A virtual power plant load AI prediction method with dynamic response on the demand side

Abstract

The present invention relates to the technical field of electric load prediction, and specifically to an AI prediction method for virtual power plant load with dynamic response on the demand side. The method obtains the electric load data of the equipment in the virtual power plant and the charge and discharge bias value of the energy storage system in real time; obtains the predicted mutation degree according to the change of the predicted electric load data of the equipment relative to the electric load data at the current moment; corrects the predicted mutation degree according to the change of the electric load data and the periodic distribution of the predicted electric load data to obtain the overall predicted mutation degree of the energy storage system; obtains the corrected predicted charge and discharge bias value of the energy storage system according to the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree to predict the charge and discharge of the virtual power plant. The present invention effectively improves the accuracy of the charge and discharge prediction of the virtual power plant by accurately obtaining the corrected predicted charge and discharge bias value of the energy storage system, which is conducive to maintaining the stability of the overall operation of the power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of electric load forecasting, and in particular to an AI-based load forecasting method for a virtual power plant with dynamic demand-side response. Background Art

[0002] In the context of energy transition, virtual power plants (VPPs) enable flexible grid control by aggregating distributed power sources (such as photovoltaic and wind power), energy storage systems, and adjustable loads. Their core function is to accurately predict both the generation load (especially fluctuating renewable energy sources) and the demand load within the controllable range. This is fundamental to maintaining real-time grid balance and ensuring safe and stable operation. Accurate predictions enable VPPs to proactively optimize the dispatch of distributed resources, effectively smoothing fluctuations and shifting peak loads. This significantly reduces the need for backup capacity and overall operating costs, thereby improving grid utilization efficiency and economic benefits.

[0003] Existing methods use unscented Kalman filtering to predict the overall load of the power grid. However, in reality, different devices in a virtual power plant will have different electrical loads at different times due to external environmental influences. For example, cloud cover may cause a sudden drop in photovoltaic power. Therefore, the result of directly predicting the overall load of the power grid deviates greatly from the actual load, which in turn causes energy storage to malfunction, resulting in problems such as local voltage exceeding the limit and redundant backup capacity. Summary of the Invention

[0004] In order to solve the technical problem of inaccurate prediction of the overall power grid load using unscented Kalman filtering, the purpose of the present invention is to provide an AI-based load prediction method for virtual power plants with dynamic demand-side response. The technical solutions adopted are as follows:

[0005] An embodiment of the present invention provides an AI-based load prediction method for a virtual power plant with dynamic demand-side response, the method comprising the following steps:

[0006] Obtain the power load data of each device in the virtual power plant at each moment and the charge and discharge bias value of the energy storage system at each moment during the current time period;

[0007] Based on the power load data of each device in the current time period, the predicted power load data of each device at the next moment after the current moment is predicted; based on the difference between the predicted power load data of each device and its power load data at the current moment, as well as the trend change of the predicted power load data relative to its power load data at the current moment, the predicted mutation degree of each device is obtained;

[0008] Based on the change in the electric load data of each device in the current time period and the corresponding periodic distribution of the predicted electric load data of each device in the electric load data in the current time period, the predicted mutation degree is corrected to obtain the overall predicted mutation degree of the energy storage system;

[0009] Based on the charge and discharge bias value in the current time period, the predicted charge and discharge bias value of the energy storage system at the next moment after the current moment is predicted; according to the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree, the corrected predicted charge and discharge bias value of the energy storage system is obtained to predict the charge and discharge of the virtual power plant.

[0010] Furthermore, the method for obtaining the predicted mutation degree is:

[0011] For any device, the difference between the predicted power load data of the device and its current power load data is used as the predicted change value of the device;

[0012] The electrical load data of the device at the current moment, the moment before the current moment, and the moment after the current moment are all represented by coordinate points in a two-dimensional coordinate system; wherein, in the two-dimensional coordinate system, the time is the horizontal axis and the electrical load data is the vertical axis;

[0013] Obtain the slope of the line segment connecting the coordinate point at the previous moment and the current moment as the first slope;

[0014] Obtain the slope of the line segment connecting the coordinate points at the current moment and the next moment after the current moment as the second slope;

[0015] The difference between the first slope and the second slope is used as the predicted trend change degree of the device;

[0016] The result of normalizing the product of the predicted trend change degree and the predicted change value is taken as the predicted mutation degree of the device.

[0017] Furthermore, the method for obtaining the overall predicted mutation degree is:

[0018] Obtain the degree of fluctuation of the power load of each device based on the changes in the power load data of each device in the current time period;

[0019] Obtain the degree of regularity of the predicted period of each device based on the period distribution of the predicted electric load data of each device in the electric load data of the current time period;

[0020] According to the power load fluctuation degree and the prediction period regularity degree of each device, the predicted mutation degree of each device is corrected to obtain the corrected predicted mutation degree of each device;

[0021] The corrected predicted mutation levels of all devices are added together and normalized to obtain the overall predicted mutation level of the energy storage system.

[0022] Furthermore, the method for obtaining the degree of electric load fluctuation is:

[0023] For any device, the electrical load data of the device in the current time period is fitted into an electrical load curve according to the time sequence, and the difference in electrical load data corresponding to any two adjacent extreme value points on the electrical load curve is obtained as the first difference;

[0024] obtaining a difference between the maximum electric load data and the minimum electric load data on the electric load curve as a second difference;

[0025] The product of the number of extreme value points on the electric load curve, the mean of the first difference and the second difference is taken as the electric load fluctuation degree of the equipment.

[0026] Furthermore, the method for obtaining the degree of regularity of the prediction cycle is:

[0027] For any device, the time corresponding to the electric load data on the electric load curve of the device that is equal to the predicted electric load data of the device is obtained, and is used as the reference time;

[0028] Get the duration between any two adjacent reference moments as the first duration;

[0029] Obtain the electric load data corresponding to the center time between any two adjacent reference times, and use them as reference electric load data;

[0030] The sum of any two adjacent reference electric load data is used as a specific reference value;

[0031] The standard deviation of the first time period is added to the standard deviation of the specific reference value, and the result of the negative correlation is used as the predicted period regularity of the device.

[0032] Furthermore, the method for obtaining the modified predicted mutation degree is:

[0033] The negative correlation result of the power load fluctuation degree of each device and the degree of prediction period regularity are added and normalized, and the result is used as the prediction confidence level of each device;

[0034] The product of the prediction confidence level of each device and its predicted mutation level is taken as the corrected predicted mutation level of each device.

[0035] Furthermore, the method for obtaining the corrected predicted charge and discharge bias value is:

[0036] The difference between the predicted charge and discharge bias value and the current charge and discharge bias value is used as the bias prediction adjustment value;

[0037] The product of the overall predicted mutation degree and the biased prediction adjustment value is used as the corrected biased prediction adjustment value of the energy storage system;

[0038] The sum of the current charge and discharge bias value and the corrected bias prediction adjustment value is used as the corrected predicted charge and discharge bias value of the energy storage system.

[0039] Furthermore, the method for predicting the charging and discharging of the virtual power plant is:

[0040] When the modified predicted charge-discharge bias value is greater than a preset threshold, the virtual power plant is predicted to perform a discharge operation and the magnitude of the discharge load is the modified predicted charge-discharge bias value;

[0041] When the corrected predicted charge-discharge bias value is less than a preset threshold, the virtual power plant is predicted to perform charging operations and the magnitude of the charging load is the absolute value of the corrected predicted charge-discharge bias value;

[0042] When the corrected predicted charge and discharge bias value is equal to the preset specified threshold, the charge and discharge operation of the predicted virtual power plant is not changed.

[0043] Furthermore, the method for obtaining the predicted electric load data is:

[0044] Based on the power load data of each device in the current time period, the predicted power load data of each device at the next moment after the current moment is predicted through unscented Kalman filtering;

[0045] The method for obtaining the predicted charge and discharge bias value is:

[0046] Based on the charge and discharge bias value in the current time period, the predicted charge and discharge bias value of the energy storage system at the next moment after the current moment is predicted through unscented Kalman filtering.

[0047] Furthermore, the method for obtaining the charge and discharge bias value is:

[0048] The difference between the power load data and the power generation load data of the energy storage system at each moment is used as the charge and discharge bias value of the energy storage system at each moment.

[0049] The present invention has the following beneficial effects:

[0050] The present invention first predicts the predicted electric load data of each device at the next moment after the current moment based on the electric load data of each device in the current time period, so as to prepare for the subsequent analysis of the mutation degree of the predicted electric load data of each device, which is conducive to the subsequent accurate analysis of the overall charging and discharging mutation predicted by the energy storage system; and then obtains the predicted mutation degree of each device according to the difference between the predicted electric load data of each device and its electric load data at the current moment, as well as the trend change of the predicted electric load data relative to its electric load data at the current moment, and accurately reflects the mutation of the predicted electric load data of each device; taking into account the deviation between the predicted electric load data directly predicted in actual conditions and the actual conditions, and then according to the change of the electric load data of each device in the current time period, as well as the corresponding trend of the predicted electric load data of each device in the electric load data of its current time period, the predicted mutation degree of each device is obtained, and the mutation of the predicted electric load data of each device is accurately reflected. The cycle distribution of each device is analyzed, and the reliability of the predicted mutation degree is analyzed to correct the predicted mutation degree, thereby more accurately obtaining the overall predicted mutation degree of the energy storage system, accurately reflecting the degree to which the predicted full charge situation of the energy storage system should change; further based on the charge and discharge bias value in the current time period, the predicted charge and discharge bias value of the energy storage system at the next moment of the current moment is predicted, preparing for the subsequent accurate acquisition of the predicted charge and discharge bias value that is more in line with the actual situation; and then according to the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree, the corrected predicted charge and discharge bias value of the energy storage system is obtained, accurately reflecting the charge and discharge situation of the energy storage system at the next moment of the current moment, and then accurately predicting the charge and discharge of the virtual power plant, effectively improving the accuracy of the charge and discharge prediction of the virtual power plant, which is conducive to maintaining the stability of the overall operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A schematic flow chart of an AI-based load prediction method for a virtual power plant with dynamic demand-side response, provided by one embodiment of the present invention;

[0053] Figure 2 A flow chart of a method for obtaining an overall predicted mutation degree provided by one embodiment of the present invention;

[0054] Figure 3 This is a structural diagram of a virtual power plant load AI prediction system with dynamic response on the demand side provided by one embodiment of the present invention;

[0055] Figure 4 A schematic diagram of a computer device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation method, structure, features and effects of a virtual power plant load AI prediction method with dynamic response on the demand side proposed by the present invention. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics of one or more embodiments may be combined in any suitable form.

[0057] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0058] The following describes in detail a specific solution of a virtual power plant load AI prediction method with dynamic response on the demand side provided by the present invention with reference to the accompanying drawings.

[0059] Example 1:

[0060] This paper proposes a virtual power plant load AI prediction method with dynamic response on the demand side. Figure 1 , which shows a schematic flow chart of a method for predicting virtual power plant load with AI based on dynamic demand-side response provided by one embodiment of the present invention, the method comprising the following steps:

[0061] Step S1: Obtain the power load data of each device in the virtual power plant at each moment and the charge and discharge bias value of the energy storage system at each moment in the current time period.

[0062] It is known that when the charge and discharge power of the energy storage system changes suddenly, it will cause the voltage of the grid node to fluctuate violently, especially during peak load periods. For example, a sudden discharge of the energy storage system will cause a sudden increase in local voltage, while a sudden charge will cause a sudden drop in local voltage, exceeding the safe voltage range. Therefore, in order to ensure the stability of the power grid, it is necessary to predict the charge and discharge of the energy storage system. In order to predict the charge and discharge of the virtual power plant, this embodiment obtains the power load data of each device in the virtual power plant at each moment in the current time period, which is conducive to the subsequent accurate analysis of the predicted power load data of each device, and then accurately analyzes the overall power load of the entire power grid.

[0063] In this embodiment, the energy storage system is assumed to be the overall electric load control system of the power grid. The energy storage system controls the charging and discharging of the entire power grid. In order to predict the charging and discharging status of the energy storage system at the next moment after the current moment, this embodiment obtains the power load data and power generation load data of the energy storage system at each moment in the current time period. It is known that in a virtual power plant, when the overall power generation load is less than the power load, the actual power demand can be met by discharging the energy storage system. In order to accurately and efficiently predict the charging and discharging status of the energy storage system, this embodiment uses the difference between the power load data and the power generation load data of the energy storage system at each moment in the current time period as the charging and discharging bias value of the energy storage system at each moment, that is, obtains the charging and discharging bias value of the energy storage system at each moment in the current time period, and subsequently determines the charging and discharging status of the energy storage system by predicting the charging and discharging bias value.

[0064] This embodiment sets the duration of the current time period to 15 days, and the interval between two adjacent moments to 5 minutes. Implementers can adjust the duration of the current time period and the interval between two adjacent moments based on actual circumstances, and these are not specified here. However, the current time period always ends at the current moment. It should be noted that the power load data for each device in the virtual power plant at each moment, as well as the power consumption and power generation load data of the energy storage system at each moment, can be directly obtained through the installed smart meters.

[0065] Step S2: Based on the power load data of each device in the current time period, predict the predicted power load data of each device at the next moment after the current moment; according to the difference between the predicted power load data of each device and its power load data at the current moment, as well as the trend change of the predicted power load data relative to its power load data at the current moment, obtain the predicted mutation degree of each device.

[0066] Specifically, in order to accurately analyze the predicted degree of change of the energy storage system, this embodiment first predicts the predicted electric load data of each device at the next moment of the current moment based on the electric load data of each device in the current time period through unscented Kalman filtering; wherein, unscented Kalman filtering is a well-known technology and will not be described in detail. Then, based on the difference between the predicted electric load data of each device and its electric load data at the current moment, as well as the trend change of the predicted electric load data relative to its electric load data at the current moment, the predicted mutation degree of each device is obtained. The greater the predicted mutation degree, the greater the change in the predicted electric load data of the corresponding device relative to its electric load data at the current moment. When the predicted mutation degree of all devices is greater, it means that the charge and discharge bias value predicted at the next moment of the current moment of the energy storage system must change more than the charge and discharge bias value at the current moment.

[0067] Preferably, in a manner that can be implemented in this embodiment, the method for obtaining the predicted mutation degree is: for any device, the absolute value of the difference between the predicted electric load data of the device and its electric load data at the current moment is used as the predicted change value of the device; the larger the predicted change value, the greater the mutation degree of the predicted electric load data of the device; in order to further analyze the mutation of the predicted electric load data of the device, the electric load data of the device at the current moment, the moment before the current moment and the moment after the current moment are all represented by coordinate points in a two-dimensional coordinate system; wherein, in the two-dimensional coordinate system, the time is the horizontal axis and the electric load is the horizontal axis. The load data is taken as the vertical axis; the slope of the line segment connecting the coordinate points corresponding to the previous moment and the current moment is then obtained as the first slope; the slope of the line segment connecting the coordinate points corresponding to the current moment and the next moment of the current moment is obtained as the second slope; wherein, the method for obtaining the slope is a well-known technique and will not be described in detail; when the first slope and the second slope are more equal, the degree of mutation in the predicted electric load data of the device is smaller, and thus, in this embodiment, the absolute value of the difference between the first slope and the second slope is used as the degree of change in the predicted trend of the device; the greater the degree of change in the predicted trend, the greater the degree of mutation in the predicted electric load data of the device;

[0068] In order to accurately characterize the sudden change in the predicted electrical load data of the device, the product of the predicted trend change degree and the predicted change value is normalized and used as the predicted sudden change degree of the device. In this embodiment, the product of the predicted trend change degree and the predicted change value is normalized using the norm normalization function.

[0069] At this point, the predicted mutation level of each device is obtained.

[0070] Step S3: Based on the change of the electric load data of each device in the current time period and the corresponding periodic distribution of the predicted electric load data of each device in the electric load data in the current time period, the predicted mutation degree is corrected to obtain the overall predicted mutation degree of the energy storage system.

[0071] Specifically, when a certain device is a distributed power source such as photovoltaic or wind power, its electric load is closely related to the environment, and the electric load data of the device may change frequently. Therefore, the predicted electric load data of the device may be significantly different from the actual data, and the confidence level is low, which is not conducive to the subsequent accurate analysis of the charge and discharge prediction of the energy storage system based on the predicted mutation level of each device. It is known that when the electric load data of a certain device changes more stably in the current time period, and the predicted electric load data of the device conforms more to the periodic change law in the electric load data in the current time period, the predicted electric load data of the device is more accurate, which indirectly reflects that the predicted mutation level of the device is more accurate. Therefore, this embodiment corrects the predicted mutation level according to the changes in the electric load data of each device in the current time period, and the corresponding periodic distribution of the predicted electric load data of each device in its electric load data in the current time period, so as to accurately obtain the predicted mutation level that each device should actually correspond to.

[0072] Considering that the energy storage system's power consumption and generation are based on the charge and discharge conditions of each device in the virtual power plant, this embodiment obtains the overall predicted mutation level of the energy storage system based on the actual predicted mutation level corresponding to each device. Therefore, this embodiment modifies the predicted mutation level based on the changes in the electrical load data of each device within the current time period, as well as the corresponding periodic distribution of each device's predicted electrical load data within its electrical load data within the current time period, to obtain the overall predicted mutation level of the energy storage system. A greater overall predicted mutation level indicates a greater predicted charge and discharge bias value for the energy storage system.

[0073] Preferably, in one possible implementation of this embodiment, the method for obtaining the overall predicted mutation degree is as follows: Figure 2 , which shows a flow chart of a method for obtaining an overall predicted mutation degree provided by this embodiment, the method comprising the following steps:

[0074] Step S201: Obtain the degree of fluctuation of the electric load of each device according to the change of the electric load data of each device in the current time period.

[0075] The more stable the changes in a device's power load data during the current time period, the more accurate the predicted power load data for that device is likely to be, indirectly indicating a higher accuracy in the predicted mutation level for that device. Furthermore, this embodiment determines the power load fluctuation level for each device based on the changes in the power load data for each device during the current time period. The smaller the power load fluctuation, the more accurate the predicted mutation level for that device.

[0076] In one implementation of this embodiment, the method for obtaining the degree of electric load fluctuation is as follows: for any device, the electric load data of the device in the current time period is fitted into an electric load curve according to the time sequence from the front to the back, wherein the curve fitting method is a well-known technology and will not be described in detail. The absolute value of the difference between the electric load data corresponding to any two adjacent extreme points on the electric load curve is obtained as the first difference; when the first difference is smaller, the electric load curve is flatter, which indirectly indicates that the electric load data of the device in the current time period changes more stably; in order to further analyze the change of the electric load data of the device, the difference between the maximum electric load data and the minimum electric load data on the electric load curve is obtained as the second difference; the smaller the second difference, the more stable the change of the electric load data of the device in the current time period; on the other hand, when the number of extreme points on the electric load curve is smaller, it also indicates that the change of the electric load data of the device in the current time period is more stable. Furthermore, this embodiment uses the product of the number of extreme points on the electric load curve, the average of the first difference, and the second difference as the degree of electric load fluctuation of the device.

[0077] At this point, the degree of fluctuation of the power load of each device is obtained.

[0078] Step S202: Obtain the predicted period regularity of each device according to the period distribution corresponding to the predicted electric load data of each device in the electric load data in the current time period.

[0079] The more consistent the predicted load data of a device is with the periodic distribution pattern, the more consistent it is with the actual situation. This embodiment then determines the degree of predicted periodic regularity for each device based on the periodic distribution pattern corresponding to the predicted load data of each device within its current time period. The greater the degree of predicted periodic regularity, the more accurate the predicted mutation degree for the corresponding device.

[0080] In one possible implementation of this embodiment, the method for obtaining the degree of predicted periodic regularity is as follows: for any device, the time corresponding to the electric load data on the device's electric load curve that is equal to the device's predicted electric load data is obtained as the reference time; the duration between any two adjacent reference times is obtained as the first duration; the electric load data corresponding to the center time between any two adjacent reference times (here, assuming that the electric load data corresponding to the extreme point) is obtained as the reference electric load data; the result of adding any two adjacent reference electric load data is used as the specific reference value; the more equal all first durations and all specific reference values ​​are, the more consistent the device's predicted electric load data is with the periodic distribution law, and the result of the negative correlation between the standard deviation of the first duration and the standard deviation of the specific reference value is used as the predicted periodic regularity of the device. In this embodiment, the inverse of the result of the addition of the standard deviation of the first duration and the standard deviation of the specific reference value is raised to the power of an exponential function with a natural constant as the base, and the output of the exponential function is the result of the negative correlation.

[0081] At this point, the degree of regularity of the prediction cycle for each device is obtained.

[0082] Step S203: According to the degree of fluctuation of the electric load of each device and the degree of regularity of the predicted period, the predicted mutation degree of each device is corrected to obtain the corrected predicted mutation degree of each device.

[0083] It is known that the smaller the degree of electric load fluctuation, the more accurate the predicted mutation degree of the corresponding device; the greater the degree of prediction cycle regularity, the more accurate the predicted mutation degree of the corresponding device. Therefore, this embodiment corrects the predicted mutation degree of each device based on the degree of electric load fluctuation and the degree of prediction cycle regularity of each device, and obtains the corrected predicted mutation degree of each device, which is conducive to the subsequent accurate prediction of the charge and discharge bias value of the energy storage system.

[0084] In one possible implementation of this embodiment, the method for obtaining the corrected predicted mutation degree is as follows: the negative correlation result of the electric load fluctuation degree of each device and the predicted periodic regularity are added together and the normalized result is used as the predicted confidence degree of each device; this embodiment respectively uses the inverse of the electric load fluctuation degree of each device as the power of an exponential function with a natural constant as the base, and the output result of the exponential function is the negative correlation result of the electric load fluctuation degree; in addition, this embodiment uses the norm normalization function to normalize the sum of the negative correlation result of the electric load fluctuation degree of each device and the predicted periodic regularity. Among them, the greater the prediction confidence degree, the more accurate the predicted mutation degree of the corresponding device, and the smaller the degree of correction required. Therefore, this embodiment uses the product of the prediction confidence degree of each device and its predicted mutation degree as the corrected predicted mutation degree of each device.

[0085] At this point, the corrected predicted mutation level of each device is obtained.

[0086] Step S204: add up the corrected predicted mutation levels of all devices and perform normalization to obtain the result as the overall predicted mutation level of the energy storage system.

[0087] By using the corrected predicted mutation level for each device, we can accurately analyze the overall predicted mutation level of the energy storage system, effectively avoiding inaccurate predictions for the overall energy storage system due to incomplete device analysis. Furthermore, this embodiment adds the corrected predicted mutation levels for all devices and normalizes the result to serve as the overall predicted mutation level for the energy storage system. This embodiment normalizes the sum of the corrected predicted mutation levels for all devices using the norm normalization function.

[0088] At this point, the overall predicted mutation degree of the energy storage system is obtained.

[0089] Step S4: Based on the charge and discharge bias value in the current time period, the predicted charge and discharge bias value of the energy storage system at the next moment of the current moment is predicted; according to the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree, the corrected predicted charge and discharge bias value of the energy storage system is obtained to predict the charge and discharge of the virtual power plant.

[0090] Specifically, in order to accurately predict the charge and discharge conditions of the energy storage system at the next moment after the current moment, this embodiment first predicts the predicted charge and discharge bias value of the energy storage system at the next moment after the current moment based on the charge and discharge bias value in the current time period through an unscented Kalman filter. It is known that the predicted charge and discharge bias value directly obtained through the unscented Kalman filter will deviate from the actual existence. This embodiment corrects the difference between the predicted charge and discharge bias value and the charge and discharge bias value at the current moment through the overall predicted mutation degree, so that the predicted change of the charge and discharge bias value can be obtained more accurately, and ultimately the more realistic charge and discharge bias value of the energy storage system can be accurately predicted, thereby more accurately predicting the charge and discharge conditions of the energy storage system. Therefore, this embodiment obtains the corrected predicted charge and discharge bias value of the energy storage system based on the charge and discharge bias value at the current moment, the predicted charge and discharge bias value, and the overall predicted mutation degree, and predicts the charge and discharge of the virtual power plant.

[0091] Preferably, in one achievable method of this embodiment, the method for obtaining the corrected predicted charge and discharge bias value is: taking the difference between the predicted charge and discharge bias value and the charge and discharge bias value at the current moment as the bias prediction adjustment value; taking the product of the overall predicted mutation degree and the bias prediction adjustment value as the corrected bias prediction adjustment value of the energy storage system; and then taking the sum of the charge and discharge bias value at the current moment and the corrected bias prediction adjustment value as the corrected predicted charge and discharge bias value of the energy storage system, which is conducive to the subsequent accurate prediction of the charge and discharge conditions of the energy storage system, that is, the prediction of the charge and discharge of the virtual power plant.

[0092] Preferably, in one possible implementation of this embodiment, the method for predicting the charging and discharging of the virtual power plant is as follows: this embodiment sets the preset specified threshold value to 0, and the implementer can set the size of the preset specified threshold value according to actual conditions, which is not limited here. When the corrected predicted charge and discharge bias value is greater than the preset specified threshold value, it means that the overall power consumption of the virtual power plant is greater than the power generation. At this time, it is predicted that the virtual power plant, i.e., the energy storage system, performs a discharge operation and the size of the discharge load is the corrected predicted charge and discharge bias value, ensuring that the power consumption in the virtual power plant remains stable; when the corrected predicted charge and discharge bias value is less than the preset specified threshold value, it means that the overall power consumption of the virtual power plant is less than the power generation. In order to avoid excessive power causing excessive voltage in some parts of the power grid, causing losses to some equipment, the virtual power plant is predicted to perform a charging operation and the size of the charging load is the absolute value of the corrected predicted charge and discharge bias value; when the corrected predicted charge and discharge bias value is equal to the preset specified threshold value, it means that the overall power consumption and power generation of the virtual power plant are maintained in a stable stage, and at this time, it is predicted that the charge and discharge operation of the virtual power plant will not change.

[0093] In summary, this embodiment obtains the electric load data of the equipment in the virtual power plant and the charge and discharge bias value of the energy storage system in real time; obtains the predicted mutation degree based on the change of the predicted electric load data of the equipment relative to the electric load data at the current moment; corrects the predicted mutation degree based on the change of the electric load data and the periodic distribution of the predicted electric load data to obtain the overall predicted mutation degree of the energy storage system; obtains the corrected predicted charge and discharge bias value of the energy storage system based on the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree to predict the charge and discharge of the virtual power plant. The present invention effectively improves the accuracy of the charge and discharge prediction of the virtual power plant by accurately obtaining the corrected predicted charge and discharge bias value of the energy storage system, which is beneficial to maintaining the stability of the overall operation of the power grid.

[0094] Example 2:

[0095] The present invention also proposes a virtual power plant load AI prediction system with dynamic response on the demand side, please refer to Figure 3, which shows a structural diagram of a virtual power plant load AI prediction system with dynamic response on the demand side provided by an embodiment of the present invention. The system includes: a data acquisition module 10, a prediction mutation degree acquisition module 20, an overall prediction mutation degree acquisition module 30 and a charge and discharge prediction module 40.

[0096] The data acquisition module 10 is used to obtain the power load data of each device in the virtual power plant at each moment and the charge and discharge bias value of the energy storage system at each moment in the current time period.

[0097] The predicted mutation degree acquisition module 20 is used to predict the predicted electric load data of each device at the next moment after the current moment based on the electric load data of each device in the current time period; according to the difference between the predicted electric load data of each device and its electric load data at the current moment, and the trend change of the predicted electric load data relative to its electric load data at the current moment, the predicted mutation degree of each device is obtained.

[0098] The overall predicted mutation degree acquisition module 30 is used to correct the predicted mutation degree based on the changes in the electric load data of each device in the current time period and the corresponding periodic distribution of the predicted electric load data of each device in the electric load data in its current time period, so as to obtain the overall predicted mutation degree of the energy storage system.

[0099] The charge and discharge prediction module 40 is used to predict the predicted charge and discharge bias value of the energy storage system at the next moment after the current moment based on the charge and discharge bias value in the current time period; according to the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree, the corrected predicted charge and discharge bias value of the energy storage system is obtained to predict the charge and discharge of the virtual power plant.

[0100] It should be noted that the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer equipment can be divided into different functional modules to complete all or part of the functions described above. In addition, the above embodiment provides a virtual power plant load AI prediction system with dynamic response on the demand side and an embodiment of a virtual power plant load AI prediction method with dynamic response on the demand side, which belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0101] Example 3:

[0102] The present invention also proposes an AI-powered load prediction device for a virtual power plant with dynamic demand-side response, comprising a memory and a processor. The memory stores executable program code, and the processor is configured to call and execute the executable program code to implement an AI-powered load prediction method for a virtual power plant with dynamic demand-side response, as provided in an embodiment of the present application. The device can be a chip, component, or module, and the chip can include a connected processor and memory. The memory is configured to store instructions, and when the processor calls and executes the instructions, the chip can execute an AI-powered load prediction method for a virtual power plant with dynamic demand-side response, as provided in the above embodiment.

[0103] In addition, the present application also protects a computer device, see Figure 4 The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402, wherein when the processor 402 executes the computer program 403, the computer device can execute any one of the above-mentioned demand-side dynamic response virtual power plant load AI prediction methods.

[0104] Example 4:

[0105] This embodiment also provides a computer-readable storage medium, which stores computer program code. When the computer program code runs on a computer, the computer executes the above-mentioned related method steps to implement a virtual power plant load AI prediction method with dynamic response on the demand side provided by the above embodiment.

[0106] Example 5:

[0107] This embodiment also provides a computer program product. When the computer program product runs on a computer, it enables the computer to execute the above-mentioned related steps to implement a virtual power plant load AI prediction method with dynamic response on the demand side provided by the above embodiment.

[0108] Among them, the device, computer-readable storage medium, computer program product or chip provided in this embodiment are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0109] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0110] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A virtual power plant load AI prediction method with dynamic response on the demand side, characterized by: The method comprises the following steps: Obtain the power load data of each device in the virtual power plant at each moment and the charge and discharge bias value of the energy storage system at each moment during the current time period; Based on the power load data of each device in the current time period, the predicted power load data of each device at the next moment after the current moment is predicted; based on the difference between the predicted power load data of each device and its power load data at the current moment, as well as the trend change of the predicted power load data relative to its power load data at the current moment, the predicted mutation degree of each device is obtained; Based on the change in the electric load data of each device in the current time period and the corresponding periodic distribution of the predicted electric load data of each device in the electric load data in the current time period, the predicted mutation degree is corrected to obtain the overall predicted mutation degree of the energy storage system; Based on the charge and discharge bias value in the current time period, the predicted charge and discharge bias value of the energy storage system at the next moment after the current moment is predicted; according to the charge and discharge bias value at the current moment, the predicted charge and discharge bias value and the overall predicted mutation degree, the corrected predicted charge and discharge bias value of the energy storage system is obtained to predict the charge and discharge of the virtual power plant.

2. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 1, characterized in that: The method for obtaining the predicted mutation degree is: For any device, the difference between the predicted power load data of the device and its current power load data is used as the predicted change value of the device; The electrical load data of the device at the current moment, the moment before the current moment, and the moment after the current moment are all represented by coordinate points in a two-dimensional coordinate system; wherein, in the two-dimensional coordinate system, the time is the horizontal axis and the electrical load data is the vertical axis; Obtain the slope of the line segment connecting the coordinate point at the previous moment and the current moment as the first slope; Obtain the slope of the line segment connecting the coordinate points at the current moment and the next moment after the current moment as the second slope; The difference between the first slope and the second slope is used as the predicted trend change degree of the device; The result of normalizing the product of the predicted trend change degree and the predicted change value is taken as the predicted mutation degree of the device.

3. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 1, characterized in that: The method for obtaining the overall predicted mutation degree is: Obtain the degree of fluctuation of the power load of each device based on the changes in the power load data of each device in the current time period; Obtain the degree of regularity of the predicted period of each device based on the period distribution of the predicted electric load data of each device in the electric load data of the current time period; According to the power load fluctuation degree and the prediction period regularity degree of each device, the predicted mutation degree of each device is corrected to obtain the corrected predicted mutation degree of each device; The corrected predicted mutation levels of all devices are added together and normalized to obtain the overall predicted mutation level of the energy storage system.

4. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 3, characterized in that: The method for obtaining the degree of electric load fluctuation is: For any device, the electrical load data of the device in the current time period is fitted into an electrical load curve according to the time sequence, and the difference in electrical load data corresponding to any two adjacent extreme value points on the electrical load curve is obtained as the first difference; obtaining a difference between the maximum electric load data and the minimum electric load data on the electric load curve as a second difference; The product of the number of extreme value points on the electric load curve, the mean of the first difference and the second difference is taken as the electric load fluctuation degree of the equipment.

5. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 4, characterized in that: The method for obtaining the degree of regularity of the prediction cycle is: For any device, the time corresponding to the electric load data on the electric load curve of the device that is equal to the predicted electric load data of the device is obtained, and is used as the reference time; Get the duration between any two adjacent reference moments as the first duration; Obtain the electric load data corresponding to the center time between any two adjacent reference times, and use them as reference electric load data; The sum of any two adjacent reference electric load data is used as a specific reference value; The standard deviation of the first time period is added to the standard deviation of the specific reference value, and the result of the negative correlation is used as the predicted period regularity of the device.

6. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 3, characterized in that: The method for obtaining the modified predicted mutation degree is: The negative correlation result of the power load fluctuation degree of each device and the degree of prediction period regularity are added and normalized, and the result is used as the prediction confidence level of each device; The product of the prediction confidence level of each device and its predicted mutation level is taken as the corrected predicted mutation level of each device.

7. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 1, characterized in that: The method for obtaining the corrected predicted charge and discharge bias value is: The difference between the predicted charge and discharge bias value and the current charge and discharge bias value is used as the bias prediction adjustment value; The product of the overall predicted mutation degree and the biased prediction adjustment value is used as the corrected biased prediction adjustment value of the energy storage system; The sum of the current charge and discharge bias value and the corrected bias prediction adjustment value is used as the corrected predicted charge and discharge bias value of the energy storage system.

8. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 1, characterized in that: The method for predicting the charging and discharging of the virtual power plant is as follows: When the modified predicted charge-discharge bias value is greater than a preset threshold, the virtual power plant is predicted to perform a discharge operation and the magnitude of the discharge load is the modified predicted charge-discharge bias value; When the corrected predicted charge-discharge bias value is less than a preset threshold, the virtual power plant is predicted to perform charging operations and the magnitude of the charging load is the absolute value of the corrected predicted charge-discharge bias value; When the corrected predicted charge and discharge bias value is equal to the preset specified threshold, the charge and discharge operation of the predicted virtual power plant is not changed.

9. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 1, characterized in that: The method for obtaining the predicted electric load data is: Based on the power load data of each device in the current time period, the predicted power load data of each device at the next moment after the current moment is predicted through unscented Kalman filtering; The method for obtaining the predicted charge and discharge bias value is: Based on the charge and discharge bias value in the current time period, the predicted charge and discharge bias value of the energy storage system at the next moment after the current moment is predicted through unscented Kalman filtering.

10. The AI ​​prediction method for virtual power plant load with dynamic demand response according to claim 1, characterized in that: The method for obtaining the charge and discharge bias value is: The difference between the power load data and the power generation load data of the energy storage system at each moment is used as the charge and discharge bias value of the energy storage system at each moment.

Images