize itself using operational data, forming a virtuous cycle of becoming smarter with use.

[0061] Corresponding to the above system embodiments, this application also provides a load control method, and the load control method described below can be referred to in correspondence with the load control system described above.

[0062] See Figure 5 , Figure 5 This is a flowchart illustrating an implementation of a load control method according to an embodiment of this application. The method may include the following steps: S501: When a load control command is received, the virtual controllable unit corresponding to each edge node receives the terminal data uploaded by each terminal in the corresponding jurisdiction in the terminal layer.

[0063] S502: Calculate the first adjustable power range corresponding to each terminal based on the data from each terminal.

[0064] In one specific embodiment of this application, step S502 may include the following steps: Step 1: Receive the comfort zones corresponding to each terminal uploaded by the terminal layer; where each comfort zone is the comfort zone set by the terminal layer according to the received comfort zone setting instructions; Step 2: Calculate the first adjustable power range for each terminal based on the data from each terminal and each comfort zone.

[0065] In one specific embodiment of this application, step S502 may include the following steps: Step 1: Calculate the personalized model parameters for each terminal based on the data from each terminal; Step 2: Calculate the first adjustable power range for each terminal based on the parameters of each personalized model and each comfort zone.

[0066] In one specific embodiment of this application, step S502 may include the following steps: The first adjustable power range for each terminal is calculated based on the current dynamic aggregation model parameters and data from each terminal.

[0067] S503: Generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit based on each first adjustable power range.

[0068] S504: Report each second adjustable power range and each dynamic adjustable potential curve to the cloud decision layer.

[0069] S505: Receives the power adjustment reference values ​​for each virtual controllable unit within a future cycle, returned by the cloud decision-making layer.

[0070] Among them, each power adjustment reference value is a reference value generated by the cloud decision layer based on each second adjustable power range and each dynamic adjustable potential curve.

[0071] S506: Generate a continuous seamless adjustment factor for each terminal based on each power adjustment reference value, and send each continuous seamless adjustment factor to the terminal layer so that the terminal layer can adjust the power of each terminal according to each continuous seamless adjustment factor.

[0072] As can be seen from the above technical solution, at the edge layer, by identifying terminal characteristics in real time, the first adjustable power range corresponding to each terminal is dynamically calculated, resulting in the seamless adjustment range corresponding to each terminal. This is then aggregated to generate a precise dynamic adjustable potential curve for the virtual controllable unit. User comfort is established as a hard constraint for regulation from the source, thereby improving the accuracy of load regulation. The cloud decision layer generates power adjustment reference values ​​for each virtual controllable unit based on each second adjustable power range and each dynamic adjustable potential curve. The edge controller decomposes the power adjustment reference values ​​into continuous seamless adjustment factors for each terminal through a rigorous quadratic programming problem, ensuring that the changes in each continuous seamless adjustment factor are constrained within its own seamless adjustment range. This mechanism mathematically guarantees that macroscopic power grid commands are transformed into microscopic, comfort-strictly constrained, fine-tuning actions for massive loads. By constructing an innovative three-layer closed-loop collaborative architecture of cloud decision-making, edge aggregation, and terminal execution, a paradigm shift from rigid intervention to flexible seamless regulation is fundamentally achieved, avoiding power impact on the power grid and improving resource utilization.

[0073] In one specific embodiment of this application, the method may further include the following steps: Step 1: Receive the actual execution indicator set corresponding to each terminal uploaded by the terminal layer; wherein, the actual execution indicator set includes actual execution power, actual comfort index and actual comfort over-limit flag; Step 2: Obtain the predicted execution indicator set corresponding to each terminal, and determine the deviation between each actual execution indicator set and the corresponding predicted execution indicator set; Step 3: Update the parameters of each personalized model based on each deviation; the predicted execution index set includes predicted execution power, predicted comfort index, and predicted comfort overshoot flag; the predicted execution index set is an index set calculated based on the actual execution index set corresponding to the previous period and the current dynamic aggregation model parameters corresponding to the previous period.

[0074] In one specific embodiment of this application, the method may further include the following steps: Receive the updated current dynamic aggregation model parameters issued by the cloud decision layer; wherein, the updated current dynamic aggregation model parameters are the model parameters obtained by the cloud decision layer based on each second adjustable power range to update the current dynamic aggregation model parameters.

[0075] Corresponding to the above system embodiments, this application also provides a load control device, and the load control device described below can be referred to in correspondence with the load control system described above.

[0076] See Figure 6 , Figure 6 This is a structural block diagram of a load regulation device according to an embodiment of this application. The device may include: The terminal data receiving module 61 is used to receive terminal data uploaded by each terminal in the corresponding jurisdiction in the terminal layer through the virtual controllable unit corresponding to each edge node when a load control command is received. The first adjustable power range calculation module 62 is used to calculate the first adjustable power range corresponding to each terminal based on the data of each terminal. The dynamic adjustable potential curve generation module 63 is used to aggregate and generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit according to each first adjustable power range. The adjustable power range and curve uploading module 64 is used to report each second adjustable power range and each dynamic adjustable potential curve to the cloud decision layer. The power adjustment reference value receiving module 65 is used to receive the power adjustment reference values ​​corresponding to each virtual controllable unit in a future cycle returned by the cloud decision layer; wherein, each power adjustment reference value is a reference value generated by the cloud decision layer based on each second adjustable power range and each dynamic adjustable potential curve; The power regulation module 66 is used to generate a continuous and imperceptible regulation factor corresponding to each terminal based on each power regulation reference value, and to send each continuous and imperceptible regulation factor to the terminal layer so that the terminal layer can perform power regulation on each terminal based on each continuous and imperceptible regulation factor.

[0077] As can be seen from the above technical solution, at the edge layer, by identifying terminal characteristics in real time, the first adjustable power range corresponding to each terminal is dynamically calculated, resulting in the seamless adjustment range corresponding to each terminal. This is then aggregated to generate a precise dynamic adjustable potential curve for the virtual controllable unit. User comfort is established as a hard constraint for regulation from the source, thereby improving the accuracy of load regulation. The cloud decision layer generates power adjustment reference values ​​for each virtual controllable unit based on each second adjustable power range and each dynamic adjustable potential curve. The edge controller decomposes the power adjustment reference values ​​into continuous seamless adjustment factors for each terminal through a rigorous quadratic programming problem, ensuring that the changes in each continuous seamless adjustment factor are constrained within its own seamless adjustment range. This mechanism mathematically guarantees that macroscopic power grid commands are transformed into microscopic, comfort-strictly constrained, fine-tuning actions for massive loads. By constructing an innovative three-layer closed-loop collaborative architecture of cloud decision-making, edge aggregation, and terminal execution, a paradigm shift from rigid intervention to flexible seamless regulation is fundamentally achieved, avoiding power impact on the power grid and improving resource utilization.

[0078] In one specific embodiment of this application, the first adjustable power range calculation module 62 may include: The comfort zone receiving submodule is used to receive the comfort zones corresponding to each terminal uploaded by the terminal layer; wherein, each comfort zone is the comfort zone set by the terminal layer according to the received comfort zone setting instruction; The first range calculation submodule is used to calculate the first adjustable power range corresponding to each terminal based on the data of each terminal and each comfort zone.

[0079] In one specific embodiment of this application, the first adjustable power range calculation module 62 may include: The personalized model parameter calculation submodule is used to calculate the personalized model parameters corresponding to each terminal based on the data of each terminal. The second range calculation submodule is used to calculate the first adjustable power range corresponding to each terminal based on the parameters of each personalized model and each comfort zone.

[0080] In one specific embodiment of this application, the device may further include: The actual execution indicator set receiving module is used to receive the actual execution indicator sets corresponding to each terminal uploaded by the terminal layer; wherein, the actual execution indicator set includes actual execution power, actual comfort index and actual comfort over-limit flag; The deviation determination module is used to obtain the predicted execution indicator set corresponding to each terminal, and to determine the deviation between each actual execution indicator set and the corresponding predicted execution indicator set. The parameter update module is used to update the parameters of each personalized model according to each deviation; among them, the predicted execution index set includes predicted execution power, predicted comfort index and predicted comfort overshoot flag; the predicted execution index set is an index set calculated based on the actual execution index set corresponding to the previous period and the current dynamic aggregation model parameters corresponding to the previous period.

[0081] In one specific embodiment of this application, the device may further include: The parameter receiving module is used to receive the updated current dynamic aggregation model parameters issued by the cloud decision layer; wherein, the updated current dynamic aggregation model parameters are the model parameters obtained by the cloud decision layer updating the current dynamic aggregation model parameters according to each second adjustable power range.

[0082] In one specific embodiment of this application, the first adjustable power range calculation module 62 is specifically a module that calculates the first adjustable power range corresponding to each terminal based on the current dynamic aggregation model parameters and the data of each terminal.

[0083] For the method embodiments described above, see [link to relevant documentation]. Figure 7 , Figure 7 This is a schematic diagram of the load control equipment provided in this application, which may include: Memory 332 is used to store computer programs; The processor 322 is used to implement the steps of the load control method of the above method embodiment when executing a computer program.

[0084] For details, please refer to Figure 8 , Figure 8 This is a schematic diagram illustrating the specific structure of a load control device provided in this embodiment. The load control device can vary significantly due to different configurations or performance characteristics. It may include a processor (central processing unit, CPU) 322 (e.g., one or more processors) and a memory 332. The memory 332 stores one or more computer programs 342 or data 344. The memory 332 can be temporary or permanent storage. The program stored in the memory 332 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the data processing device. Furthermore, the processor 322 may be configured to communicate with the memory 332 and execute the series of instruction operations stored in the memory 332 on the load control device 301.

[0085] The load control device 301 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input / output interfaces 358, and / or one or more operating systems 341.

[0086] The steps in the load control method described above can be implemented by the structure of the load control equipment.

[0087] Corresponding to the above method embodiments, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps: When a load control command is received, the virtual controllable units corresponding to each edge node receive terminal data uploaded by each terminal within the corresponding jurisdiction in the terminal layer; calculate the first adjustable power range corresponding to each terminal based on the terminal data; aggregate the first adjustable power ranges to generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit; report the second adjustable power ranges and dynamic adjustable potential curves to the cloud decision layer; receive the power adjustment reference values ​​corresponding to each virtual controllable unit for the next cycle returned by the cloud decision layer; wherein, each power adjustment reference value is a reference value generated by the cloud decision layer based on each second adjustable power range and dynamic adjustable potential curve; generate a continuous seamless adjustment factor corresponding to each terminal based on each power adjustment reference value, and send the continuous seamless adjustment factors to the terminal layer so that the terminal layer can perform power control on each terminal based on the continuous seamless adjustment factors.

[0088] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0089] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0090] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The methods, devices, and computer-readable storage media disclosed in the embodiments are described simply because they correspond to the systems disclosed in the embodiments; relevant details can be found in the system section.

[0091] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the technical solutions and core ideas of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A load control system, characterized in that, include: The cloud decision layer (1) is used to receive load control instructions and send load control commands to the edge aggregation layer (2) according to the load control instructions; Based on each second adjustable power range and each dynamic adjustable potential curve, generate power adjustment reference values ​​for each virtual controllable unit corresponding to each edge node in the next cycle; and send each power adjustment reference value to the corresponding virtual controllable unit. The edge aggregation layer (2) is used to receive terminal data uploaded by each terminal in the corresponding jurisdiction of the terminal layer (3) through each virtual controllable unit; calculate the first adjustable power range corresponding to each terminal according to the terminal data; aggregate and generate the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit according to each first adjustable power range; report each second adjustable power range and each dynamic adjustable potential curve to the cloud decision layer (1); generate the continuous seamless adjustment factor corresponding to each terminal according to each power adjustment reference value, and send each continuous seamless adjustment factor to the terminal layer (3). The terminal layer (3) is used to regulate the power of each terminal according to each continuous sensorless adjustment factor.

2. The load control system according to claim 1, characterized in that, The terminal layer (3) is also used to receive comfort zone setting instructions and set comfort zones according to the comfort zone setting instructions; and upload the comfort zones corresponding to each terminal to the edge aggregation layer (2). The edge aggregation layer (2) is specifically used to calculate the first adjustable power range corresponding to each terminal based on the data of each terminal and each comfort zone.

3. The load control system according to claim 2, characterized in that, The edge aggregation layer (2) is specifically used to calculate the personalized model parameters corresponding to each terminal based on the data of each terminal; and to calculate the first adjustable power range corresponding to each terminal based on the personalized model parameters and each comfort zone.

4. The load control system according to claim 3, characterized in that, The terminal layer (3) is also used to collect the actual execution index set corresponding to each terminal and upload each actual execution index set to the edge aggregation layer (2); wherein, the actual execution index set includes actual execution power, actual comfort index and actual comfort over-limit flag; The edge aggregation layer (2) is also used to obtain the predicted execution index set corresponding to each terminal, and to determine the deviation between each actual execution index set and the corresponding predicted execution index set; and to update each personalized model parameter according to each deviation; wherein, the predicted execution index set includes predicted execution power, predicted comfort index and predicted comfort boundary flag; the predicted execution index set is an index set calculated based on the actual execution index set corresponding to the previous period and the current dynamic aggregation model parameter corresponding to the previous period.

5. The load control system according to claim 1, characterized in that, The cloud decision layer (1) is also used to update the current dynamic aggregation model parameters according to each second adjustable power range, obtain the updated dynamic aggregation model parameters, and determine the updated dynamic aggregation model parameters as the new current dynamic aggregation model parameters; and send the current dynamic aggregation model parameters to the edge aggregation layer (2).

6. The load control system according to claim 5, characterized in that, The edge aggregation layer (2) is specifically used to calculate the first adjustable power range corresponding to each terminal based on the current dynamic aggregation model parameters and the data of each terminal.

7. The load control system according to claim 1, characterized in that, The terminal layer (3) is specifically used to perform power regulation on each terminal based on each continuous imperceptible adjustment factor using a smooth execution algorithm.

8. A load control method, characterized in that, include: When a load control command is received, the virtual controllable unit corresponding to each edge node receives the terminal data uploaded by each terminal in the corresponding jurisdiction in the terminal layer. Calculate the first adjustable power range for each terminal based on the data from each terminal. Based on each first adjustable power range, the second adjustable power range and dynamic adjustable potential curve corresponding to each virtual controllable unit are generated. Each second adjustable power range and each dynamic adjustable potential curve are reported to the cloud decision layer; The system receives power adjustment reference values ​​for each virtual controllable unit within a future cycle, returned by the cloud-based decision-making layer. Each power adjustment reference value is a reference value generated by the cloud-based decision-making layer based on each second adjustable power range and each dynamic adjustable potential curve. Based on each power adjustment reference value, a continuous seamless adjustment factor is generated for each terminal, and each continuous seamless adjustment factor is sent to the terminal layer so that the terminal layer can adjust the power of each terminal according to each continuous seamless adjustment factor.

9. A load control device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the load control method as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the load control method as described in claim 8.

Images