Power Internet of Things load forecasting method and system based on pre-trained large model

Abstract

The present invention relates to the technical field of power consumption data processing, and in particular to a method and system for load forecasting of an electric power Internet of Things based on a pre-trained large model. The present invention analyzes the power consumption distribution of different devices at different times within the neighborhood range of each moment, obtains the power consumption correlation degree between devices at each moment, and obtains the associated devices of each device; obtains the fault attention factor of each device at each moment according to the changing trend of the number of associated devices at different times within the neighborhood range of each device at each moment, the power consumption correlation degree of each device at each moment, and the distribution characteristics at each moment; obtains the overall correction amount of power consumption at each moment; and performs load forecasting on the electric power Internet of Things at the next moment according to the distribution of the overall correction amount of power consumption at each moment within the neighborhood range of the real time moment. The present invention analyzes the correlation and fluctuation contribution between devices, accurately obtains the attention situation at each moment, and improves the effectiveness of the overall load forecast.

Description

Technical Field

[0001] The present invention relates to the technical field of power consumption data processing, and in particular to a method and system for power Internet of Things load prediction based on a pre-trained large model. Background Art

[0002] Power Internet of Things load forecasting refers to the process of using Internet of Things technology to collect real-time power consumption data of various equipment in the power system, and combining big data analysis, machine learning, artificial intelligence and other technologies to predict and analyze the power load in the future. By predicting the load change trend, it provides support for the safe operation of the power system, resource optimization and scientific decision-making.

[0003] In the existing technology, the overall load changes of power supply equipment are predicted through pre-trained large models. However, the operating modes of different devices vary greatly, and the load changes of a certain device may have a significant impact on the overall load. Predicting only the overall load changes will result in the inability to capture the differentiated load demands between devices, and the overall load prediction effect is poor. Summary of the Invention

[0004] In order to solve the technical problem that only predicting the overall load change will fail to capture the differentiated load demands between devices, resulting in poor overall load prediction effect, the purpose of the present invention is to provide a power Internet of Things load prediction method and system based on a pre-trained large model. The technical solution adopted is as follows:

[0005] The present invention proposes a method for load forecasting of the power Internet of Things based on a pre-trained large model, the method comprising:

[0006] Obtain the power consumption of various devices in the power Internet of Things at every moment;

[0007] Based on the power consumption distribution of each device at different times within its neighborhood at each moment, the power consumption fluctuation degree of each device at each moment is obtained; based on the similarity of the power consumption fluctuation degree of all moments within the neighborhood of different times, the power consumption correlation degree between devices at each moment is obtained, and the associated devices of each device are obtained;

[0008] Based on the power consumption correlation between different devices at each moment and the number distribution of associated devices, the fluctuation contribution factor of each device at each moment is obtained. Based on the changing trend of the number of associated devices at different moments in the neighborhood of each device at each moment, the fluctuation contribution factor of each device at each moment, and the distribution characteristics at each moment, the fault attention factor of each device at each moment is obtained. Based on the fault attention factors and power consumption of different devices at each moment, the overall power consumption correction value at each moment is obtained.

[0009] According to the distribution of the overall correction amount of power consumption at each moment in the neighborhood range of the real time, the overall correction amount of power consumption is input into the prediction model to perform load prediction on the power Internet of Things.

[0010] Furthermore, the method for obtaining the power consumption fluctuation degree includes:

[0011] Obtain a fitting curve based on the power consumption of the device at all times within the neighborhood of each time, obtain the extreme value in the fitting curve, and use the time corresponding to the extreme value as the target time;

[0012] Based on the number of target moments in the neighborhood of each moment, the absolute values ​​of the derivatives between adjacent target moments, and the corresponding power consumption differences between adjacent target moments, the power consumption fluctuation degree of each device at each moment is obtained. The number of target moments, the absolute values ​​of the derivatives, and the power consumption differences are all positively correlated with the power consumption fluctuation degree.

[0013] Furthermore, the method for obtaining the power consumption correlation degree includes:

[0014] According to the similarity of power consumption fluctuations at all times between devices in the neighborhood at different times, the power consumption related time period within the neighborhood at each time between the devices is obtained;

[0015] Based on the total duration of the power consumption-related periods within the neighborhood range between devices at each moment, the correlation coefficient of the power consumption fluctuation degree at all moments within different power consumption-related periods, and the duration of the power consumption-related periods, the power consumption correlation degree between devices at each moment is obtained. The total duration of the power consumption-related periods, the correlation coefficient of the power consumption fluctuation degree at all moments within the power consumption-related periods, and the duration of the power consumption-related periods are all positively correlated with the power consumption correlation degree.

[0016] Furthermore, the method for obtaining the power consumption related time period includes:

[0017] In the neighborhood of each moment, the correlation coefficient of the series of power consumption fluctuations between devices at all moments is obtained as the local correlation between devices at each moment.

[0018] The signs of the local correlation values ​​between the devices at each moment are obtained. If the signs of consecutive adjacent moments are the same, the range of the corresponding consecutive adjacent moments is used as the power consumption correlation period.

[0019] Furthermore, the method for obtaining the associated device includes:

[0020] If the power consumption correlation between the devices at each moment is greater than a preset threshold, the corresponding devices are considered as associated devices.

[0021] Furthermore, the method for obtaining the fluctuation contribution factor includes:

[0022] Based on the number of associated devices of each device at each moment and the degree of power consumption correlation between each device and other associated devices at each moment, the fluctuation contribution factor of each device at each moment is obtained. The number of associated devices and the degree of power consumption correlation are both positively correlated with the fluctuation contribution factor.

[0023] Furthermore, the method for obtaining the fault attention factor includes:

[0024] Obtain the failure factor of each device at each moment based on the changing trend of the number of associated devices at different moments within the neighborhood of each device at each moment;

[0025] If the failure factor of a device at any moment is greater than the preset failure threshold, the corresponding device will be regarded as the faulty device at each moment;

[0026] According to the failure factors and fluctuation contribution factors of different faulty equipment at each moment and the time difference between each moment and the real time, the failure attention factor at each moment is obtained. The failure factor and fluctuation contribution factor are both positively correlated with the failure attention factor, while the time difference is negatively correlated with the failure attention factor.

[0027] Furthermore, the method for obtaining the fault factor includes:

[0028] Obtain a quantity fitting line for fitting the quantity of associated devices at all times within the neighborhood of each device at each moment, and obtain the absolute value of the slope of the quantity fitting line as the quantity fluctuation coefficient of each device at each moment;

[0029] Based on the difference fluctuation degree of the number of associated devices between different adjacent moments within the neighborhood of each device at each moment, as well as the quantity fluctuation coefficient, the failure factor of each device at each moment is obtained. The difference fluctuation degree and the first fluctuation coefficient are both positively correlated with the failure factor.

[0030] Furthermore, the method for obtaining the overall correction amount of power consumption includes:

[0031] The cumulative power consumption of all devices at each moment is obtained as the overall power consumption at each moment; the product of the overall power consumption at each moment and the fault attention factor is obtained as the overall power consumption correction at each moment.

[0032] The present invention also proposes a power Internet of Things load forecasting system based on a pre-trained large model, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, it implements any one of the steps of the power Internet of Things load forecasting method based on a pre-trained large model.

[0033] The present invention has the following beneficial effects:

[0034] The present invention obtains the power consumption fluctuation degree of each device at each moment based on the power consumption distribution of each device at different moments in the neighborhood range of each moment, accurately captures the power consumption fluctuation of the device, and thus understands the dynamic changes in the power consumption of the device; according to the similarity of the power consumption fluctuation degree of all moments in the neighborhood range of different moments between the devices, the power consumption correlation degree between the devices at each moment is obtained, and the associated devices of each device are obtained, revealing the power consumption correlation relationship between the devices. Identifying the associated devices is helpful for more comprehensive analysis of the power consumption pattern of the power Internet of Things; according to the power consumption correlation degree between different devices at each moment and the number distribution of associated devices, the fluctuation pattern of each device at each moment is obtained. Dynamic contribution factor, more accurately evaluate the contribution of each device to the overall power consumption fluctuation; according to the changing trend of the number of associated devices at different times within the neighborhood of each device at each moment, the fluctuation contribution factor of each device at each moment and the distribution characteristics at the moment, the fault attention factor of each device at each moment is obtained, which can more comprehensively reflect the failure risk of the device; according to the fault attention factor and power consumption of different devices at each moment, the overall power consumption correction amount at each moment is obtained, which more accurately reflects the overall power consumption of the power Internet of Things at each moment; according to the distribution of the overall power consumption correction amount at each moment within the neighborhood of the real time moment, the load of the power Internet of Things is predicted. The present invention analyzes the association and fluctuation contribution between devices, accurately obtains the attention situation at each moment, and improves the effectiveness of the overall load prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] 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.

[0036] Figure 1 A flowchart of a method for load forecasting of the power Internet of Things based on a pre-trained large model provided by one embodiment of the present invention;

[0037] Figure 2 A flowchart of a method for obtaining a fault attention factor provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effectiveness of a method and system for load forecasting in the power Internet of Things based on a pre-trained large model proposed by the present invention. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0039] 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.

[0040] The following describes in detail a specific solution of a power Internet of Things load forecasting method and system based on a pre-trained large model provided by the present invention in conjunction with the accompanying drawings.

[0041] See also Figure 1 , which shows a method flow chart of a method for load forecasting of the power Internet of Things based on a pre-trained large model provided by one embodiment of the present invention, specifically including:

[0042] Step S1: Obtain the power consumption of various devices in the power Internet of Things at each moment.

[0043] In this embodiment of the present invention, to avoid only considering overall load changes while ignoring the differentiated load demands between devices, which would result in prediction results that fail to accurately reflect the impact of sudden load fluctuations on certain devices on the power system and thus affect the accuracy of scheduling decisions, it is necessary to analyze the power consumption of different devices to more accurately understand their operating modes and load fluctuation characteristics. First, smart meters are connected to the power supply circuits of all devices to obtain the power consumption of multiple devices at each moment for analysis.

[0044] It should be noted that, in an embodiment of the present invention, the collection time range is set to one week, and the collection frequency is once per second, that is, data collection is performed with an interval of 1 second between moments. Timestamp synchronization is used to ensure that the power consumption of devices at the same moment is strictly aligned, thereby accurately analyzing the load coupling relationship between devices. In other embodiments of the present invention, the intervals between moments can be set according to specific circumstances, and are not limited or elaborated here.

[0045] It should be noted that in order to facilitate subsequent data processing, the power consumption data is normalized by decimal calibration to have similar scale and distribution for better comparison. The specific decimal calibration and normalization process is a technical means well known to those skilled in the art and will not be elaborated here.

[0046] Step S2: Based on the power consumption distribution of each device at different times within the neighborhood of each device at each moment, the power consumption fluctuation degree of each device at each moment is obtained; based on the similarity of the power consumption fluctuation degree of all moments within the neighborhood of different moments between the devices, the power consumption correlation degree between the devices at each moment is obtained, and the associated devices of each device are obtained.

[0047] The operating modes of different devices vary greatly, and the load demand fluctuates with changes in the production process. By analyzing the changing trend of power consumption within the range, the dynamic changes in power consumption of devices within the range can be reflected. According to the power consumption distribution of each device at different times within the neighborhood range at each moment, the degree of power consumption fluctuation of each device at each moment can be obtained.

[0048] Preferably, in one embodiment of the present invention, the method for obtaining the power consumption fluctuation degree includes:

[0049] Obtain a fitting curve based on the power consumption of the device at all times within the neighborhood of each time, obtain the extreme value in the fitting curve, and use the time corresponding to the extreme value as the target time;

[0050] Among them, in some embodiments of the present invention, the fitting is performed using the least squares method, and polynomial fitting can also be used for curve fitting. The method for obtaining the extreme value is: by calculating the derivative of the curve at each moment to reflect the rate of change of the curve at a certain moment, when the derivative is 0, it means that the rate of change of the curve near the moment is extremely small, and the power consumption at the corresponding moment is at the maximum or minimum situation nearby, that is, the data with a derivative of 0 on the curve is the extreme value, which allows a more intuitive and clear understanding of the dynamic change trend of power consumption. The specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0051] Based on the number of target moments in the neighborhood of each moment, the absolute values ​​of the derivatives between adjacent target moments, and the difference in corresponding power consumption between adjacent target moments, the power consumption fluctuation degree of each device at each moment is obtained. The number of target moments, the absolute values ​​of the derivatives, and the difference in power consumption are all positively correlated with the power consumption fluctuation degree.

[0052] It should be noted that since the target moment represents the moment corresponding to the extreme value within the field, the more extreme moments there are, the more rise and fall changes will occur, and the greater the fluctuation in power consumption; the absolute value of the derivative reflects the instantaneous intensity or rate of data change. The larger the absolute value of the derivative, the greater the rate of change in power consumption and the greater the fluctuation in data; the power consumption difference reflects the deviation of the power consumption of the equipment when it changes at any moment. The greater the difference, the more inconsistent the power consumption of the equipment at different moments, and the greater the degree of power consumption fluctuation. Therefore, the number of target moments, the absolute value of the derivative and the difference in power consumption can be comprehensively analyzed to analyze the power consumption situation. The number of target moments, the absolute value of the derivative and the difference in power consumption are positively correlated with the degree of power consumption fluctuation.

[0053] In one embodiment of the present invention, in order to eliminate the influence of the size of the neighborhood range, the number of target moments in the neighborhood range of each moment is normalized, that is, the ratio of the number of target moments in the neighborhood range of each moment to the number of all moments in the neighborhood range is calculated as the first fluctuation coefficient; the mean of the absolute values ​​of the derivatives of all moments between adjacent target moments is obtained, and the difference in corresponding power consumption between adjacent target moments is obtained to reflect the data fluctuation characteristics between adjacent moments; the product of the mean of the absolute values ​​of the derivatives between adjacent target moments and the difference in power consumption is calculated as the data fluctuation coefficient between adjacent target moments; all adjacent target moments in the neighborhood range are analyzed to obtain the cumulative sum of the data fluctuation coefficients between all adjacent target moments as the second fluctuation coefficient; the product of the first fluctuation coefficient and the second fluctuation coefficient is obtained as the degree of power consumption fluctuation; therefore, based on the above-mentioned basic number operations, a correlation between the number of target moments, the absolute value of the derivative, the difference in power consumption and the degree of power consumption fluctuation is constructed, that is, the larger the number of target moments, the larger the absolute value of the derivative, and the larger the difference in power consumption, the greater the change in data, the more unstable the change in power consumption, and the greater the degree of power consumption fluctuation.

[0054] It should be noted that, in one embodiment of the present invention, the method for obtaining the neighborhood range is based on each moment and a range consisting of a preset number of historical moments, where the preset number is 50; in other embodiments of the present invention, the neighborhood range can be set according to specific circumstances, and is not limited or elaborated here.

[0055] The similarity of power consumption fluctuations between devices reflects the degree of mutual influence between them. The greater the similarity, the greater the mutual influence between the devices, and the stronger the correlation between their power consumption. Based on the similarity of power consumption fluctuations between devices at all times within their neighborhood, the power consumption correlation between devices at each moment is determined, along with each device's associated devices.

[0056] Preferably, in one embodiment of the present invention, the method for obtaining the power consumption correlation degree includes:

[0057] According to the similarity of power consumption fluctuations at all times between devices in the neighborhood at different times, the power consumption related time period within the neighborhood at each time between the devices is obtained;

[0058] It should be noted that, in one embodiment of the present invention, the method for obtaining the power consumption related time period includes:

[0059] In the neighborhood of each moment, the correlation coefficient of the series of power consumption fluctuations between devices at all moments is obtained as the local correlation between devices at each moment.

[0060] The signs of the local correlation values ​​between the devices at each moment are obtained. If the signs of consecutive adjacent moments are the same, the range of the corresponding consecutive adjacent moments is used as the power consumption correlation period.

[0061] It should be noted that in the embodiment of the present invention, the correlation coefficient is the Pearson correlation coefficient, and the value range is -1 to 1. The closer it is to 1, the greater the positive correlation, and the closer it is to -1, the greater the negative correlation. The specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0062] For example, if there is a local similarity symbol of -1, -1, 1, -1, 1, 1, -1, 1, 1 in the neighborhood range of each moment, and if there are consecutive adjacent moments with the same symbol (-1 -1), (1 1), (1 1), the corresponding time range will be used as the power consumption related period for subsequent analysis.

[0063] Based on the total duration of the power consumption-related periods within the neighborhood range between devices at each moment, the correlation coefficient of the power consumption fluctuation degree at all moments within different power consumption-related periods, and the duration of the power consumption-related periods, the power consumption correlation degree between devices at each moment is obtained. The total duration of the power consumption-related periods, the correlation coefficient of the power consumption fluctuation degree at all moments within the power consumption-related periods, and the duration of the power consumption-related periods are all positively correlated with the power consumption correlation degree.

[0064] In one embodiment of the present invention, the total duration of the power-consumption-related time periods within the neighborhood range of each moment is normalized, that is, the ratio of the total duration of the power-consumption-related time periods within the neighborhood range of each moment to the duration within the neighborhood range is calculated as the power-consumption-related time proportion; the average of the products of the duration of all power-consumption-related time periods within the neighborhood range of each moment and the correlation coefficient of the power consumption fluctuation degree at all moments is obtained as the overall correlation coefficient; the product between the power-consumption-related time proportion and the overall correlation coefficient is obtained and normalized as the power consumption correlation degree; therefore, based on the above basic mathematical operations, the total duration of the power-consumption-related time periods, the correlation coefficient of the power consumption fluctuation degree at all moments within the power-consumption-related time periods, and the correlation between the duration of the power-consumption-related time periods and the power consumption correlation degree are constructed, that is, the greater the total duration of the power-consumption-related time periods, the greater the correlation coefficient of the power consumption fluctuation degree at all moments within the power-consumption-related time periods, the longer the duration of the power-consumption-related time periods, the more moments with greater mutual influence, and the greater the power consumption correlation degree.

[0065] It should be noted that, in the embodiment of the present invention, normalization can be performed through linear normalization or a normalization function. The specific means are technical means well known to those skilled in the art and will not be described in detail here.

[0066] Preferably, in one embodiment of the present invention, the method for obtaining an associated device includes:

[0067] If the power consumption correlation between the devices at each moment is greater than a preset threshold, the corresponding devices are considered as associated devices.

[0068] It should be noted that, in one embodiment of the present invention, the size of the preset degree threshold is 0.7; in other embodiments of the present invention, the size of the preset degree threshold can be set according to specific circumstances, which is not limited or elaborated here.

[0069] Step S3: Based on the power consumption correlation degree between different devices at each moment and the number distribution of associated devices, obtain the fluctuation contribution factor of each device at each moment; based on the changing trend of the number of associated devices at different moments in the neighborhood range of each device at each moment, the fluctuation contribution factor of each device at each moment and the distribution characteristics at each moment, obtain the fault attention factor of each device at each moment; based on the fault attention factor and power consumption of each device at different moments in the neighborhood range of each device at different moments, obtain the overall power consumption correction of each device at each moment.

[0070] Multiple devices together constitute the source of power load. The status of each device has different degrees of influence on the fluctuation of the overall power load. The more associated devices each device has, the greater the power consumption correlation, and the greater the degree of mutual influence between devices. When there is abnormal fluctuation in a device, it will affect other devices to a greater extent. The more significant the impact on the fluctuation of the overall load, the greater the fluctuation contribution factor. Therefore, by comprehensively analyzing the number of associated devices and the size of the correlation degree, the fluctuation contribution factor of each device is quantified. Therefore, according to the power consumption correlation degree and the number distribution of associated devices at each moment between different devices, the fluctuation contribution factor of each device at each moment is obtained. The number of associated devices and the power consumption correlation degree are both positively correlated with the fluctuation contribution factor.

[0071] Preferably, in one embodiment of the present invention, the method for obtaining the fluctuation contribution factor includes:

[0072] Based on the number of associated devices of each device at each moment and the degree of power consumption correlation between each device and other associated devices at each moment, the fluctuation contribution factor of each device at each moment is obtained. The number of associated devices and the degree of power consumption correlation are both positively correlated with the fluctuation contribution factor.

[0073] In one embodiment of the present invention, the number of associated devices of each device at each moment is normalized, that is, the ratio of the number of associated devices of each device at each moment to the number of all devices is calculated to quantify the association ratio of each device at each moment; the average power consumption association degree between each device and all other associated devices at each moment is obtained as the influence weight of each device at each moment; the product of the association ratio and the influence weight of each device at each moment is obtained and normalized as the fluctuation contribution factor of each device at each moment; therefore, based on the above-mentioned basic mathematical operations, a correlation between the number of associated devices, the power consumption association degree and the fluctuation contribution factor is constructed, that is, the larger the number of associated devices, the greater the power consumption association degree, the greater the impact of the device on the whole, and the greater the fluctuation contribution factor.

[0074] Under normal operating conditions, the power consumption data of mutually related devices show a stable correlation. If the degree of correlation between a device and its related devices suddenly drops, it may indicate that the device is operating abnormally. The changing trend of the number of related devices can reflect the fluctuation of related devices within the neighborhood. The fluctuation contribution factor reflects the contribution of the device to the overall fluctuation. The larger the fluctuation contribution factor, the greater the impact on the overall load, and the more attention it needs. Based on the changing trend of the number of related devices at different times within the neighborhood of each device at each moment, the fluctuation contribution factor of each device at each moment, and the distribution characteristics at the moment, the fault attention factor of each device at each moment is obtained.

[0075] Preferably, in one embodiment of the present invention, the method for obtaining the fault attention factor is as follows: Figure 2 , which shows a flow chart of a method for obtaining a fault attention factor, including:

[0076] Step S201: Obtain the failure factor of each device at each moment according to the change trend of the number of associated devices at different moments within the neighborhood of each device at each moment.

[0077] In one embodiment of the present invention, a method for obtaining a fault factor includes:

[0078] Obtain a quantity fitting line for fitting the quantity of associated devices at all times within the neighborhood of each device at each moment, and obtain the absolute value of the slope of the quantity fitting line as the quantity fluctuation coefficient of each device at each moment;

[0079] The failure factor of each device at each moment is obtained based on the difference fluctuation degree of the number of associated devices between different adjacent moments within the neighborhood of each device at each moment, as well as the quantity fluctuation coefficient. Both the difference fluctuation degree and the quantity fluctuation coefficient are positively correlated with the failure factor.

[0080] It should be noted that, in the embodiments of the present invention, the degree of difference fluctuation can be reflected by calculating the variance or standard deviation of the difference. The larger the variance or standard deviation, the greater the degree of difference fluctuation. The specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0081] In one embodiment of the present invention, data fitting is performed using existing fitting methods such as the least squares method. The slope of the straight line can reflect the change in the number of associated devices. The absolute value of the slope is obtained by calculating the difference between the data at both ends of the quantity fitting straight line and the difference between the corresponding moments at both ends. The larger the absolute value of the slope, the greater the change in the number of associated devices, the larger the quantity fluctuation coefficient, and the greater the impact on the fault. The product of the difference fluctuation degree of the number of associated devices between different adjacent moments and the quantity fluctuation coefficient is calculated and normalized to obtain the fault factor of each device at each moment. Therefore, based on the above basic mathematical operations, a correlation between the difference fluctuation degree and the quantity fluctuation coefficient and the fault factor is constructed, that is, the greater the difference fluctuation degree, the larger the quantity fluctuation coefficient, the more unstable the change in the number of associated devices, the more likely a fault will occur, and the larger the fault factor.

[0082] Step S202: If there is a device whose failure factor at each moment is greater than a preset failure threshold, the corresponding device is regarded as a failed device at each moment.

[0083] It should be noted that, in one embodiment of the present invention, the preset fault threshold is 0.8. In other embodiments of the present invention, the preset fault threshold may be set according to specific circumstances, which is not limited or elaborated herein.

[0084] Step S203: Obtain the fault concern factor at each moment based on the fault factor, fluctuation contribution factor of different faulty devices at each moment, and the moment difference between each moment and the real time moment. The fault factor and fluctuation contribution factor are both positively correlated with the fault concern factor, and the moment difference is negatively correlated with the fault concern factor.

[0085] In one embodiment of the present invention, the difference between each moment and the real-time moment is obtained and normalized and mapped as the time weight at each moment. The product of the fault factor and the fluctuation contribution factor of all faulty devices at each moment is calculated and accumulated as the overall fault influence coefficient; the product of the time weight at each moment and the overall fault influence coefficient is obtained and normalized as the fault attention factor at each moment; therefore, based on the above basic mathematical operations, a correlation between the moment, the fault factor and the fluctuation contribution factor and the fault attention factor is constructed, that is, the larger the moment, the farther each moment is from the real-time moment, the smaller the fault factor, the smaller the fluctuation contribution factor, the smaller the impact between devices, and the smaller the fault attention factor.

[0086] The fault attention factor can reflect the degree of mutual influence between devices. The stronger the correlation, the more attention should be paid to abnormal fluctuations in the device. The larger the fault attention factor, the greater the proportion of power consumption analysis. Based on the fault attention factor and power consumption of different devices at each moment, the overall power consumption correction at each moment is obtained.

[0087] Preferably, in one embodiment of the present invention, the method for obtaining the overall correction amount of power consumption includes:

[0088] The cumulative power consumption of all devices at each moment is obtained as the overall power consumption at each moment; the product of the overall power consumption at each moment and the fault attention factor is obtained as the overall power consumption correction at each moment.

[0089] Step S4: Based on the distribution of the overall correction amount of power consumption at each moment within the neighborhood range of the real time moment, the overall correction amount of power consumption is input into the prediction model to perform load prediction on the power Internet of Things.

[0090] By comprehensively analyzing the degree of impact at each moment based on the correlation between the above-mentioned devices and making an overall correction to the power consumption at each moment, the data can be input into a large prediction model, such as a long short-term memory network (LSTM), for training to establish a load forecasting model to predict the overall load value at the next moment. This ensures that the model accurately captures the differentiated load demands between devices and improves the accuracy of power load forecasting.

[0091] In summary, the present invention analyzes the power consumption distribution of different devices at different times within the neighborhood range of each moment, obtains the power consumption correlation degree between devices at each moment, and obtains the associated devices of each device; according to the changing trend of the number of associated devices at different times within the neighborhood range of each device at each moment, the power consumption correlation degree of each device at each moment and the distribution characteristics at each moment, obtains the fault attention factor of each device at each moment; obtains the overall power consumption correction amount at each moment; according to the overall power consumption correction amount distribution at each moment within the neighborhood range of the real time moment, the load forecast of the power Internet of Things at the next moment is performed. The present invention analyzes the correlation and fluctuation contribution between devices, accurately obtains the attention situation at each moment, and improves the effectiveness of the overall load forecast.

[0092] The present invention also proposes a power Internet of Things load forecasting system based on a pre-trained large model, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, it implements any one of the steps of a power Internet of Things load forecasting method based on a pre-trained large model.

[0093] 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.

[0094] 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 method for load forecasting of power Internet of Things based on a pre-trained large model, characterized in that: The method comprises: Obtain the power consumption of various devices in the power Internet of Things at every moment; Based on the power consumption distribution of each device at different times within its neighborhood at each moment, the power consumption fluctuation degree of each device at each moment is obtained; based on the similarity of the power consumption fluctuation degree of all moments within the neighborhood of different times, the power consumption correlation degree between devices at each moment is obtained, and the associated devices of each device are obtained; Based on the power consumption correlation between different devices at each moment and the number distribution of associated devices, the fluctuation contribution factor of each device at each moment is obtained. Based on the changing trend of the number of associated devices at different moments in the neighborhood of each device at each moment, the fluctuation contribution factor of each device at each moment, and the distribution characteristics at each moment, the fault attention factor of each device at each moment is obtained. Based on the fault attention factors and power consumption of different devices at each moment, the overall power consumption correction value at each moment is obtained. According to the distribution of the overall correction amount of power consumption at each moment in the neighborhood of the real time, the overall correction amount of power consumption is input into the prediction model to perform load forecasting on the power Internet of Things; The method for obtaining the neighborhood range is to use each moment as a benchmark and a range formed by a preset number of historical moments.

2. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 1 is characterized in that: The method for obtaining the power consumption fluctuation degree includes: Obtain a fitting curve based on the power consumption of the device at all times within the neighborhood of each time, obtain the extreme value in the fitting curve, and use the time corresponding to the extreme value as the target time; Based on the number of target moments in the neighborhood of each moment, the absolute values ​​of the derivatives between adjacent target moments, and the corresponding power consumption differences between adjacent target moments, the power consumption fluctuation degree of each device at each moment is obtained. The number of target moments, the absolute values ​​of the derivatives, and the power consumption differences are all positively correlated with the power consumption fluctuation degree.

3. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 1 is characterized in that: The method for obtaining the power consumption correlation degree includes: According to the similarity of power consumption fluctuations at all times between devices in the neighborhood at different times, the power consumption related time period within the neighborhood at each time between the devices is obtained; Based on the total duration of the power consumption-related periods within the neighborhood range between devices at each moment, the correlation coefficient of the power consumption fluctuation degree at all moments within different power consumption-related periods, and the duration of the power consumption-related periods, the power consumption correlation degree between devices at each moment is obtained. The total duration of the power consumption-related periods, the correlation coefficient of the power consumption fluctuation degree at all moments within the power consumption-related periods, and the duration of the power consumption-related periods are all positively correlated with the power consumption correlation degree.

4. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 3 is characterized in that: The method for obtaining the power consumption related time period includes: In the neighborhood of each moment, the correlation coefficient of the series of power consumption fluctuations between devices at all moments is obtained as the local correlation between devices at each moment. The signs of the local correlation values ​​between the devices at each moment are obtained. If the signs of consecutive adjacent moments are the same, the range of the corresponding consecutive adjacent moments is used as the power consumption correlation period.

5. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 1 is characterized in that: The method for obtaining the associated device includes: If the power consumption correlation between the devices at each moment is greater than a preset threshold, the corresponding devices are considered as associated devices.

6. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 1, characterized in that: The method for obtaining the fluctuation contribution factor includes: Based on the number of associated devices of each device at each moment and the degree of power consumption correlation between each device and other associated devices at each moment, the fluctuation contribution factor of each device at each moment is obtained. The number of associated devices and the degree of power consumption correlation are both positively correlated with the fluctuation contribution factor.

7. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 1 is characterized in that: The method for obtaining the fault attention factor includes: Obtain the failure factor of each device at each moment based on the changing trend of the number of associated devices at different moments within the neighborhood of each device at each moment; If the failure factor of a device at any moment is greater than the preset failure threshold, the corresponding device will be regarded as the faulty device at each moment; According to the failure factors and fluctuation contribution factors of different faulty equipment at each moment and the time difference between each moment and the real time, the failure attention factor at each moment is obtained. The failure factor and fluctuation contribution factor are both positively correlated with the failure attention factor, while the time difference is negatively correlated with the failure attention factor.

8. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 7 is characterized in that: The method for obtaining the fault factor includes: Obtain a quantity fitting line for fitting the quantity of associated devices at all times within the neighborhood of each device at each moment, and obtain the absolute value of the slope of the quantity fitting line as the quantity fluctuation coefficient of each device at each moment; Based on the difference fluctuation degree of the number of associated devices between different adjacent moments within the neighborhood of each device at each moment, as well as the quantity fluctuation coefficient, the failure factor of each device at each moment is obtained. The difference fluctuation degree and the first fluctuation coefficient are both positively correlated with the failure factor.

9. The method for load forecasting of the power Internet of Things based on a pre-trained large model according to claim 1, characterized in that: The method for obtaining the overall correction amount of power consumption includes: The cumulative power consumption of all devices at each moment is obtained as the overall power consumption at each moment; the product of the overall power consumption at each moment and the fault attention factor is obtained as the overall power consumption correction at each moment.

10. A power Internet of Things load forecasting system based on a pre-trained large model, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the power Internet of Things load forecasting method based on a pre-trained large model as described in any one of claims 1 to 9 are implemented.

Images