Flexible load regulation and control method and system

Abstract

The invention provides a flexible load regulation and control method and system, and relates to the technical field of power distribution and utilization of a power system. Comprising the following steps: S1, acquiring electric power, indoor temperature and environment temperature data of each flexible load user, preprocessing the data, and uploading the data to an edge calculation layer; s2, the edge calculation layer obtains a transformer area total load curve in a specific time period in the future by using the data after data preprocessing and adopting an ARIMA time sequence prediction model; s3, the edge calculation layer compares the total load curve with the upper limit of the safety load of the transformer area through a decision control module, judges whether an overload risk exists or not, and generates a cooperative control strategy according to the risk level; s4, the user side execution equipment receives the cooperative control strategy and adjusts the operation state of the flexible load equipment, and meanwhile, the actual load and temperature changes are monitored; systematized and intelligent management of the electric heating load of the transformer area is achieved, and a fine regulation and control mechanism of multi-dimensional cooperation and flexible adjustment is constructed.

Description

Technical Field

[0001] This invention relates to the field of power system distribution technology, and in particular to a flexible load control method and system. Background Technology

[0002] With the deepening of clean heating policies such as "coal-to-electricity" conversion, the penetration rate of electric heating equipment (such as air source heat pumps, thermal storage electric heaters, and direct-heating electric heaters) in power distribution areas, especially in rural areas and urban-rural fringe areas, has increased dramatically. These devices are generally characterized by high power and concentrated operating time. Their load characteristics highly overlap with the daily electricity load of residents, especially during the morning and evening peak hours in winter. This can easily lead to instantaneous load spikes in power distribution areas, posing a serious challenge to distribution transformers, lines, and other facilities, and causing safety problems such as overload and low voltage.

[0003] Currently, the regulation and management of electric heating loads suffers from the following main deficiencies: 1. Existing electric heating equipment is mostly set to start and stop independently by users, lacking unified coordination. The power grid lacks effective proactive intervention methods, typically only resorting to simple time-of-use pricing guidance after overload occurs, or ultimately being forced to adopt mandatory measures such as "power rationing." This approach is slow to respond, unable to prevent risks, and severely sacrifices users' heating comfort, impacting people's livelihoods. 2. Existing technical solutions generally lack the ability to predict the overall load situation of the distribution area, especially short-term load changes. Due to the inability to predict peak loads in advance, control decisions lack foresight, only passively responding to overload events that have already occurred, missing the optimal window for regulation. 3. Traditional load control methods often treat each electric heating device as an independent entity, performing isolated "point-to-point" control, lacking a systemic perspective that aggregates dispersed loads into a dispatchable, flexible resource pool. This results in low control efficiency, an inability to achieve collaborative optimization between groups, and difficulty in achieving a refined balance between ensuring power grid security and maintaining user comfort. Summary of the Invention

[0004] The main objective of this invention is to provide a flexible load control method and system that solves the technical problems of existing control methods being passive and inefficient, having low accuracy in load prediction, low control efficiency, and limited applicability.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a flexible load regulation method, comprising the following steps: S1: Deploy data acquisition equipment in the distribution area to obtain power consumption, indoor temperature and ambient temperature data of each flexible load user, and upload the data to the edge computing layer after data preprocessing; S2: The edge computing layer uses the load forecasting module to obtain the total load curve of the transformer area for a specific future period based on historical load data, real-time load data and weather forecast information, and adopts the ARIMA time series forecasting model. S3: The edge computing layer compares the total load curve with the safe load limit of the transformer area through the decision control module to determine whether there is an overload risk. If there is an overload risk, a collaborative control strategy is generated according to the risk level. S4: The user-side execution device receives the collaborative control strategy and adjusts the operating status of the flexible load equipment, while monitoring the actual load and temperature changes. If the overload risk is eliminated, the normal operation of the flexible load equipment is restored.

[0006] In the preferred embodiment, the data preprocessing in S1 includes: Outlier Handling: The raw load data is screened for thresholds. If a load value exceeds 1.5 times the maximum load value for the same period in the past 7 days, it is considered an outlier and replaced with the average load value of the previous hour. The formula is as follows: (1.1); In the formula, t-15k represents k 15-minute intervals before time t; Missing value handling: For load data with a missing duration of no more than 30 minutes, imputation is performed using the load average at one time granularity before and after the missing time; for data with a missing duration of more than 30 minutes, the data is deleted, using the following formula: (1.2); In the formula, t-15 represents 15 minutes before time t, and t+15 represents 15 minutes after time t; Special load screening: Using Pearson correlation analysis, features weakly correlated with the load are removed, retaining only strongly correlated core features. The formula is as follows: (1.3); In the formula, , These are the feature values ​​(such as humidity, wind speed, and holiday indicators) and load values ​​of the i-th sample, respectively. , These are the sample means of the corresponding variables; N is the number of samples.

[0007] In a preferred embodiment, S2 includes: S21: Input the preprocessed data into the ARIMA time series prediction model, and obtain a stationary series through differencing, using the following formula: (1.4); In the formula, t-15 represents 15 minutes before time t. It is sequence data; S22: Determine the load change based on the autoregressive coefficient and the moving average coefficient, where the autoregressive coefficient represents the weight of the impact of historical load changes on the future, and the moving average coefficient represents the correction weight for the prediction error. The formula is as follows: (1.5); In the formula: t is the current time. For constant terms, , They represent the autoregressive coefficients, Represents the moving average coefficient. Here, h represents the error term and the prediction duration. S23: The total load curve is obtained by reconstructing the total load curve based on the load change and the actual load at the previous moment, wherein the total load curve is a time-series load sequence, and the formula is: (1.6); In the formula, This refers to the change in load. S24: Train the parameters of the time series prediction model based on the historical load data using the least squares method, and update the parameters weekly to obtain the updated model, as shown in the formula: (1.7); In the formula, N is the number of samples in the training set.

[0008] In a preferred embodiment, S3 includes: S31: The upper limit of the safe load of the distribution area is determined based on the rated capacity of the distribution transformer, the safe load rate of the transformer, the rated current carrying capacity of the main line of the distribution area, the average supply voltage of the distribution area, and the average power factor of the distribution area. The formula is as follows: (1.9); In the formula, S is the rated capacity of the distribution transformer, η is the safe load rate of the transformer, I is the rated current carrying capacity of the main line of the distribution area, and U is the average power supply voltage of the distribution area. The average power factor of the transformer area; S32: Compare the predicted load in the total load curve with a preset multiple of the safe load limit of the transformer area. If the predicted load exceeds the preset multiple, it is determined that there is an overload risk, and the risk level is classified according to the degree of exceedance, using the following formula: (1.10); S33: If an overload risk is determined, a combination of control strategies is selected based on the risk level, wherein the combination of control strategies is determined from fair round-robin control, priority control, and temperature range control.

[0009] In the preferred embodiment, the fair round-robin control for selecting a combination of control strategies includes: For the fair rotation control, flexible load equipment is grouped according to geographical region and equipment power to obtain multiple groups, and the control duration and control interval are set, wherein the control duration represents the duration of a single power reduction or pause, and the control interval represents the time between two controls in the same group; The fair round-robin control is executed cyclically according to the group to obtain the round-robin sequence, wherein the round-robin sequence is determined by sequentially adjusting each group and restoring the operation of the previous group.

[0010] In the preferred embodiment, the priority control of selecting the combination of control strategies includes: The priority control is based on the user's temperature sensitivity and power consumption duration to divide the priority level into high-priority users and low-priority users. The high-priority users correspond to a lower control frequency, and the low-priority users correspond to a higher control frequency. The priority level is updated periodically based on power consumption data.

[0011] In the preferred embodiment, the temperature range control for selecting the combination of control strategies is determined, including: For the temperature range control, a comfortable temperature range is determined, and the user-set temperature is adjusted to the lower limit of the comfortable temperature range to obtain the adjusted target temperature, wherein the adjusted target temperature is determined by extending the equipment downtime; If the risk level is high risk, then a dual regulation is obtained by combining the adjusted target temperature with a power reduction, wherein the power reduction is determined by reducing the operating power of the equipment. If the indoor temperature is monitored by the user-side execution device and is lower than a preset threshold, the temperature range control is stopped and the original user-set temperature is restored.

[0012] In the preferred embodiment, the S3 step of generating a collaborative control strategy based on the risk level includes: If the risk level is low risk, a strategy based on temperature range control is generated, wherein the temperature range control is obtained by adjusting the target temperature to the lower limit of the comfort range, and the original target temperature is maintained for high-priority users in the priority control. If the risk level is high risk, a combined strategy of round-robin control, priority control and temperature range control is generated. The round-robin control reduces power by performing grouped cyclic execution, the priority control adjusts the control frequency according to the user priority, and the temperature range control adjusts the target temperature before the round-robin. If the risk level is extreme risk, a strategy based on priority control and round-robin control is generated. The priority control increases the intensity of regulation for low-priority users, and the round-robin control pauses execution by shortening the interval and is supplemented by temperature range control to adjust to a lower limit.

[0013] In the preferred embodiment, step S4 includes: The collaborative control strategy is obtained from the edge computing layer via a communication module, wherein the collaborative control strategy includes temperature adjustment instructions or start / stop instructions. The set temperature or operating power of the flexible load equipment is adjusted according to the collaborative control strategy to obtain the adjusted operating state, wherein the adjusted operating state is executed by an intelligent temperature controller. The system monitors the actual load and temperature changes. If the actual load is lower than the upper limit of the safe load of the transformer area, the overload risk is determined to be eliminated, and a recovery command is generated. The recovery command is determined by switching the flexible load device back to the user-set state.

[0014] Secondly, the present invention provides a flexible load control system, applicable to the aforementioned flexible load control method, comprising: The data acquisition module is used to deploy data acquisition equipment in the distribution area to obtain the power consumption, indoor temperature and ambient temperature data of each flexible load user, and upload the data to the edge computing layer after data preprocessing. The load forecasting module is used by the edge computing layer to obtain the total load curve of the transformer area for a specific future period based on historical load data, real-time load data and weather forecast information using the ARIMA time series forecasting model. The decision control module is used by the edge computing layer to compare the total load curve with the safe load limit of the transformer area to determine whether there is an overload risk. If there is an overload risk, a collaborative control strategy is generated according to the risk level. The adjustment module is used by the user-side execution device to receive the collaborative control strategy and adjust the operating status of the flexible load equipment, while monitoring the actual load and temperature changes. If the overload risk is eliminated, the normal operation of the flexible load equipment is restored.

[0015] This invention provides a flexible load control method and system, comprising: S1: Deploying data acquisition equipment in the distribution transformer area to acquire power consumption, indoor temperature, and ambient temperature data of each flexible load user, performing data preprocessing, and uploading the data to the edge computing layer; S2: The edge computing layer, through a load forecasting module, uses an ARIMA time series forecasting model based on historical load data, real-time load data, and weather forecast information to obtain the total load curve of the transformer area for a specific future period; S3: The edge computing layer, through a decision control module, compares the total load curve with the safe load limit of the transformer area to determine whether there is an overload risk. If an overload risk exists, a collaborative control strategy is generated according to the risk level; S4: User The side-execution device receives the collaborative control strategy and adjusts the operating status of the flexible load equipment, while monitoring actual load and temperature changes. If the overload risk is eliminated, the normal operation of the flexible load equipment is restored. This achieves systematic and intelligent management of the electric heating load in the distribution area, overcoming the shortcomings of isolated and disconnected links in existing technologies. It constructs a refined control mechanism of multi-dimensional collaboration and flexible adjustment, effectively aggregating the dispersed electric heating loads into a controllable flexible resource pool. Through a refined group control strategy, it achieves peak shaving and valley filling of the load curve, smooths out load spikes, alleviates the power supply pressure on the distribution equipment, delays capacity expansion and renovation investment, and improves the adaptability and resource utilization efficiency of the distribution area to high-proportion renewable energy access. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the overall implementation process of the flexible load control method of the present invention; Figure 2 This is a flowchart of the method steps of the present invention; Figure 3 This is a comparison diagram of the predicted load of the transformer substation using the method of this invention and the traditional linear regression method; Figure 4 This is a diagram illustrating the expected effect of the adjustment strategy for flexible load electric heating according to the present invention. Detailed Implementation

[0017] Example 1 like Figure 1-4 As shown, a flexible load control method includes the following steps: S1: Deploy data acquisition equipment in the distribution area to obtain power consumption, indoor temperature and ambient temperature data of each flexible load user, and upload the data to the edge computing layer after data preprocessing; S2: The edge computing layer uses the load forecasting module to obtain the total load curve of the transformer area for a specific future period based on historical load data, real-time load data and weather forecast information, and adopts the ARIMA time series forecasting model. S3: The edge computing layer compares the total load curve with the safe load limit of the transformer area through the decision control module to determine whether there is an overload risk. If there is an overload risk, a collaborative control strategy is generated according to the risk level. S4: The user-side execution device receives the collaborative control strategy and adjusts the operating status of the flexible load equipment, while monitoring the actual load and temperature changes. If the overload risk is eliminated, the normal operation of the flexible load equipment is restored.

[0018] In this embodiment, by deploying a data acquisition layer and performing high-precision short-term load forecasting at the edge based on the ARIMA model, the overload risk of the distribution area can be identified 1-4 hours in advance, improving the proactive safety of the power grid operation. The decision control module compares the total load curve with the upper limit of the safe load of the distribution area to determine whether there is an overload risk. It constructs a refined control mechanism with multi-dimensional collaboration and flexible adjustment, effectively aggregating the dispersed electric heating load into a controllable flexible resource pool. Through a refined group control strategy, the load curve is peak-shaving and valley-filling is achieved, the load peaks are smoothed, the power supply pressure of the distribution equipment is alleviated, the investment in capacity expansion and renovation is delayed, and the adaptability and resource utilization efficiency of the distribution area to the high proportion of renewable energy access are improved. This realizes the systematic and intelligent management of the electric heating load of the distribution area, overcoming the defects of isolated and disconnected links in the existing technology.

[0019] In this embodiment, smart meters, temperature sensors, etc. are deployed in the distribution area of ​​S1 to collect data such as power consumption, indoor temperature, and ambient temperature of each electric heating user in real time, as shown in Table 1.

[0020] Table 1: Sample Data Collection Table for Transmission Station Areas (Time Granularity: 15 minutes)

[0021] In this embodiment, the essence of load forecasting is "deriving unknown loads based on known information". It is necessary to clarify the source of input data and the form of output results to ensure that the model is consistent with the actual engineering scenario.

[0022] (1) Input information: , ,

[0023] It represents the total load power of the distribution substation at time t, in units of kW. The data comes from the total smart meter data of the substation or the user-side smart meter data aggregated by the edge gateway. It includes electric heating load (accounting for 60%~80%) and daily electricity load of residents (lighting, home appliances, etc.), and is the core indicator reflecting the electricity status of the substation. This represents a time variable, in units of minutes, with a minimum time granularity of 15 minutes. The selection criteria are: ① It must comply with the State Grid smart meter standard acquisition frequency (DL / T645-2007 "Multi-functional Energy Meter Communication Protocol" stipulates that 15 minutes is the minimum data upload interval); ② The operating cycle of electric heating equipment (such as air source heat pumps) is usually 20~30 minutes. A 15-minute granularity can capture load changes caused by equipment start-up and shutdown, and avoid measurement noise introduced by excessively fine granularity. The temperature of the substation at time t is expressed in °C. This data comes from ① the hourly forecast for the next 24 hours obtained by connecting to the meteorological bureau's public API (accuracy ±1 °C); ② real-time data from digital temperature sensors deployed outdoors in the substation (accuracy ±0.5 °C). The two data are weighted and fused (in this embodiment, the sensor data weight is 0.7 and the forecast data weight is 0.3), which reduces the error of a single data source and improves the accuracy of the data.

[0024] (2) Output information: ,

[0025] This indicates the total load power of the transformer area predicted at the next (t+h) time, in units of kW. After prediction, the output is a load sequence every 15 minutes (e.g., 4 values ​​are output for a 1-hour prediction, and 16 values ​​are output for a 4-hour prediction), which facilitates the decision control module to formulate control strategies according to time segments (e.g., priority control for peak load periods). The predicted duration is set at 1-4 hours, based on the following criteria: ① Control response window requirements: From detecting overload risk to executing control commands (user notification, equipment adjustment), 1-4 hours of preparation time is required. Too short a time will prevent the process from being completed, while too long a time will lead to accumulated prediction errors due to changes in temperature and user behavior. In the preferred scheme, data preprocessing in step S1 includes: (1) Outlier handling: handling the original load data Perform threshold screening if If the load value is 1.5 times the maximum load value for the same period in the past 7 days, it is considered an outlier and is replaced by the average load value of the previous hour. The formula is as follows: (1.1); In the formula, t-15k represents k 15-minute intervals before time t: for example, k=1 corresponds to t-15 minutes, and k=4 corresponds to t-60 minutes.

[0026] (2) Missing value handling: For load data missing due to communication interruption during data acquisition, load data with a missing duration of ≤2 time granularities (i.e., a missing duration of no more than 30 minutes) will be filled using the average load value of one time granularity before and after the missing time; data with a missing duration of more than 30 minutes will be deleted, as shown in the formula: (1.2); In the formula, t-15 represents 15 minutes before time t, and t+15 represents 15 minutes after time t; The ambient temperature change within 30 minutes is usually ≤2℃, and the impact on the electric heating load is ≤16%. Based on the characteristics of air source heat pumps: for every 1℃ drop in temperature, the power increases by 8%. The error filled by the average of adjacent time points is ≤8%, which can meet the model input requirements. If the missing time is >30 minutes, the temperature and user behavior, such as adjusting the set temperature, may change significantly. In this case, the data for that time period should be discarded to avoid the filling error being transmitted to the model.

[0027] (3) Special load screening: Features weakly correlated with the load (such as humidity, wind speed, and holiday indicators) are removed, and only "strongly correlated core features" are retained to reduce the dimensionality of the model input. The formula obtained through Pearson correlation analysis is: (1.3); In the formula, , These are the feature values ​​(such as humidity, wind speed, and holiday indicators) and load values ​​of the i-th sample, respectively. , These are the sample means of the corresponding variables; N is the sample size. r The value range is [-1, 1], and the closer its absolute value is to 1, the stronger the linear correlation.

[0028] This embodiment significantly reduces noise interference and improves the accuracy and stability of the input data for the load forecasting module by automating outlier and missing value processing and irrelevant feature filtering, thereby enhancing the quality and reliability of the input data.

[0029] The model used in this embodiment is the ARIMA model. The core of the ARIMA model is to transform a non-stationary load sequence into a stationary sequence and then predict the future using historical information.

[0030] In this embodiment, the edge computing layer includes a load prediction module and a decision control module.

[0031] In S2, the load forecasting module uses the ARIMA algorithm to predict the total load curve of the transformer area for the next 1-4 hours based on historical load data, real-time load data, weather forecast data, and especially temperature information.

[0032] The ARIMA (Autoregressive Integral Moving Average) model is a classic model in the field of time series forecasting. It has the characteristics of having few parameters, low computational complexity, and strong interpretability. It has also been verified by a large number of scientific research experiments that it improves the accuracy of load forecasting in the short term (1-4 hours) of distribution substations, thus meeting the actual control needs.

[0033] In the preferred embodiment, step S2 includes: S21: Differential Processing: Differential processing transforms a non-stationary sequence into a stationary sequence. A stationary sequence is a fundamental assumption of the ARIMA model; its mean and variance do not change over time, and future patterns can be inferred from historical data. Original transformer load sequence. There is an intraday cyclical trend: for example, the load peaks at 8:00 AM and 7:00 PM, and the trough occurs in the early morning; it is a non-stationary sequence (the mean and variance change over time), and needs to be transformed into a stationary sequence through first-order differencing. The formula is: (1.4); In the formula, t-15 represents 15 minutes before time t. It is sequence data; S22: Formula for predicting the load in the next period. The prediction is based on a dual mechanism of autoregression and moving average. The autoregression utilizes the pattern of historical load changes, while the moving average corrects the prediction error.

[0034] The load change is determined based on the autoregressive coefficient and the moving average coefficient, where the autoregressive coefficient represents the weight of the impact of historical load changes on the future, and the moving average coefficient represents the correction weight for the prediction error. The formula is as follows: (1.5); In the formula: This is a constant term, also known as the intercept, which represents the baseline level of load variation. It indicates the average trend of load variation when there are no external disturbances. The value range is usually -5 to 5kW, and the value is based on the load scale of the transformer area; , These represent the autoregressive coefficients, which respectively represent the influence weights of the load changes in the previous 15 minutes and the previous 30 minutes on the future changes, with values ​​ranging from 0.3 to 0.7. q represents the moving average coefficient, which is the weighting of the prediction error of the previous 15 minutes on the current prediction. The value ranges from 0.1 to 0.3. It is used to smooth random errors, such as meter measurement noise. q>1 will lead to model overfitting. The error term follows a mean of 0 and a variance of . The normal distribution Typically ≤25kW², This represents random disturbances that the model cannot capture, such as when a user temporarily turns off electric heating equipment.

[0035] S23: Prediction result restoration (stationary → non-stationary): The essence of the restoration process is based on the cumulative change of the actual load at the previous moment. For example, to predict the load at (t+15) minutes, the actual load at time t (known) needs to be added to the predicted change over 15 minutes. This ensures that the restored load sequence is continuous and without breaks, avoiding abnormal values ​​such as negative load or sudden load jumps that do not conform to reality.

[0036] The total load curve is obtained by reconstructing the load change and the actual load at the previous moment, wherein the total load curve is a time-series load sequence, and the formula is: (1.6); In the formula, It is a "load change", which needs to be restored to the actual load power through the inverse operation of the first-order difference. To ensure that the results conform to physical meaning, the load is a non-negative real number.

[0037] S24: Model Training and Update: Model Parameters ( , , , It needs to be trained using historical data and needs to be updated regularly to adapt to changes in load, such as seasonal changes and users adding new electric heating equipment.

[0038] (1) Objective function of training set Using historical data from the past 30 days as the training set (including load (P(t)) and temperature (T(t))), the parameters of the time series prediction model are trained based on the historical load data using the least squares method, and the parameters are updated weekly to obtain the updated model. The formula is as follows: (1.7) In the formula, N is the number of training set samples (30 days × 96 time granularities = 2880 samples).

[0039] Note: The 30-day data includes a complete weekly cycle (5 working days + 2 weekends), which can cover the differences in users' electricity consumption habits on different days: for example, on working days, users are away during the day, and the electric heating load is 20%~30% lower than on weekends; the least squares method is a classic method for parameter estimation, which has unbiasedness (the mean of the parameter estimate is equal to the true value) and effectiveness (the estimation error is minimal), and has low computational complexity, with fitting time in the transformer area gateway ≤ 3 minutes.

[0040] (2) Evaluation accuracy indicators The Mean Absolute Percentage Error (MAPE) is used as the accuracy evaluation index, which intuitively reflects the proportion of prediction error to actual load. The formula is as follows: (1.8) Note: The accuracy requirement for load forecasting in transformer area control does not need to be too high. For example, if the safe load limit of a transformer area is 1200kW, and the predicted load is 1100kW, even if MAPE=15%, the actual load may be as high as 1265kW (only 5.4% above the limit), which can be mitigated by short-term rotating control. If the predicted load is 1300kW, the actual load may be as low as 1105kW (without exceeding the limit), and no overload risk will be missed.

[0041] (3) Adaptive update Establish a "weekly update" mechanism—automatically perform the following operations every weekend at 23:00: ① Add real-time data (load, temperature) for the 7 days of the current week; ② Remove the oldest 7 days of data from the training set; ③ Refit the parameters using the least squares method and update the model.

[0042] This embodiment utilizes the characteristics of ARIMA model having few parameters and low computational complexity, combined with the complete process of differencing, prediction, and restoration, and performs periodic model updates at the edge side. Under the premise of meeting the real-time requirements of engineering, it provides a reliable load curve with small errors, thereby improving the accuracy and efficiency of future short-term load forecasting.

[0043] like Figure 3 As shown, the load values ​​differ, with the horizontal axis representing time and the vertical axis representing load power (unit: kW): 1) Actual load curve (represented by a solid red line): Based on the actual operating data collected by the smart meters in the distribution area, it reflects the actual load changes of the distribution area during the corresponding time period; 2) ARIMA model prediction curve (represented by a solid blue line): The curve closely matches the actual load curve as a whole, with only slight deviations at load abrupt change nodes (such as concentrated equipment start-up and shutdown), and a low mean absolute percentage error (MAPE); 3) Linear regression prediction curve (represented by a solid pink line): The curve has poor smoothness and obvious deviations, especially during peak / valley periods, deviating significantly from the actual curve. It cannot accurately capture the periodic fluctuations and temperature-sensitive characteristics of the load, and has a large MAPE.

[0044] By comparing the curve fit, it is intuitively demonstrated that the ARIMA time series prediction model in this embodiment can more accurately capture the temporal patterns of transformer load and environmental influencing factors, and the prediction accuracy is significantly better than that of the traditional linear regression method.

[0045] In this embodiment, the core function of the decision control module is to connect the load forecast results with the actual control actions: based on the "total load curve of the distribution area for the next 1-4 hours" output by the load forecast module, it is compared in real time with the preset "safe load limit" of the distribution area to determine whether there is an overload risk; if there is a risk, it follows the two principles of fairness (i.e., all users have equal opportunities for control, avoiding excessive control for some users) and minimum impact (minimizing the interference of control actions on users' heating comfort), generates executable collaborative control instructions, and sends them to the user-side smart thermostat. This module solves the defects of the extensive power rationing and indiscriminate control in the existing technology. Through refined strategy design, it minimizes the impact on users' lives while ensuring the safe operation of the distribution network, which meets the requirements of the "Distribution Network Operation Regulations" (DL / T1573-2016) for balancing user rights protection and grid safety.

[0046] In the preferred embodiment, step S3 includes: S31: Overload Risk Assessment ① First, determine the upper limit of the safe load for the transformer area. The safe load of the transformer area is not a fixed value.

[0047] The upper limit of the safe load of the distribution area is determined based on the rated capacity of the distribution transformer, the safe load rate of the transformer, the rated current carrying capacity of the main line of the distribution area, the average supply voltage of the distribution area, and the average power factor of the distribution area. The formula is as follows: (1.9); In the formula, S is the rated capacity of the distribution transformer (kVA), which is taken from the transformer nameplate parameters; η is the safe load factor of the transformer, ranging from 0.8 to 0.9 (according to the "Operating Regulations for Power Transformers" DL / T5729-2016, the long-term operating load factor should not exceed 80%, and can be relaxed to 90% for short-term operation within 1 hour to avoid overheating and damage to the transformer); η is the safe load factor of the transformer. The rated current carrying capacity of the main line in the distribution area is determined by referring to the table in the "Low Voltage Distribution Design Code" GB50054-2011 based on the conductor material (e.g., copper core wire) and cross-section (e.g., 120mm², 150mm²). For example, the current carrying capacity of 120mm² copper core wire is 210A. The average power supply voltage (kV) for the distribution area is taken as 0.38kV (standard voltage for low-voltage distribution areas). The average power factor for the distribution area is 0.85 to 0.9 (electric heating equipment is an inductive load, and its power factor is usually 0.85. It can be increased to 0.9 after adding reactive power compensation).

[0048] ② Overload risk level classification To avoid either "emphasizing regulation over risk" or "emphasizing regulation over risk," the decision control module is based on predicted load. With the upper limit of safe load in the transformer area Based on the relationship, four risk levels are divided, with each level corresponding to a different level of regulatory intensity to ensure the precision of regulation.

[0049] S32: Compare the predicted load in the total load curve with a preset multiple of the safe load limit of the transformer area. If the predicted load exceeds the preset multiple, it is determined that there is an overload risk, and the risk level is classified according to the degree of exceedance, using the following formula: (1.10); No control is required for risk-free loads; the equipment can operate freely without adjustment. For low-risk loads, slight control is possible (e.g., fine-tuning within a temperature range). For high-risk loads, moderate / heavy control is required (e.g., a combination of rotational and priority control). Explanation: Setting 0.9 times the safe load as the "warning threshold" is because load forecasting has an error of ≤15%. Using 1.0 times directly would be insufficient. As a trigger point, the actual load may exceed the upper limit due to prediction errors; a threshold of 0.9 times can reserve a buffer space to ensure that the control action has enough time to take effect (15~30 minutes).

[0050] S33: If an overload risk is determined, a combination of control strategies is selected based on the risk level, wherein the combination of control strategies is determined from fair round-robin control, priority control, and temperature range control.

[0051] This embodiment dynamically calculates the upper limit of safe load by integrating transformer and line capacity, and finely classifies risk levels based on the ratio of predicted load to the upper limit. This enables dynamic and quantitative assessment of the operational risks of the distribution area, providing a precise basis for triggering and intensity control of the control strategy, avoiding blind control and improving the accuracy of control.

[0052] In this embodiment, the module generates differentiated control instructions for different overload risk levels, which are divided into three core control strategies.

[0053] (1) Core control strategy one: fairness-based round-robin control strategy In the preferred scheme, the fair round-robin control that selects a combination of control strategies includes: Based on a fair rotational control strategy, flexible load devices are grouped according to geographical region and device power to obtain multiple groups. A control duration and control interval are set, where the control duration represents the duration of a single power reduction or pause, and the control interval represents the time between two controls within the same group. The fair rotational control is executed cyclically according to the groups to obtain a rotational sequence, where the rotational sequence is determined by sequentially controlling each group and restoring the operation of the previous group.

[0054] The specific implementation includes: Step 1: Grouping equipment: Grouping by both "geographical area of ​​the transformer substation" and "equipment power" avoids centralized control of the same area / high-power equipment (leading to sudden drops in localized temperature). For example: ①: Divide the area into 3 to 5 geographical zones (such as East Zone and West Zone, with each zone having ≤20 users).

[0055] ②: Within each zone, the equipment is divided into "low-power group" (≤2kW, such as wall-mounted electric heaters) and "high-power group" (>2kW, such as air source heat pumps) according to power.

[0056] ③: The final number of groups = number of zones × 2 (e.g., 3 zones × 2 = 6 groups), and the difference in the number of devices in each group is ≤ 2 households, to ensure that the load is evenly distributed.

[0057] Step 2: Setting the control parameters: ①: Set the adjustment duration: 10~15 minutes / group (single adjustment duration).

[0058] ②: Set the control interval: 60~90 minutes (the interval between two controls in the same group).

[0059] ③: Control method: When the risk is low, the power is reduced by 30%; when the risk is high, the power is paused briefly (10 minutes).

[0060] Step 3: Round-robin execution logic: The process is executed in a cyclical order of "Group 1 → Group 2 → ... → Group n → Group 1". For example, the execution flow for 6 groups of devices under high risk is as follows: 0-10 minutes: Control group 1 (paused), other groups operate normally.

[0061] 10-20 minutes: Control group 2 (pause), group 1 resumes operation.

[0062] ... 50-60 minutes: Control group 6 (pause), group 5 resumes operation.

[0063] 60-70 minutes: Adjust Group 1 again (pause), completing one cycle.

[0064] Note: According to the human thermal comfort response model in "Building Environment Science", the human body's perception of changes in indoor temperature lags by about 15 to 20 minutes. During the short-term adjustment period of 10 to 15 minutes, the indoor temperature drops by ≤1.5℃ (after the electric heating equipment is stopped, the heat storage of the building envelope (walls and windows) will slow down the temperature drop), and users will not feel a significant cold sensation.

[0065] The frequency (once per hour) and duration (10 minutes) of each group of devices are completely consistent to avoid over-regulation of any group of devices (e.g., one group is regulated twice in one hour, while other groups are regulated once), which is in line with the principle of "equal rights for users".

[0066] This embodiment uses a dual-dimensional grouping based on geography and power, along with a fixed-time rotation mechanism, to evenly distribute the control pressure to all users and devices. While achieving the peak shaving target, it maximizes the fair electricity rights of the user group and improves the fairness and uniformity of load control.

[0067] (2) Core control strategy two: priority control: In the preferred scheme, priority control includes: The priority control is based on the user's temperature sensitivity and power consumption duration to divide the priority level into high-priority users and low-priority users. The high-priority users correspond to a lower control frequency, and the low-priority users correspond to a higher control frequency. The priority level is updated periodically based on power consumption data.

[0068] In the preferred scheme, the temperature range for the selected control strategy combination is determined, including: The temperature range is controlled to determine a comfortable temperature range, and the user-set temperature is adjusted to the lower limit of the comfortable temperature range to obtain the adjusted target temperature, wherein the adjusted target temperature is determined by extending the equipment downtime.

[0069] If the risk level is high, then a dual regulation is achieved by combining the adjusted target temperature with a power reduction, wherein the power reduction is determined by reducing the operating power of the equipment.

[0070] If the indoor temperature is monitored by the user-side execution device and is lower than a preset threshold, the temperature range control is stopped and the original user-set temperature is restored.

[0071] The specific implementation includes: Priority control refers to "classifying users into different priorities based on their temperature sensitivity and living needs, with higher-priority users receiving less regulation and lower-priority users receiving more regulation," ensuring that temperature-sensitive groups such as the elderly and young children are not affected. The specific steps are as follows: 1) Priority 1: Residential users with elderly people (≥65 years old) or young children (≤6 years old) in their households, based on the special population identification registered in the electricity marketing system + winter electricity consumption ≥16 hours / day (indicating high dependence on heating); the adjustment frequency for this priority is no adjustment when the risk is low, and adjustment once every 2 hours (10 minutes) when the risk is high.

[0072] 2) Priority 2: Ordinary residential users without special groups. These users have no special identification and their electricity usage time in winter is 10-16 hours / day. The priority handling method for this user is to adjust the power supply once every 1.5 hours when the risk is low and once every hour when the risk is high.

[0073] 3) Priority 3: For users in public places, the electricity usage is commercial / public + winter electricity usage time ≤ 10 hours / day, not 24-hour heating. For this priority, the control is adjusted once every 1 hour when the risk is low and once every 45 minutes when the risk is high.

[0074] The priority is updated every 7 days based on user electricity consumption data, such as electricity usage duration and power fluctuations. For example, if a regular residential user's electricity usage duration increases from 12 hours / day to 18 hours / day in winter without any special indication, the user can be temporarily promoted to priority level 1. If it is suspected that an elderly or young child has been added, the priority will be officially adjusted after the user confirms the change, so as to avoid unfairness caused by static priority.

[0075] This embodiment achieves precise control by identifying and protecting special groups in the home such as the elderly and young children who are sensitive to temperature. Under the premise of ensuring the safety of the power grid, it prioritizes meeting the comfort needs of users who need continuous heating the most, thereby improving the acceptance of the technology and the user experience.

[0076] (3) Core control strategy three: fine-tuning based on the comfort temperature range In the preferred embodiment, step S3 generates a collaborative control strategy based on the risk level, including: If the risk level is low, a strategy based on temperature range control is generated, wherein the temperature range control is obtained by adjusting the target temperature to the lower limit of the comfort range, and the original target temperature is maintained for high-priority users in the priority control.

[0077] If the risk level is high risk, a combined strategy of polling control, priority control and temperature range control is generated. The polling control reduces power by performing grouped cyclic execution, the priority control adjusts the control frequency according to the user priority, and the temperature range control adjusts the target temperature before polling.

[0078] If the risk level is extreme risk, a strategy based on priority control and round-robin control is generated. The priority control increases the intensity of regulation for low-priority users, and the round-robin control pauses execution by shortening the interval and is supplemented by temperature range control to adjust to a lower limit.

[0079] The specific implementation includes: Temperature range control refers to: instead of directly shutting down the electric heating equipment, extending the precise temperature set by the user to a comfortable temperature range for the human body. In this embodiment, 20~23℃ is used. When there is an overload risk, the target temperature of the equipment is temporarily adjusted to the lower limit of the range, 20℃. By extending the equipment downtime, load shaving is achieved. The specific steps are as follows: Step 1: Referring to the "Evaluation Standard for Indoor Thermal and Humidity Environment of Civil Buildings" (GB / T50785-2012), the comfortable indoor temperature range for civil buildings in winter is 18~24℃. Combining the temperature control accuracy of electric heating equipment (±1℃), the final comfort range for regulation is determined to be 20~23℃. The lower limit of 20℃ can avoid discomfort caused by excessively low temperatures, and the upper limit of 23℃ can avoid wasting electricity due to excessively high temperatures.

[0080] Step 2: Setting Adjustment Parameters ① Normal operation: The target temperature of the equipment is equal to the user-set temperature. For example, if the temperature is 22℃, it should be within the range of 20~23℃. If the user sets the temperature to 24℃, it will automatically adjust to 23℃.

[0081] ②Low-risk control: The target temperature is lowered to the lower limit of the range, 20℃, and the equipment is operated at 20℃. The downtime is extended: for example, the downtime is 5 minutes / cycle when the temperature is 22℃, and 8 minutes / cycle when the temperature is 20℃.

[0082] ③ High-risk regulation: Target temperature reduced to 20℃ + power reduced by 20%: dual peak shaving, downtime extended to 12 minutes / cycle.

[0083] ④ Risk resolved: The target temperature will automatically return to the user's original set temperature of 22℃.

[0084] Step 3: Temperature Feedback Closed Loop The intelligent thermostat collects the indoor temperature in real time. If the indoor temperature is lower than 19°C during the adjustment period, the comfort range will be lowered by 1°C, the adjustment will be stopped immediately, and the original target temperature will be restored to avoid excessive temperature drop.

[0085] This embodiment extends equipment downtime by fine-tuning the set temperature within a small range, achieving load reduction with almost imperceptible results for the user. Combined with a closed-loop temperature feedback system, it prevents excessive drops in indoor temperature, balances energy efficiency and comfort, and achieves flexible load regulation, minimizing the impact on users.

[0086] This embodiment adopts corresponding control strategies according to different risk scenarios.

[0087] In a preferred embodiment, step S4 includes: obtaining the collaborative control strategy from the edge computing layer via a communication module, wherein the collaborative control strategy includes a temperature adjustment command or a start / stop command.

[0088] The set temperature or operating power of the flexible load equipment is adjusted according to the collaborative control strategy to obtain the adjusted operating state, wherein the adjusted operating state is executed by an intelligent temperature controller.

[0089] The system monitors the actual load and temperature changes. If the actual load is lower than the upper limit of the safe load of the transformer area, the overload risk is determined to be eliminated, and a recovery command is generated. The recovery command is determined by switching the flexible load device back to the user-set state.

[0090] The specific implementation includes: ① For low-risk scenarios, a strategy combination of "primarily temperature range control and secondarily priority control" is adopted: First, temperature range control is activated for all Level 2 and Level 3 users, lowering the target temperature from the user-set value of 22℃ to the lower limit of the comfort range of 20℃. This achieves a power reduction of approximately 20% per device by extending the device's downtime. Meanwhile, for Level 1 users: those with elderly people, young children, or other special groups in their households, the original set temperature of 22℃ is maintained, and they do not participate in the power reduction.

[0091] This combined strategy achieves a total peak reduction rate of approximately 16% while maximizing user comfort and avoiding fluctuations in equipment start-up and shutdown caused by frequent polling.

[0092] ② For high-risk scenarios, a triple collaborative strategy of polling control + priority control + temperature range control is adopted: The first step is to group users based on priority control: Level 1 users are divided into Group A, Level 2 users are evenly divided into Groups B1 to B4, and Level 3 users are divided into Groups C1 to C2.

[0093] The second step is to activate the temperature range control 5 minutes before each group's rotation begins, lowering the target temperature for that group of users from 22℃ to 20℃, so that the indoor temperature drops gradually.

[0094] The third step is to implement rotational control: Group A pauses for 10 minutes every 2 hours, Group B pauses for 10 minutes every 1 hour, and Group C pauses for 12 minutes every 45 minutes.

[0095] This combined strategy achieves a peak reduction rate of 35% to 40% through sequential timing, while striking an optimal balance between fairness, protection of special groups, and user comfort.

[0096] ③ For extreme scenarios, an enhanced strategy is adopted, which primarily uses priority control and polling control, supplemented by temperature range control: First, temperature range control is implemented only for Level 1 users: the temperature is lowered to 20℃, exempting them from the obligation to pause patrols, reflecting public concern. Second, enhanced patrols are implemented for Level 2 and Level 3 users: Level 2 users patrol every 45 minutes with a 10-minute pause; Level 3 users patrol every 30 minutes with a 12-minute pause. Simultaneously, the target temperature for all users participating in patrols is lowered to the lower limit of the comfort range of 19℃, and the pause duration is appropriately extended to maximize peak shaving effects within the acceptable range for users.

[0097] This embodiment dynamically invokes and combines different control strategies for different risk levels, such as low, high, and extreme, achieving a smooth transition from gentle guidance to enhanced intervention. It ensures that the dual goals of safety and comfort can be achieved in the best way under any circumstances, improving the intelligence and refined control capabilities for dealing with different risk scenarios.

[0098] like Figure 4 As shown, the system achieves user-initiated adjustments and collaborative dynamic adjustments with the proposed strategy to jointly meet user load demands. It demonstrates the effect of the collaborative control strategy (a combination of temperature range control, fair rotation control, and priority control) on the load changes in the transformer substation before and after implementation, validating the technical goals of "peak shaving and valley filling" and "balancing safety and comfort." The horizontal axis represents time, and the vertical axis represents the total load power of the transformer substation (unit: kW).

[0099] The load curve (blue rectangle) after the strategy adjustment in this embodiment: The load change curve after implementing the coordinated control strategy is that the original sharp peak is smoothed out, the overall load is controlled below the upper limit of the safe load, and the curve is smooth without drastic fluctuations: Low-risk period (e.g., 7:00-7:30 am): Only by fine-tuning the temperature range control, the load drops slightly to below 1080 kW without significant fluctuations; High-risk period (e.g., 8:00-9:00 am): Through the combination of rotation + temperature range + priority control, the load is stabilized between 1100-1200 kW, achieving peak shaving (the original peak value of 1400 kW drops to below 1200 kW); After the risk is eliminated (e.g., after 9:30 am): The load naturally falls back, and the curve returns to a stable state within the safe range.

[0100] The collaborative control strategy of this invention can effectively smooth out instantaneous peaks in the electric heating load of the distribution area, stabilizing the total load within a safe range, while avoiding the impact of large load fluctuations on user comfort. This achieves the dual objectives of "safe grid operation" and "user heating comfort," verifying the effectiveness of the refined control mechanism. In this embodiment, a smart thermostat is installed on the user side and connected to the electric heating equipment. It receives and executes control commands issued by the edge computing layer, while simultaneously uploading local data. The smart thermostat includes a communication module, a main control module, a sensing module, a temperature control execution module, a human-machine interaction module, and a power supply module.

[0101] The generated control commands are sent to the target user's smart temperature controller (execution layer) via communication methods such as power line carrier communication (HPLC) or radio frequency (RF).

[0102] The smart thermostat receives and executes commands to adjust the operating status of the connected electric heating equipment, for example, temporarily adjusting the set temperature to the lower limit of the comfort range.

[0103] The system enters a monitoring loop, continuously monitoring the actual value of the total load in the distribution area and changes in the indoor temperature of users.

[0104] Determine if the overload risk has been eliminated. If not, dynamically adjust the control strategy (e.g., adjust the patrol group or control intensity) and return to the previous step; if eliminated, generate a recovery command, end the current control, restore the equipment to normal operation, and return to the initial data acquisition stage.

[0105] Example 2 Further illustrating with reference to Embodiment 1, a flexible load control system, applicable to the flexible load control method described in Embodiment 1, includes: The data acquisition module is used to deploy data acquisition equipment in the distribution area to obtain the power consumption, indoor temperature and ambient temperature data of each flexible load user, and upload the data to the edge computing layer after data preprocessing. The load forecasting module is used by the edge computing layer to obtain the total load curve of the transformer area for a specific future period based on historical load data, real-time load data and weather forecast information using the ARIMA time series forecasting model.

[0106] The decision control module is used by the edge computing layer to compare the total load curve with the upper limit of the safe load of the transformer area to determine whether there is an overload risk. If there is an overload risk, a collaborative control strategy is generated according to the risk level.

[0107] The adjustment module is used by the user-side execution device to receive the collaborative control strategy and adjust the operating status of the flexible load equipment, while monitoring the actual load and temperature changes. If the overload risk is eliminated, the normal operation of the flexible load equipment is restored.

[0108] This embodiment provides the working process, working details and technical effects of a flexible load control method, which can be found in Embodiment 1 and will not be repeated here.

[0109] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method of flexible load regulation, comprising: The method comprises the following steps: S1: deploying a data acquisition device in a power distribution area to obtain power consumption, indoor temperature and ambient temperature data of each flexible load user, and uploading the data to an edge computing layer after data preprocessing; S2: the edge computing layer obtains a total load curve of the area in a specific period in the future based on historical load data, real-time load data and weather forecast information by a load prediction module using an ARIMA time series prediction model; S3: the edge computing layer compares the total load curve with the upper limit of the safe load of the area by a decision control module to determine whether there is an overload risk, and generates a collaborative control strategy according to the risk level if there is an overload risk; S4: the user side execution device receives the collaborative control strategy and adjusts the operation state of the flexible load device, while monitoring the actual load and temperature changes, and if the overload risk is eliminated, the normal operation of the flexible load device is restored.

2. The method of claim 1, wherein, The data preprocessing in S1 comprises: Outlier processing: threshold screening is performed on the original load data, and if the load value exceeds 1.5 times the maximum load value of the same period in the past 7 days, it is determined as an outlier, and the average load value of the previous 1 hour is used to replace it, the formula is: (1.1)； In the formula, t-15k represents k 15 minutes before time t; Missing value processing: for load data with a missing duration of not more than 30 minutes, the average load value of each time granularity before and after the missing time is used to fill in; for data with a missing duration of more than 30 minutes, it is deleted, the formula is: (1.2)； In the formula, t-15 represents 15 minutes before time t, and t+15 represents 15 minutes after time t; Special load screening: through Pearson correlation analysis, weakly correlated features are removed, and only strongly correlated core features are retained, the formula is: (1.3)； In the formula, , are the feature values (such as humidity, wind speed, holiday identifier) and load values of the i-th sample, respectively; , are the sample mean values of the corresponding variables; N is the sample size.

3. The method of claim 1, wherein, S2 comprises: S21: input the data processed in the data preprocessing into the ARIMA time series prediction model, and obtain a stationary sequence through difference processing, the formula is: (1.4)； In the formula, t-15 represents 15 minutes before time t, is sequence data; S22: determine the load change amount according to the autoregressive coefficient and the moving average coefficient, wherein the autoregressive coefficient represents the influence weight of historical load change on the future, and the moving average coefficient represents the correction weight of prediction error, the formula is: (1.5)； where t is the current time, is a constant term, , respectively represent autoregressive coefficients, represent moving average coefficients, is an error term, and h is the prediction length. S23: restore the total load curve according to the load change amount and the actual load at the previous time, wherein the total load curve is a load sequence of each time granularity, the formula is: (1.6)； In the formula, is the load change amount; S24: train the parameters of the time series prediction model based on the historical load data by the least square method, and update the parameters every week to obtain an updated model, the formula is: (1.7)； In the formula, N is the number of training set samples.

4. The method of claim 1, wherein, S3 comprises: S31: determine the upper limit of the safe load of the area according to the rated capacity of the power distribution transformer, the transformer safety load rate, the rated carrying capacity of the main line of the area, the average power supply voltage of the area and the average power factor of the area, the formula is: (1.9)； In the formula, S is the rated capacity of distribution transformer, η is the safe load ratio of transformer, I is the rated current-carrying capacity of main line of transformer station, U is the average power supply voltage of transformer station, is the average power factor of transformer station; S32: compare the predicted load in the total load curve with a preset multiple of the upper limit of the safe load of the area, if the predicted load exceeds the preset multiple, it is determined that there is an overload risk, and the risk level is divided according to the overrun degree, the formula is: (1.10)； S33: If it is determined that there is an overload risk, a control strategy combination is selected according to the risk level, wherein the control strategy combination is determined from fair round-robin control, priority control and temperature interval control.

5. The method of claim 4, wherein, The fair round-robin control for selecting the control strategy combination comprises: For the fair round-robin control, flexible load devices are grouped according to geographical areas and device power to obtain a plurality of groups, and a regulation duration and a regulation interval are set, wherein the regulation duration represents the duration of a single power reduction or suspension, and the regulation interval represents the time between two regulations of the same group; The fair round-robin control is executed according to the group cycle to obtain a round-robin sequence, wherein the round-robin sequence is determined by sequentially regulating each group and restoring the operation of the previous group.

6. The method of claim 4, wherein, The priority control for selecting the control strategy combination comprises: For the priority control, priority levels are divided according to user temperature sensitivity and power consumption duration to obtain high-priority users and low-priority users, wherein the high-priority users correspond to a lower regulation frequency, the low-priority users correspond to a higher regulation frequency, and the priority levels are updated periodically according to power consumption data.

7. The method of claim 4, wherein the flexible load regulation is characterized by, In the temperature interval control for selecting the control strategy combination, it is determined that: For the temperature interval control, a comfortable temperature interval is determined, and the user set temperature is adjusted to the lower limit of the comfortable temperature interval to obtain an adjusted target temperature, wherein the adjusted target temperature is determined by extending the device downtime; If the risk level is high, a double regulation is obtained by combining power reduction on the basis of the adjusted target temperature, wherein the power reduction is determined by reducing the device operating power; The indoor temperature is monitored by the user side execution device, and if the indoor temperature is lower than a preset threshold, the temperature interval control is stopped and the original user set temperature is restored.

8. The method of claim 4, wherein the flexible load regulation is characterized by, In S3, a cooperative control strategy is generated according to the risk level, comprising: If the risk level is low, a strategy mainly based on temperature interval control is generated, wherein the temperature interval control is obtained by adjusting the target temperature to the lower limit of the comfortable interval, and the high-priority users in the priority control maintain the original target temperature; If the risk level is high, a combined strategy of round-robin control, priority control and temperature interval control is generated, wherein the round-robin control is executed by grouping and cycling power reduction, the priority control adjusts the regulation frequency according to user priority, and the temperature interval control adjusts the target temperature before round-robin; If the risk level is extreme, a strategy mainly based on priority control and round-robin control is generated, wherein the priority control increases the regulation intensity for low-priority users, and the round-robin control executes suspension by shortening the interval, and is supplemented by temperature interval control to a lower lower limit.

9. The method of claim 1, wherein, In S4, it comprises: The cooperative control strategy is obtained from the edge computing layer through the communication module, wherein the cooperative control strategy comprises temperature adjustment instructions or start-stop instructions; The set temperature or operating power of the flexible load device is adjusted according to the cooperative control strategy to obtain an adjusted operating state, wherein the adjusted operating state is executed by an intelligent temperature controller; The actual load and the temperature change are monitored, if the actual load is lower than the upper limit of the safety load of the distribution area, it is judged that the overload risk is removed, and a recovery instruction is generated, wherein the recovery instruction is determined by switching the flexible load device back to the user setting state.

10. A flexible load regulation system adapted for use in a method of flexible load regulation according to any one of claims 1-9, characterized in that, Comprise: A data acquisition module is used to deploy data acquisition equipment in a power distribution area, to obtain the power consumption, indoor temperature and ambient temperature data of each flexible load user, and to upload the data to the edge computing layer after data preprocessing; A load prediction module is used to obtain the total load curve of the distribution area in a specific future period by the edge computing layer through the load prediction module based on historical load data, real-time load data and weather forecast information, using an ARIMA time series prediction model; A decision control module is used to compare the total load curve with the upper limit of the safety load of the distribution area by the edge computing layer through the decision control module to determine whether there is an overload risk, and to generate a collaborative control strategy according to the risk level if there is an overload risk; An execution adjustment module is used for user-side execution equipment to receive the collaborative control strategy and adjust the operating state of the flexible load device, while monitoring the actual load and temperature change, and if the overload risk is removed, the normal operation of the flexible load device is restored.

Images