WO2020237539A1 - 电力负荷的预测方法、装置及存储介质 - Google Patents

电力负荷的预测方法、装置及存储介质 Download PDF

Info

Publication number
WO2020237539A1
WO2020237539A1 PCT/CN2019/089102 CN2019089102W WO2020237539A1 WO 2020237539 A1 WO2020237539 A1 WO 2020237539A1 CN 2019089102 W CN2019089102 W CN 2019089102W WO 2020237539 A1 WO2020237539 A1 WO 2020237539A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
time scale
load
power load
time
Prior art date
Application number
PCT/CN2019/089102
Other languages
English (en)
French (fr)
Inventor
华文韬
李晶
刘浩
王丹
李昂
Original Assignee
西门子股份公司
西门子（中国）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 西门子股份公司, 西门子（中国）有限公司 filed Critical 西门子股份公司
Priority to EP19931378.4A priority Critical patent/EP3965027A4/en
Priority to US17/614,401 priority patent/US11740603B2/en
Priority to PCT/CN2019/089102 priority patent/WO2020237539A1/zh
Priority to CN201980096604.1A priority patent/CN113853623A/zh
Publication of WO2020237539A1 publication Critical patent/WO2020237539A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
    • G05B19/042—Programme control other than numerical control, i.e. in sequence controllers or logic controllers using digital processors
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/20—Pc systems
    • G05B2219/26—Pc applications
    • G05B2219/2639—Energy management, use maximum of cheap power, keep peak load low
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Definitions

  • the invention relates to the field of energy, in particular to a method, a device, a cloud platform, a server and a storage medium for power load prediction.
  • Electric load (such as the power of equipment such as transformers, etc.) is an important part of the electric power industry, which has a great influence on the stability of grid operation. Continuous overload will cause damage to electrical equipment such as transformers. In order to ensure the normal operation of the power grid, it is necessary to monitor the electrical load in advance.
  • power load forecasting is usually based on the growth rate, and the growth rate is calculated based on user tags.
  • the tags of registered users of the power supply bureau are relatively fixed, and these tags do not reflect the latest situation of users. Therefore, the power load forecast based on the growth rate severely limits the accuracy of the forecast.
  • one aspect of the embodiments of the present invention proposes a power load forecasting method, and on the other hand, a power load forecasting device, cloud platform, server, storage medium, and computer program product are proposed to increase power Load forecast accuracy.
  • the method for predicting power load proposed in the embodiment of the present invention includes: obtaining historical power load data of a one-dimensional time series meeting a set time length, which is composed of data corresponding to each time point; and combining the one-dimensional time
  • the historical power load data of the sequence is converted into a three-dimensional matrix composed of a set time scale, the days included in each time scale, and the time points included in each day; according to the size of the time scale, the historical power of the three-dimensional matrix
  • the load data is divided into at least one operation mode; in each operation mode, the time scale is used as a unit, and the electric power of the next time scale to be predicted in the operation mode is derived based on the historical power load data in each time scale
  • the daily value band of the load data includes: obtaining historical power load data of a one-dimensional time series meeting a set time length, which is composed of data corresponding to each time point; and combining the one-dimensional time
  • the historical power load data of the sequence is converted into a three-dimensional matrix composed of
  • the time scale in each operation mode, is used as a unit, and the power load of the next time scale to be predicted in the operation mode is derived based on the historical power load data in each time scale.
  • the daily value band of the data includes: in each operation mode, perform the following operations: use the time scale as a unit to determine the load of the historical power load data in each time scale at each time point of the day Representative value to obtain a load dominant curve; for each two adjacent time scales, calculate the change value of the load dominant curve of the next time scale relative to the load dominant curve of the previous time scale to obtain a load change curve; according to all Load dominance curve and all load change curves, derive the load dominance curve of the next time scale to be predicted; determine the load dominance of the next time scale to be predicted according to the value of historical power load data in each time scale The confidence interval of the curve; based on the load dominant curve of the next time scale to be predicted and the confidence interval, the daily value band of the power load data of the next time scale to be predicted in the operation
  • the dividing the historical power load data of the three-dimensional matrix into at least one operation mode according to the size of the time scale includes: when the time scale is less than a set first threshold, dividing the The historical power load data of the three-dimensional matrix is divided into one operation mode; when the time scale is greater than or equal to the first threshold, the historical power load data of the three-dimensional matrix is clustered with days as individuals to obtain a plurality of categories , Each category corresponds to a running mode.
  • the method further includes: calculating the proportion of the number of days contained in each class in the total number of days, and discarding the classes whose proportion is less than the set second threshold, and at the same time The operation mode corresponding to the class is discarded.
  • the time scale less than the first threshold includes months or weeks; the time scale greater than the first threshold includes years.
  • the time scale is a month or a week; the historical power load data of a one-dimensional time series is converted into a set time scale, the days included in each time scale, and the time points included in each day
  • the three-dimensional matrix formed includes: converting the historical power load data of a one-dimensional time series into an initial three-dimensional matrix composed of years, days included in each year, and time points included in each day; converting the initial three-dimensional matrix into the time scale , A three-dimensional matrix composed of the days included in each time scale and the time points included in each day.
  • the external data after obtaining a load change curve, it further includes: obtaining external data for the corresponding time period for the load change curve; the external data includes weather data and/or production plan data; and calculating the external data Correlation with the load change curve; when the correlation is greater than the set third threshold, determine that the external data is related to the load change curve; when all load change curves are related to the external data of the corresponding time period When the ratio reaches the set fourth threshold, it is determined that the external data is related to the electric load data; the load dominating curve for the next time scale to be predicted based on all load dominating curves and all load change curves is: Load dominant curve, all load change curves and current external data, derive the load dominant curve of the next time scale to be predicted.
  • a power load forecasting device proposed in an embodiment of the present invention includes: a one-dimensional data acquisition module for acquiring historical power load data of a one-dimensional time series that satisfies a set time length, which consists of data corresponding to each time point Composition; a data conversion module for converting the historical power load data of the one-dimensional time series into a three-dimensional matrix composed of a set time scale, days included in each time scale, and time points included in each day; operating mode
  • the dividing module is configured to divide the historical power load data of the three-dimensional matrix into at least one operation mode according to the size of the time scale; the data prediction module is configured to use the time scale as a unit in each operation mode According to the historical power load data in each time scale, the daily value band of the power load data of the next time scale to be predicted in the operation mode is derived.
  • the data prediction module includes: a first unit, configured to determine the historical power load data of each time scale in each day by using the time scale as a unit in each operation mode The representative value of the load at a time point is used to obtain a load-dominant curve; the second unit is used to calculate the load-dominant curve of the latter time scale relative to the previous one for each two adjacent time scales in each operating mode The change value of the load dominant curve on the time scale is used to obtain a load change curve; the third unit is used to derive the next time scale to be predicted based on all load dominant curves and all load change curves in each operation mode Load-dominant curve; the fourth unit is used to determine the confidence interval of the load-dominant curve for the next time scale to be predicted according to the value of historical power load data in each time scale in each operation mode; and The fifth unit is used to obtain the power load data of the next time scale to be predicted in the operation mode based on the load dominance curve of the next time scale to be predicted and the confidence interval in
  • the operation mode dividing module divides the historical power load data of the three-dimensional matrix into an operation mode when the time scale is less than a set first threshold; when the time scale is greater than or When it is equal to the first threshold, cluster the historical power load data of the three-dimensional matrix with days as an individual to obtain a plurality of classes, and each class corresponds to an operating mode.
  • it further includes: a simplification module for calculating the proportion of the number of days contained in each class in the total number of days, and discarding the class whose proportion is less than the set second threshold, and discarding the The operating mode corresponding to the class.
  • the time scale less than the first threshold includes months or weeks; the time scale greater than the first threshold includes years.
  • the time scale is a month or a week; the data conversion module first converts the historical power load data of the one-dimensional time series into a year, a day included in each year, and a time point included in each day The initial three-dimensional matrix, and then convert the initial three-dimensional matrix into a three-dimensional matrix composed of the time scale, the days included in each time scale, and the time points included in each day.
  • the data prediction module further includes: a sixth unit for obtaining external data of a corresponding time period for each load change curve obtained by the second unit; the external data includes weather data and/or Production plan data; calculating the correlation between the external data and the load change curve; when the correlation is greater than a set third threshold, determining that the external data is related to the load change curve; and the seventh unit , Used to determine that the external data is related to the electrical load data when the proportion of all load change curves related to the external data in the corresponding time period reaches the set fourth threshold; the fifth unit is based on all load dominant curves and all load changes Curves and current external data to derive the load dominant curve for the next time scale to be predicted.
  • Another power load prediction device proposed in an embodiment of the present invention includes: at least one memory and at least one processor, wherein: the at least one memory is used to store a computer program; the at least one processor is used to call the At least one computer program stored in the memory executes the power load prediction method described in any of the above embodiments.
  • a cloud platform or server proposed in an embodiment of the present invention includes the power load forecasting device described in any of the foregoing embodiments.
  • a computer-readable storage medium provided in an embodiment of the present invention has a computer program stored thereon; it is characterized in that the computer program can be executed by a processor and implement the power load described in any of the above embodiments. method of prediction.
  • a computer program product proposed in an embodiment of the present invention which is stored on a computer-readable storage medium, includes computer program instructions, and when executed, enables at least one processor to execute the power load described in any of the foregoing embodiments Forecasting method.
  • one-dimensional time series data is converted into three-dimensional matrix data based on time scale, and then the operating modes are divided based on the three-dimensional matrix data, and each operating mode According to the historical power load data in each time scale, the power load data of the next time scale is predicted. Since this forecasting method is based on historical power load data instead of user tags, it avoids the use of one-dimensional time series data for time series analysis technology that is easy to lose features, and uses three-dimensional matrix data-based technology for prediction. It can improve the accuracy of power load forecasting.
  • the time scale determines the load dominance curve of the historical power load data in each time scale, and calculate the latter time scale for every two adjacent time scales
  • the load-dominant curve of the load-dominant curve is relative to the load-dominant curve of the previous time scale, and then the load-dominant curve of the next time scale to be predicted is derived; the value of the historical power load data in each time scale is determined.
  • the confidence interval of the load dominance curve of the next time scale to be predicted; based on the load dominance curve of the next time scale to be predicted and the confidence interval, the power of the next time scale to be predicted in the operation mode is obtained
  • the daily value band of the load data mainly uses statistical techniques, so it is easy to implement and can guarantee the accuracy of prediction.
  • the converted three-dimensional matrix data is clustered with days as individuals, so that the The days with similar characteristics are divided into one category, and the same category corresponds to the same operating mode.
  • the historical power load data in each time scale is reused to predict the power load of the next time scale. , You can further improve the accuracy of the forecast.
  • the classes whose proportions are less than the set second threshold are discarded, and the operation mode corresponding to the class is discarded at the same time, It can reduce the amount of calculation in invalid operating modes and save calculation and storage resources.
  • the three-dimensional matrix when the three-dimensional matrix is converted, by first converting the historical power load data of the one-dimensional time series into a three-dimensional matrix based on a larger time scale, the three-dimensional matrix can be made more versatile, that is, the one-dimensional data is This kind of complex conversion of three-dimensional matrix data only needs to be performed once. After that, whether it is a larger time scale or a smaller time scale, the simpler further conversion can be performed through the three-dimensional matrix based on the larger time scale without the need to re-convert. The complex conversion of one-dimensional data to three-dimensional matrix data is performed, thereby reducing the difficulty of data processing.
  • Fig. 1 is an exemplary flowchart of a method for predicting power load in an embodiment of the present invention.
  • Fig. 2 is a schematic flowchart of a method for deriving power load data of the next time scale to be predicted based on historical power load data in each time scale in an embodiment of the present invention.
  • Fig. 3 is a schematic diagram of the daily value band of the power load data of the next time scale to be predicted in an operation mode obtained in an example of the present invention.
  • Fig. 4 is an exemplary structure diagram of a power load forecasting device in an embodiment of the present invention.
  • FIG. 5 is a schematic structural diagram of the data prediction module shown in FIG. 3 in an embodiment of the present invention.
  • Fig. 6 is an exemplary structure diagram of yet another power load forecasting device in an embodiment of the present invention.
  • the power load data (such as power data, etc.) has the characteristics of time series
  • the power load data is one-dimensional time series data
  • time series analysis is only based on one-dimensional forecasts, it will lose some of the characteristics of historical data.
  • power load prediction can also be performed based on different time scales (such as year, month, week, etc.).
  • Fig. 1 is an exemplary flowchart of a method 1 for predicting power load in an embodiment of the present invention. As shown in Figure 1, the method 1 may include the following steps:
  • Step S102 Obtain historical power load data of a one-dimensional time series meeting a set time length, which is composed of data corresponding to each time point.
  • the set time length can be determined according to actual needs. For example, when the time scale is a year, the set time length can be 2 years, etc.
  • Step S104 Convert the historical power load data of the one-dimensional time series into a three-dimensional matrix consisting of a set time scale, days included in each time scale, and time points included in each day.
  • the historical power load data of the one-dimensional time series can be converted into a year, a day included in a year, and a day included
  • a set first threshold such as a year
  • the historical power load data of the one-dimensional time series can be converted into a year, a day included in a year, and a day included
  • the three-dimensional matrix of time points For example, if there are 96 time points for data collection in one day, and one year is 365 days, and the time length of historical power load data is 2 years, a three-dimensional matrix of 2, 365, and 96 points can be obtained.
  • the historical power load data of the one-dimensional time series can be converted into months, days included in each month, and time included in each day.
  • the time scale is a week
  • the historical power load data of the one-dimensional time series can be converted into a three-dimensional matrix composed of weeks, days included in each week, and time points included in each day. matrix.
  • the historical power load data of a one-dimensional time series can also be converted into years, days included in each year, and days included in each day.
  • the initial three-dimensional matrix composed of time points, and then the initial three-dimensional matrix is transformed into a three-dimensional matrix composed of time scales such as month (or week), days contained in time scales such as month (or week), and time points contained in every day .
  • Step S106 According to the size of the time scale, divide the historical power load data of the three-dimensional matrix into at least one operation mode.
  • the operating mode here can be understood as a class with similar characteristics. For example, if the daily power load data curve is drawn in units of days, the power load data of each day with similar curve lines is a class, that is, a class. Operating mode.
  • the specific implementation takes into account that when the time scale is relatively large, such as when the year is the time scale, because the data changes within a year are relatively large, if only one operating mode is divided, the whole year's data is used to predict In the next year's power load situation, the accuracy of the forecast is not high enough because the regularity is not prominent. For this reason, for the case where the time scale is relatively large, consider clustering the historical power load data of the three-dimensional matrix with days as individuals, so that days with similar characteristics can be clustered into one class, so that multiple differences can be obtained. Each category corresponds to a running mode. When the time scale is relatively small, such as month or week, you can only divide one operation mode, and use the data of the whole month or week to predict the data of the subsequent whole month or week.
  • the historical power load data of the three-dimensional matrix can be divided into an operation mode; when the time scale is greater than or equal to the first threshold
  • the setting of the first threshold can be set according to actual conditions, for example, it can be set to any reasonable value such as 50 days, 60 days, 100 days, 200 days, etc., which is not limited here.
  • the proportion of the number of days contained in each category in the total number of days can be further calculated, and the proportion of Classes whose ratios are less than the set second threshold are discarded, and the operating mode corresponding to the class is discarded at the same time. For example, for a situation of 730 days in two years, if a category includes only one day, or three, five days, or even ten and a half months, etc., the category and the corresponding operation mode can be discarded.
  • the setting of the second threshold can be set according to the actual situation, and it is not limited here.
  • Step S108 In each operation mode, using the time scale as a unit, derive the daily fetch of the power load data of the next time scale to be predicted in the operation mode according to the historical power load data in each time scale. Value band.
  • a trained neural network prediction model can be used for prediction, that is, historical power load data in each time scale is used as the input of the neural network prediction model, and the output of the neural network prediction model in the operating mode is received. The daily value band of the predicted power load data on the next time scale.
  • the following method can also be used for prediction, that is, for each operating mode, the power load data of the next time scale to be predicted can be derived from the historical power load data in each time scale as shown in Figure 2. Steps in Method 2.
  • Step S202 using the time scale as a unit, determine the load representative value of the historical power load data in each time scale at each time point of the day to obtain a load dominant curve.
  • the load representative value of the first time point can be determined according to the value of the first time point of each day in a month.
  • the value at the second time point in each day determines a representative load value at the second time point
  • a representative load value at the third time point is determined according to the value at the third time point every day in a month.
  • a load representative value at the 96th time point is determined according to the value at the 96th time point of each day in a month, thereby obtaining a load dominant curve composed of 96 load representative values.
  • Step S204 For each two adjacent time scales, calculate the change value of the load dominating curve of the next time scale relative to the load dominating curve of the previous time scale to obtain a load change curve.
  • Step S206 according to all load dominant curves and all load change curves, derive the load dominant curve of the next time scale to be predicted.
  • Step S208 Determine the confidence interval of the load dominance curve of the next time scale to be predicted according to the value of the historical power load data in each time scale.
  • Step S210 based on the load dominance curve of the next time scale to be predicted and the confidence interval, obtain the daily value band of the power load data of the next time scale to be predicted in the operation mode.
  • FIG. 3 shows a schematic diagram of the daily value band of the power load data of the next time scale to be predicted in an operation mode obtained in an example.
  • Figure 3 we take the situation of 96 time points in a day as an example. It can be seen that the curve with dots in the middle is the load-dominant curve of the next time scale to be predicted. The dots on the line are the time points, and the upper and lower The two curves correspond to the upper and lower confidence intervals.
  • step S204 it may further include: obtaining external data for the corresponding time period for the load change curve; the external data includes weather data and/or production plan data; and calculating the external data Correlation with the load change curve; when the correlation is greater than the set third threshold, determine that the external data is related to the load change curve; when all load change curves are related to the external data of the corresponding time period When the ratio reaches the set fourth threshold, it is determined that the external data is related to the electrical load data.
  • step S206 the load dominant curve of the next time scale to be predicted can be derived based on all load dominant curves, all load change curves, and current external data.
  • step S206 still only needs to derive the load dominating curve of the next time scale to be predicted based on all load dominating curves and all load change curves.
  • the setting of the third threshold and the fourth threshold can be set according to the actual situation, which is not limited here.
  • the power load prediction method in the embodiment of the present invention has been described in detail above, and the power load prediction device in the embodiment of the present invention will be described below.
  • the device in the embodiment of the present invention can be used to implement the above in the embodiment of the present invention.
  • Fig. 4 is an exemplary structure diagram of a power load forecasting device in an embodiment of the present invention. As shown in the solid line part in FIG. 4, the device 4 may include: a one-dimensional data acquisition module 410, a data conversion module 420, an operation mode division module 430, and a data prediction module 440.
  • the one-dimensional data acquisition module 410 is used to acquire historical power load data of a one-dimensional time series meeting a set time length, which is composed of data corresponding to each time point.
  • the data conversion module 420 is configured to convert the historical power load data of the one-dimensional time series into a three-dimensional matrix composed of a set time scale, days included in each time scale, and time points included in each day. For example, if the time scale is greater than or equal to a set first threshold, such as a year, the data conversion module 420 can convert the historical power load data of the one-dimensional time series into a year, a day included in a year, and a day. A three-dimensional matrix composed of the contained time points.
  • the data conversion module 420 may convert the historical power load data of the one-dimensional time series into a month, a day included in a month, and a day included A three-dimensional matrix composed of time points; for example, a week, the data conversion module 420 can convert the historical power load data of the one-dimensional time series into a three-dimensional matrix composed of weeks, days included in each week, and time points included in each day. matrix.
  • the historical power load data of a one-dimensional time series can also be converted into years, days included in each year, and days included in each day.
  • the initial three-dimensional matrix formed by the time points, and then the initial three-dimensional matrix is transformed into a time scale, such as month (or week), the day in the time scale such as the month (or week), and the time point in each day Three-dimensional matrix.
  • the operation mode division module 430 is configured to divide the historical power load data of the three-dimensional matrix into at least one operation mode according to the size of the time scale.
  • the operating mode dividing module 430 may divide the historical power load data of the three-dimensional matrix into one operating mode when the time scale is less than a set first threshold; when the time scale is greater than or When it is equal to the first threshold, cluster the historical power load data of the three-dimensional matrix with days as an individual to obtain a plurality of classes, and each class corresponds to an operating mode.
  • the device 3 may further include a compact module (not shown in Figure 4) for calculating the number of days included in each class in the total number of days The class whose proportion is less than the set second threshold is discarded, and the operation mode corresponding to the class is discarded.
  • the data prediction module 440 is configured to derive the value band of the power load data of the next time scale to be predicted based on the historical power load data in each time scale in each operation mode, using the time scale as a unit.
  • the data prediction module 440 may have multiple specific implementation forms.
  • the data prediction module 440 may have a structure as shown in the solid line part in FIG. 5.
  • the data prediction module 440 may include: a first unit 441, a second unit 442, a third unit 443, a fourth unit 444, and a fifth unit 445.
  • the first unit 441 is used to determine the load representative value of the historical power load data in each time scale at each time point of the day in each operation mode, using the time scale as a unit, to obtain a load Dominant curve.
  • the second unit 442 is used to calculate the change value of the load-dominant curve of the latter time-scale relative to the load-dominant curve of the previous time-scale for each two adjacent time scales in each operation mode, to obtain a load change curve.
  • the third unit 443 is used to derive the load dominant curve of the next time scale to be predicted based on all load dominant curves and all load change curves in each operation mode.
  • the fourth unit 444 is used for determining the confidence interval of the load dominance curve of the next time scale to be predicted according to the value of the historical power load data in each time scale in each operation mode.
  • the fifth unit 445 is used to obtain the power load data of the next time scale to be predicted in the operation mode based on the load dominant curve of the next time scale to be predicted and the confidence interval in each operation mode The daily value band.
  • the data prediction module 440 shown in FIG. 5 may further include a sixth unit 446 and a seventh unit 447 as shown in the dashed part in FIG. 4.
  • the sixth unit 446 is configured to obtain external data of the corresponding time period for each load change curve obtained by the second unit 442; the external data includes weather data and/or production plan data; and calculate the external data and The correlation of the load change curve; when the correlation is greater than a set third threshold, it is determined that the external data is related to the load change curve.
  • the seventh unit 447 is configured to determine that the external data is related to the electric load data when the proportion of all load change curves related to the external data of the corresponding time period reaches the set fourth threshold.
  • the fifth unit 445 can derive the load dominance curve for the next time scale to be predicted based on all load dominance curves, all load change curves, and current external data.
  • Fig. 6 is an exemplary structure diagram of yet another power load forecasting device in an embodiment of the present invention.
  • the device may include: at least one memory 61 and at least one processor 62.
  • some other components may also be included, such as communication ports. These components communicate via the bus.
  • At least one memory 61 is used to store a computer program.
  • the computer program can be understood as including various modules of the power load forecasting apparatus shown in FIG. 4.
  • at least one memory 61 may also store an operating system and the like.
  • Operating systems include but are not limited to: Android operating system, Symbian operating system, Windows operating system, Linux operating system, etc.
  • At least one processor 62 is configured to call a computer program stored in at least one memory 61 to execute the method for representing device operating data described in the embodiment of the present invention.
  • the processor 62 may be a CPU, a processing unit/module, an ASIC, a logic module, or a programmable gate array. It can receive and send data through the communication port.
  • the embodiment of the present invention also provides a server, or a server cluster, or a cloud platform that includes the power load prediction device shown in FIG. 2 or FIG. 3 above.
  • a hardware module may include specially designed permanent circuits or logic devices (such as dedicated processors, such as FPGAs or ASICs) to complete specific operations.
  • the hardware module may also include a programmable logic device or circuit (for example, including a general-purpose processor or other programmable processors) temporarily configured by software to perform specific operations.
  • a programmable logic device or circuit for example, including a general-purpose processor or other programmable processors
  • it can be determined according to cost and time considerations.
  • the embodiment of the present invention also provides a computer software that can be executed on a server or a server cluster or a cloud platform.
  • the computer software can be executed by a processor and implement the power load prediction described in the embodiment of the present invention. method.
  • an embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and the computer program can be executed by a processor and implement the power load prediction method described in the embodiment of the present invention.
  • a system or device equipped with a storage medium may be provided, and the software program code for realizing the function of any one of the above embodiments is stored on the storage medium, and the computer (or CPU or MPU of the system or device) ) Read and execute the program code stored in the storage medium.
  • an operating system operating on the computer can also be used to complete part or all of the actual operations through instructions based on the program code.
  • Implementations of storage media used to provide program codes include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), Magnetic tape, non-volatile memory card and ROM.
  • the program code can be downloaded from the server computer via a communication network.
  • a computer program product proposed in an embodiment of the present invention which is stored on a computer-readable storage medium, includes computer program instructions, and when executed, enables at least one processor to execute the power load described in any of the foregoing embodiments. method of prediction.
  • one-dimensional time series data is converted into three-dimensional matrix data based on time scale, and then the operating modes are divided based on the three-dimensional matrix data, and each operating mode According to the historical power load data in each time scale, the power load data of the next time scale is predicted. Since this forecasting method is based on historical power load data instead of user tags, it avoids the use of one-dimensional time series data for time series analysis technology that is easy to lose features, and uses three-dimensional matrix data-based technology for prediction. It can improve the accuracy of power load forecasting.
  • the time scale determines the load dominance curve of the historical power load data in each time scale, and calculate the latter time scale for every two adjacent time scales
  • the load-dominant curve of the load-dominant curve is relative to the load-dominant curve of the previous time scale, and then the load-dominant curve of the next time scale to be predicted is derived; the value of the historical power load data in each time scale is determined.
  • the confidence interval of the load dominance curve of the next time scale to be predicted; based on the load dominance curve of the next time scale to be predicted and the confidence interval, the power of the next time scale to be predicted in the operation mode is obtained
  • the daily value band of the load data mainly uses statistical techniques, so it is easy to implement and can guarantee the accuracy of prediction.
  • the converted three-dimensional matrix data is clustered with days as individuals, so that the The days with similar characteristics are divided into one category, and the same category corresponds to the same operating mode.
  • the historical power load data in each time scale is reused to predict the power load data of the next time scale. At this time, the accuracy of the forecast can be further improved.
  • the classes whose proportions are less than the set second threshold are discarded, and the operation mode corresponding to the class is discarded at the same time, It can reduce the amount of calculation in invalid operating modes and save calculation and storage resources.
  • the three-dimensional matrix when the three-dimensional matrix is converted, by first converting the historical power load data of the one-dimensional time series into a three-dimensional matrix based on a larger time scale, the three-dimensional matrix can be made more versatile, that is, the one-dimensional data is This kind of complex conversion of three-dimensional matrix data only needs to be performed once. After that, whether it is a larger time scale or a smaller time scale, the simpler further conversion can be performed through the three-dimensional matrix based on the larger time scale without the need to re-convert. The complex conversion of one-dimensional data to three-dimensional matrix data is performed, thereby reducing the difficulty of data processing.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Power Engineering (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种电力负荷的预测方法、装置及存储介质。其中，方法包括：获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成（S102）；将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵（S104）；根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式（S106）；在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段（S108）。该方法能够提高电力负荷的预测准确性。

Description

电力负荷的预测方法、装置及存储介质 技术领域
本发明涉及能源领域，特别是一种电力负荷的预测方法、装置、云平台、服务器及存储介质。
背景技术
电力工业是能源领域的主要基础设施，其对工业的发展和生活的质量起着重要作用。电力负荷(如变压器等设备的功率等)是电力工业中的一个重要组成部分，其对电网运行的稳定性影响很大。持续过载将导致电器设备如变压器等的损坏。为了保证电网的正常运行，有必要提前监测电力负荷。
目前，电力负荷预测通常以增长率为基础，而增长率是根据用户标签计算的。然而供电局的注册用户标签是相对固定的，这些标签并不能反映用户的最新情况，因此基于增长率的电力负荷预测严重限制了预测的准确性。
发明内容
有鉴于此，本发明实施例中一方面提出了一种电力负荷的预测方法，另一方面提出了一种电力负荷的预测装置、云平台、服务器、存储介质及计算机程序产品，用以提高电力负荷的预测准确性。
本发明实施例中提出的一种电力负荷的预测方法，包括：获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成；将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵；根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式；在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段。
在一个实施方式中，所述在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷 数据的每天的取值波段包括：在每个运行模式内，执行下述操作：以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据在一天中的各个时间点上的负荷代表值，得到一条负荷主导曲线；针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的变化值，得到一条负荷变化曲线；根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线；根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间；基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段。
在一个实施方式中，所述根据时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式包括：在所述时间尺度小于一设定的第一阈值时，将所述三维矩阵的历史电力负荷数据划分为一个运行模式；在所述时间尺度大于或等于所述第一阈值时，对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，得到复数个类，每个类对应一个运行模式。
在一个实施方式中，所述得到复数个类之后，进一步包括：计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式。
在一个实施方式中，小于第一阈值的时间尺度包括：月或周；大于第一阈值的时间尺度包括：年。
在一个实施方式中，所述时间尺度为月或周；所述将一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵包括：将一维时间序列的历史电力负荷数据转换为由年、每年包含的天以及每天包含的时间点构成的初始三维矩阵；将所述初始三维矩阵转换为由所述时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵。
在一个实施方式中，得到一条负荷变化曲线之后，进一步包括：针对所述负荷变化曲线，获取对应时间段的外部数据；所述外部数据包括天气数据和/或生产计划数据；计算所述外部数据与所述负荷变化曲线的相关性；在所述相关性大于设定的第三阈值时，确定所述外部数据与所述负荷变化曲线相关；在所有负荷变化曲线与对应时间段的外部数据相关的比例达到设定的第四阈值时，确定外部数据与电力负荷数据相关；所述根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线为：根据所有负荷主导曲线、所有负荷变化曲线以及当前的外部数据，推导出要预测的下一时间尺度的负荷主导曲线。
本发明实施例中提出的一种电力负荷的预测装置，包括：一维数据获取模块，用于获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成；数据转换模块，用于将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵；运行模式划分模块，用于根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式；数据预测模块，用于在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段。
在一个实施方式中，所述数据预测模块包括：第一单元，用于在每个运行模式内，以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据在一天中的各个时间点上的负荷代表值，得到一条负荷主导曲线；第二单元，用于在每个运行模式内，针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的变化值，得到一条负荷变化曲线；第三单元，用于在每个运行模式内，根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线；第四单元，用于在每个运行模式内，根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间；和第五单元，用于在每个运行模式内，基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段。
在一个实施方式中，所述运行模式划分模块在所述时间尺度小于一设定的第一阈值时，将所述三维矩阵的历史电力负荷数据划分为一个运行模式；在所述时间尺度大于或等于所述第一阈值时，对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，得到复数个类，每个类对应一个运行模式。
在一个实施方式中，进一步包括：精简模块，用于计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式。
在一个实施方式中，小于第一阈值的时间尺度包括：月或周；大于第一阈值的时间尺度包括：年。
在一个实施方式中，所述时间尺度为月或周；所述数据转换模块先将所述一维时间序列的历史电力负荷数据转换为由年、每年包含的天以及每天包含的时间点构成的初始三维矩阵，再将所述初始三维矩阵转换为由所述时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵。
在一个实施方式中，所述数据预测模块进一步包括：第六单元，用于针对第二单元得到的每条负荷变化曲线，获取对应时间段的外部数据；所述外部数据包括天气数据和/或生产计划数据；计算所述外部数据与所述负荷变化曲线的相关性；在所述相关性大于设定的第三阈值时，确定所述外部数据与所述负荷变化曲线相关；和第七单元，用于在所有负荷变化曲线与对应时间段的外部数据相关的比例达到设定的第四阈值时，确定外部数据与电力负荷数据相关；所述第五单元根据所有负荷主导曲线、所有负荷变化曲线以及当前的外部数据，推导出要预测的下一时间尺度的负荷主导曲线。
本发明实施例中提出的又一种电力负荷的预测装置，包括：至少一个存储器和至少一个处理器，其中：所述至少一个存储器用于存储计算机程序；所述至少一个处理器用于调用所述至少一个存储器中存储的计算机程序，执行上述任一实施方式中所述的电力负荷的预测方法。
本发明实施例中提出的一种云平台或服务器，包括上述任一实施方式所述的电力负荷的预测装置。
本发明实施例中提出的一种计算机可读存储介质，其上存储有计算机程序；其特征在于，所述计算机程序能够被一处理器执行并实现上述任一实施方式中所述的电力负荷的预测方法。
本发明实施例中提出的一种计算机程序产品，其存储在计算机可读存储介质上，包括计算机程序指令，被执行时能够使至少一处理器执行上述上述任一实施方式中所述的电力负荷的预测方法。
从上述方案中可以看出，由于本发明实施例中，将一维的时间序列数据转换为基于时间尺度的三维的矩阵数据，然后基于该三维的矩阵数据划分运行模式，并在每个运行模式内，根据每个时间尺度内的历史电力负荷数据进行下一时间尺度的电力负荷数据的预测。由于该预测方法是基于的历史电力负荷数据而非用户标签进行的预测，且避免采用容易丢失特征的基于一维时间序列数据进行时间序列分析的技术，采用基于三维矩阵数据的技术进行预测，因此可以提高电力负荷的预测准确性。
此外，通过在每个运行模式内，以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据的负荷主导曲线，并针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的负荷变化曲线，进而推导出要预测的下一时间尺度的负荷主导曲线；根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间；基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的 电力负荷数据的每天的取值波段。该过程主要采用统计学技术，因此易于实现，且能保证预测的准确性。
进一步地，由于充分考虑了较长的时间尺度可能面临的内部数据波动比较大的情况，针对时间尺度较大的情况，对转换后的三维矩阵数据以天为个体进行聚类，从而可将具有类似特征的天划分为一个类，同一个类对应同一个运行模式，此时在每个运行模式内再利用之前每个时间尺度内的历史电力负荷数据进行下一时间尺度的电力负荷的预测时，便可以进一步提高预测的准确性。
此外，在划分类之后，通过计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式，可以减少无效运行模式的计算量，节约计算及存储资源。
另外，在进行三维矩阵的转换时，通过先将一维时间序列的历史电力负荷数据转换为基于较大时间尺度的三维矩阵，可以使该三维矩阵具有较大的通用性，即一维数据到三维矩阵数据这种复杂转换只需进行一次，之后无论是较大的时间尺度，还是较小的时间尺度，都可以通过该基于较大时间尺度的三维矩阵进行较简单的进一步转换，而无需从头再进行一维数据到三维矩阵数据的复杂转换，因此降低了数据处理的难度。
最后，通过计算外部数据与负荷变化曲线的相关性，以进一步确定外部数据对负荷变化的影响，并在有一定影响时，进行下一时间尺度的数据预测时将外部数据也考虑进去，从而可以进一步提高数据预测的准确性。
附图说明
下面将通过参照附图详细描述本发明的优选实施例，使本领域的普通技术人员更清楚本发明的上述及其它特征和优点，附图中：
图1为本发明实施例中一种电力负荷的预测方法的示例性流程图。
图2为本发明实施例中一种根据每个时间尺度内的历史电力负荷数据推导出要预测的下一个时间尺度的电力负荷数据的方法流程示意图。
图3为本发明一个例子中得到的一运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段的示意图。
图4为本发明实施例中电力负荷的预测装置的示例性结构图。
图5为本发明实施例中图3所示数据预测模块的结构示意图。
图6为本发明实施例中又一种电力负荷的预测装置的示例性结构图。
其中，附图标记如下：
Figure PCTCN2019089102-appb-000001
具体实施方式
为了描述上的简洁和直观，下文通过描述若干代表性的实施方式来对本发明的方案进行阐述。实施方式中大量的细节仅用于帮助理解本发明的方案。但是很明显，本发明的技术方案实现时可以不局限于这些细节。为了避免不必要地模糊了本发明的方案，一些实施方式没有进行细致地描述，而是仅给出了框架。下文中，“包括”是指“包括但不限于”，“根据……”是指“至少根据……，但不限于仅根据……”。由于汉语的语言习惯，下文中没有特别指出一个成分的数量时，意味着该成分可以是一个也可以是多个，或可理解为至少一个。
本发明实施例中，考虑到电力负荷数据(如功率数据等)具有时间序列的特点，即电力负荷数据为一维的时间序列数据，因此严重依赖于时间序列分析中使用的技术。但时间序列分析由于只是基于一维的预测，因此会失去一些历史数据的特征。为此，本发明实施例中，经过创造性的劳动之后，考虑将一维的时间序列数据转换为三维的矩阵数据，然后基于三维的矩阵数据进行电力负荷的预测。并且，本发明实施例中，还可以基于不同的时间尺度(如年、月、周等)进行电力负荷的预测。例如根据前几年(如前两 年)的电力负荷数据预测接下来的一年的电力负荷数据；又如，根据前若干个月(如前24个月)的电力负荷数据预测接下来的一个月的电力负荷数据；再如，根据前若干周(如前52周)的电力负荷数据预测接下来的一个周的电力负荷数据。
为了使本发明的技术方案及优点更加清楚明白，以下结合附图及实施方式，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施方式仅仅用以阐述性说明本发明，并不用于限定本发明的保护范围。
图1为本发明实施例中一种电力负荷的预测方法1的示例性流程图。如图1所示，该方法1可包括如下步骤：
步骤S102，获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成。
本步骤中，设定时间长度可根据实际需要确定，例如，当时间尺度为年时，设定的时间长度可以为2年等。
步骤S104，将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵。
例如，若时间尺度大于或等于一设定的第一阈值，例如为年，则本步骤中可将所述一维时间序列的历史电力负荷数据转换为由年、每年内包含的天以及每天包含的时间点构成的三维矩阵。例如，若一天进行数据采集的时间点为96个，而一年为365天，历史电力负荷数据的时间长度为2年，则可得到一个三维方向的点数分别为2、365、96的矩阵。
又如，若时间尺度小于所述第一阈值，例如为月，则本步骤中可将所述一维时间序列的历史电力负荷数据转换为由月、每月内包含的天以及每天包含的时间点构成的三维矩阵；又如若时间尺度为周，则本步骤中可将所述一维时间序列的历史电力负荷数据转换为由周、每周内包含的天以及每天包含的时间点构成的三维矩阵。此外，针对时间尺度小于所述第一阈值例如为月或周的情况，具体实现时，也可以先将一维时间序列的历史电力负荷数据转换为由年、每年内包含的天以及每天包含的时间点构成的初始三维矩阵，然后再将初始三维矩阵转换为由时间尺度例如月(或周)、时间尺度内如每月(或周)内包含的天以及每天包含的时间点构成的三维矩阵。
步骤S106，根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式。这里的运行模式可以理解为具有类似特征的一个类，举例说明，若以天为单位绘制每天的电力负荷数据曲线时，具有相似曲线线形的各天的电力负荷数据即为一个类，也即一个运行模式。
本步骤中，具体实现时，考虑到在时间尺度比较大时，如以年为时间尺度时，由于一年内数据的变化波动比较大，因此若只划分一个运行模式，利用整年的数据去预测下一年的电力负荷情况时，由于规律性不突出，因此预测的准确性不够大。为此对于时间尺度比较大的情况，可考虑对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，这样具有近似特征的天便可以聚集到一个类中，从而可以得到复数个不同的类，每个类对应一个运行模式。而当时间尺度比较小时，如月或周等，则可以只划分一个运行模式，利用整月或整周的数据去进行后续整月或整周的数据的预测。
即本步骤中，可以在所述时间尺度小于一设定的第一阈值时，将所述三维矩阵的历史电力负荷数据划分为一个运行模式；在所述时间尺度大于或等于所述第一阈值时，对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，得到复数个类，每个类对应一个运行模式。其中，第一阈值的设置可根据实际情况设定，例如，设置为50天、60天、100天、200天等等任意合理的取值，此处不对其进行限制。
进一步地，针对划分多个运行模式的情况，为了减少对无效运行模式的计算，可在得到复数个类之后，进一步计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式。例如，针对两年共730天的情况，若一个类中仅包括一天，或三、五天，甚或十天半月等较少天数的情况，可舍弃该类以及该类所对应的运行模式。其中，第二阈值的设置可根据实际情况设定，此处不对其进行限制。
步骤S108，在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段。
本步骤中，具体实现时，可有多种实现方式。
例如，可利用一训练好的神经网络预测模型进行预测，即将各时间尺度内的历史电力负荷数据作为所述神经网络预测模型的输入，并接收所述神经网络预测模型输出的该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段。
又如，也可以采用如下的方法进行预测，即针对每个运行模式可执行如图2所示根据每个时间尺度内的历史电力负荷数据推导出要预测的下一个时间尺度的电力负荷数据的方法2中的各个步骤。
步骤S202，以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据在一天中的各个时间点上的负荷代表值，得到一条负荷主导曲线。
例如，以一天96个时间点的情况为例，假设时间尺度为月，则可根据一个月中每天 的第1个时间点的值确定出一个第1个时间点的负荷代表值，根据一个月中每天的第2个时间点的值确定出一个第2个时间点的负荷代表值，根据一个月中每天的第3个时间点的值确定出一个第3个时间点的负荷代表值，以此类推，根据一个月中每天的第96个时间点的值确定出一个第96个时间点的负荷代表值，从而得到一条由96个负荷代表值构成的一条负荷主导曲线。
步骤S204，针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的变化值，得到一条负荷变化曲线。
步骤S206，根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线。
步骤S208，根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间。
步骤S210，基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段。
图3示出了一个例子中得到的一运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段的示意图。图3中以一天96个时间点的情况为例，可见，中间的带有圆点的曲线为要预测的下一时间尺度的负荷主导曲线，线上的圆点为各个时间点，而上下的两条曲线对应的是上下的置信区间。
针对图2所示的方法，步骤S204之后，可进一步包括：针对所述负荷变化曲线，获取对应时间段的外部数据；所述外部数据包括天气数据和/或生产计划数据；计算所述外部数据与所述负荷变化曲线的相关性；在所述相关性大于设定的第三阈值时，确定所述外部数据与所述负荷变化曲线相关；在所有负荷变化曲线与对应时间段的外部数据相关的比例达到设定的第四阈值时，确定外部数据与电力负荷数据相关。相应地，步骤S206中可根据所有负荷主导曲线、所有负荷变化曲线以及当前的外部数据，推导出要预测的下一时间尺度的负荷主导曲线。否则，若确定外部数据与电力负荷数据不相关时，则步骤S206仍只需根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线。其中，第三阈值和第四阈值的设置可根据实际情况设定，此处不对其进行限制。
以上对本发明实施例中的电力负荷的预测方法进行了详细描述，下面再对本发明实施例中的电力负荷的预测装置进行描述，本发明实施例中的装置可用于实现本发明实施例中的上述方法，对于本发明装置实施例中未详细披露的内容请参见上述方法实施例中的对应描述，此处不再一一赘述。
图4为本发明实施例中电力负荷的预测装置的示例性结构图。如图4中的实线部分所示，该装置4可包括：一维数据获取模块410、数据转换模块420、运行模式划分模块430和数据预测模块440。
其中，一维数据获取模块410用于获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成。
数据转换模块420用于将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵。例如，若时间尺度大于或等于一设定的第一阈值，例如为年，则数据转换模块420可将所述一维时间序列的历史电力负荷数据转换为由年、每年内包含的天以及每天包含的时间点构成的三维矩阵。又如，若时间尺度小于所述第一阈值，例如为月，则数据转换模块420可将所述一维时间序列的历史电力负荷数据转换为由月、每月内包含的天以及每天包含的时间点构成的三维矩阵；又例如为周，则数据转换模块420可将所述一维时间序列的历史电力负荷数据转换为由周、每周内包含的天以及每天包含的时间点构成的三维矩阵。此外，针对时间尺度小于第一阈值的情况，例如为月或周的情况，具体实现时，也可以先将一维时间序列的历史电力负荷数据转换为由年、每年内包含的天以及每天包含的时间点构成的初始三维矩阵，然后再将初始三维矩阵转换为由时间尺度，例如月(或周)、时间尺度内如每月(或周)内包含的天以及每天包含的时间点构成的三维矩阵。
运行模式划分模块430用于根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式。
在一个实施方式中，运行模式划分模块430可在所述时间尺度小于一设定的第一阈值时，将所述三维矩阵的历史电力负荷数据划分为一个运行模式；在所述时间尺度大于或等于所述第一阈值时，对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，得到复数个类，每个类对应一个运行模式。在另一个实施方式中，对于需要划分多个类的情况，该装置3可进一步包括一精简模块(图4中未示出)，用于计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式。
数据预测模块440用于在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出要预测的下一个时间尺度的电力负荷数据的取值波段。
具体实现时，数据预测模块440可有多种具体实现形式。例如，在一个实施方式中，数据预测模块440可具有如图5中的实线部分所示的结构形式。如图5所示，该数据预 测模块440可包括：第一单元441、第二单元442、第三单元443、第四单元444和第五单元445。
其中，第一单元441用于在每个运行模式内，以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据在一天中的各个时间点上的负荷代表值，得到一条负荷主导曲线。
第二单元442用于在每个运行模式内，针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的变化值，得到一条负荷变化曲线。
第三单元443用于在每个运行模式内，根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线。
第四单元444用于在每个运行模式内，根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间。
第五单元445用于在每个运行模式内，基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段。
在一个实施方式中，图5所示数据预测模块440中可进一步如图4中的虚线部分所示，包括第六单元446和第七单元447。
其中，第六单元446用于针对第二单元442得到的每条负荷变化曲线，获取对应时间段的外部数据；所述外部数据包括天气数据和/或生产计划数据；计算所述外部数据与所述负荷变化曲线的相关性；在所述相关性大于设定的第三阈值时，确定所述外部数据与所述负荷变化曲线相关。
第七单元447用于在所有负荷变化曲线与对应时间段的外部数据相关的比例达到设定的第四阈值时，确定外部数据与电力负荷数据相关。
相应地，第五单元445可根据所有负荷主导曲线、所有负荷变化曲线以及当前的外部数据，推导出要预测的下一时间尺度的负荷主导曲线。
图6为本发明实施例中又一种电力负荷的预测装置的示例性结构图。如图5所示，该装置可包括：至少一个存储器61和至少一个处理器62。此外，还可以包括一些其它组件，例如通信端口等。这些组件通过总线进行通信。
其中，至少一个存储器61用于存储计算机程序。在一个实施方式中，该计算机程序可以理解为包括图4所示的电力负荷的预测装置的各个模块。此外，至少一个存储器61还可存储操作系统等。操作系统包括但不限于：Android操作系统、Symbian操作系统、 Windows操作系统、Linux操作系统等等。
至少一个处理器62用于调用至少一个存储器61中存储的计算机程序，以执行本发明实施例中所述的设备运行数据的表示方法。处理器62可以为CPU，处理单元/模块，ASIC，逻辑模块或可编程门阵列等。其可通过所述通信端口进行数据的接收和发送。
此外，本发明实施例中还提供一种包括上述图2或图3所示的电力负荷的预测装置的服务器、或服务器集群、或云平台等。
需要说明的是，上述各流程和各结构图中不是所有的步骤和模块都是必须的，可以根据实际的需要忽略某些步骤或模块。各步骤的执行顺序不是固定的，可以根据需要进行调整。各模块的划分仅仅是为了便于描述采用的功能上的划分，实际实现时，一个模块可以分由多个模块实现，多个模块的功能也可以由同一个模块实现，这些模块可以位于同一个设备中，也可以位于不同的设备中。
可以理解，上述各实施方式中的硬件模块可以以机械方式或电子方式实现。例如，一个硬件模块可以包括专门设计的永久性电路或逻辑器件(如专用处理器，如FPGA或ASIC)用于完成特定的操作。硬件模块也可以包括由软件临时配置的可编程逻辑器件或电路(如包括通用处理器或其它可编程处理器)用于执行特定操作。至于具体采用机械方式，或是采用专用的永久性电路，或是采用临时配置的电路(如由软件进行配置)来实现硬件模块，可以根据成本和时间上的考虑来决定。
另外，本发明实施例中还提供一种能够在服务器或服务器集群或云平台上执行的计算机软件，所述计算机软件能够被一处理器执行并实现本发明实施例中所述的电力负荷的预测方法。
此外，本发明实施例中还提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序能够被一处理器执行并实现本发明实施例中所述的电力负荷的预测方法。具体地，可以提供配有存储介质的系统或者装置，在该存储介质上存储着实现上述实施例中任一实施方式的功能的软件程序代码，且使该系统或者装置的计算机(或CPU或MPU)读出并执行存储在存储介质中的程序代码。此外，还可以通过基于程序代码的指令使计算机上操作的操作系统等来完成部分或者全部的实际操作。还可以将从存储介质读出的程序代码写到插入计算机内的扩展板中所设置的存储器中或者写到与计算机相连接的扩展单元中设置的存储器中，随后基于程序代码的指令使安装在扩展板或者扩展单元上的CPU等来执行部分和全部实际操作，从而实现上述实施方式中任一实施方式的功能。用于提供程序代码的存储介质实施方式包括软盘、硬盘、磁光盘、光盘(如CD-ROM、CD-R、CD-RW、DVD-ROM、DVD-RAM、DVD-RW、DVD+RW)、磁带、非易失性存 储卡和ROM。可选择地，可以由通信网络从服务器计算机上下载程序代码。
本发明实施例中提出的一种计算机程序产品，其存储在计算机可读存储介质上，包括计算机程序指令，被执行时能够使至少一处理器执行上述任一实施方式中所述的电力负荷的预测方法。
从上述方案中可以看出，由于本发明实施例中，将一维的时间序列数据转换为基于时间尺度的三维的矩阵数据，然后基于该三维的矩阵数据划分运行模式，并在每个运行模式内，根据每个时间尺度内的历史电力负荷数据进行下一时间尺度的电力负荷数据的预测。由于该预测方法是基于的历史电力负荷数据而非用户标签进行的预测，且避免采用容易丢失特征的基于一维时间序列数据进行时间序列分析的技术，采用基于三维矩阵数据的技术进行预测，因此可以提高电力负荷的预测准确性。
此外，通过在每个运行模式内，以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据的负荷主导曲线，并针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的负荷变化曲线，进而推导出要预测的下一时间尺度的负荷主导曲线；根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间；基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段。该过程主要采用统计学技术，因此易于实现，且能保证预测的准确性。
进一步地，由于充分考虑了较长的时间尺度可能面临的内部数据波动比较大的情况，针对时间尺度较大的情况，对转换后的三维矩阵数据以天为个体进行聚类，从而可将具有类似特征的天划分为一个类，同一个类对应同一个运行模式，此时在每个运行模式内再利用之前每个时间尺度内的历史电力负荷数据进行下一时间尺度的电力负荷数据的预测时，便可以进一步提高预测的准确性。
此外，在划分类之后，通过计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式，可以减少无效运行模式的计算量，节约计算及存储资源。
另外，在进行三维矩阵的转换时，通过先将一维时间序列的历史电力负荷数据转换为基于较大时间尺度的三维矩阵，可以使该三维矩阵具有较大的通用性，即一维数据到三维矩阵数据这种复杂转换只需进行一次，之后无论是较大的时间尺度，还是较小的时间尺度，都可以通过该基于较大时间尺度的三维矩阵进行较简单的进一步转换，而无需从头再进行一维数据到三维矩阵数据的复杂转换，因此降低了数据处理的难度。
最后，通过计算外部数据与负荷变化曲线的相关性，以进一步确定外部数据对负荷变化的影响，并在有一定影响时，进行下一时间尺度的数据预测时将外部数据也考虑进去，从而可以进一步提高数据预测的准确性。
以上所述仅为本发明的较佳实施例而已，并不用以限制本发明，凡在本发明的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (18)

  1. 电力负荷的预测方法，其特征在于，包括：
    获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成(S102)；
    将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵(S104)；
    根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式(S106)；
    在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段(S108)。
  2. 根据权利要求1所述的电力负荷的预测方法，其特征在于，所述在每个运行模式内，以所述时间尺度为单位，根据每个时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段(S108)包括：
    在每个运行模式内，执行下述操作：
    以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据在一天中的各个时间点上的负荷代表值，得到一条负荷主导曲线(S202)；
    针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的变化值，得到一条负荷变化曲线(S204)；
    根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线(S206)；
    根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间(S208)；
    基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段(S210)。
  3. 根据权利要求1或2所述的电力负荷的预测方法，其特征在于，所述根据时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式(S106)包括：
    在所述时间尺度小于一设定的第一阈值时，将所述三维矩阵的历史电力负荷数据划分为一个运行模式；
    在所述时间尺度大于或等于所述第一阈值时，对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，得到复数个类，每个类对应一个运行模式。
  4. 根据权利要求3所述的电力负荷的预测方法，其特征在于，所述得到复数个类之后，进一步包括：计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式。
  5. 根据权利要求3所述的电力负荷的预测方法，其特征在于，小于第一阈值的时间尺度包括：月或周；大于第一阈值的时间尺度包括：年。
  6. 根据权利要求5所述的电力负荷的预测方法，其特征在于，所述时间尺度为月或周；所述将一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵包括：
    将一维时间序列的历史电力负荷数据转换为由年、每年包含的天以及每天包含的时间点构成的初始三维矩阵；
    将所述初始三维矩阵转换为由所述时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵。
  7. 根据权利要求2所述的电力负荷的预测方法，其特征在于，得到一条负荷变化曲线之后，进一步包括：
    针对所述负荷变化曲线，获取对应时间段的外部数据；所述外部数据包括天气数据和/或生产计划数据；
    计算所述外部数据与所述负荷变化曲线的相关性；
    在所述相关性大于设定的第三阈值时，确定所述外部数据与所述负荷变化曲线相关；
    在所有负荷变化曲线与对应时间段的外部数据相关的比例达到设定的第四阈值时，确定外部数据与电力负荷数据相关；
    所述根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线为：根据所有负荷主导曲线、所有负荷变化曲线以及当前的外部数据，推导出要预测的下一时间尺度的负荷主导曲线。
  8. 电力负荷的预测装置，其特征在于，包括：
    一维数据获取模块(410)，用于获取满足设定时间长度的一维时间序列的历史电力负荷数据，其由对应各时间点的数据组成；
    数据转换模块(420)，用于将所述一维时间序列的历史电力负荷数据转换为由设定的时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵；
    运行模式划分模块(430)，用于根据所述时间尺度的大小，将所述三维矩阵的历史电力负荷数据划分为至少一个运行模式；
    数据预测模块(440)，用于在每个运行模式内，以所述时间尺度为单位，根据每个 时间尺度内的历史电力负荷数据推导出该运行模式内要预测的下一个时间尺度的电力负荷数据的每天的取值波段。
  9. 根据权利要求8所述的电力负荷的预测装置，其特征在于，所述数据预测模块(440)包括：
    第一单元(441)，用于在每个运行模式内，以所述时间尺度为单位，确定每个时间尺度内的历史电力负荷数据在一天中的各个时间点上的负荷代表值，得到一条负荷主导曲线；
    第二单元(442)，用于在每个运行模式内，针对每相邻的两个时间尺度，计算后一个时间尺度的负荷主导曲线相对于前一个时间尺度的负荷主导曲线的变化值，得到一条负荷变化曲线；
    第三单元(443)，用于在每个运行模式内，根据所有负荷主导曲线以及所有负荷变化曲线，推导出要预测的下一时间尺度的负荷主导曲线；
    第四单元(444)，用于在每个运行模式内，根据各时间尺度内的历史电力负荷数据的取值确定出所述要预测的下一时间尺度的负荷主导曲线的置信区间；和
    第五单元(445)，用于在每个运行模式内，基于所述要预测的下一时间尺度的负荷主导曲线以及所述置信区间，得到该运行模式内要预测的下一时间尺度的电力负荷数据的每天的取值波段。
  10. [根据细则91更正 22.07.2019]　
    根据权利要求8或9所述的电力负荷的预测装置，其特征在于，所述运行模式划分模块在所述时间尺度小于一设定的第一阈值时，将所述三维矩阵的历史电力负荷数据划分为一个运行模式；在所述时间尺度大于或等于所述第一阈值时，对所述三维矩阵的历史电力负荷数据以天为个体进行聚类，得到复数个类，每个类对应一个运行模式。
  11. [根据细则91更正 22.07.2019]　
    根据权利要求10所述的电力负荷的预测装置，其特征在于，进一步包括：精简模块，用于计算每个类中所包含的天数在总天数中的占比，并将占比小于设定的第二阈值的类舍弃，同时舍弃所述类对应的运行模式。
  12. [根据细则91更正 22.07.2019]　
    根据权利要求11所述的电力负荷的预测装置，其特征在于，小于第一阈值的时间尺度包括：月或周；大于第一阈值的时间尺度包括：年。
  13. [根据细则91更正 22.07.2019]　
    根据权利要求12所述的电力负荷的预测装置，其特征在于，所述时间尺度为月或周；所述数据转换模块先将所述一维时间序列的历史电力负荷数据转换为由年、每年包含的天以及每天包含的时间点构成的初始三维矩阵，再将所述初始三维矩阵转换为由所述时间尺度、每个时间尺度内包含的天以及每天包含的时间点构成的三维矩阵。
  14. 根据权利要求9所述的电力负荷的预测装置，其特征在于，所述数据预测模块 (440)进一步包括：
    第六单元(446)，用于针对第二单元得到的每条负荷变化曲线，获取对应时间段的外部数据；所述外部数据包括天气数据和/或生产计划数据；计算所述外部数据与所述负荷变化曲线的相关性；在所述相关性大于设定的第三阈值时，确定所述外部数据与所述负荷变化曲线相关；和
    第七单元(447)，用于在所有负荷变化曲线与对应时间段的外部数据相关的比例达到设定的第四阈值时，确定外部数据与电力负荷数据相关；
    所述第五单元根据所有负荷主导曲线、所有负荷变化曲线以及当前的外部数据，推导出要预测的下一时间尺度的负荷主导曲线。
  15. 电力负荷的预测装置，其特征在于，包括：至少一个存储器(61)和至少一个处理器(62)，其中：
    所述至少一个存储器(61)用于存储计算机程序；
    所述至少一个处理器(62)用于调用所述至少一个存储器(61)中存储的计算机程序，执行如权利要求1至7中任一项所述的电力负荷的预测方法。
  16. [根据细则91更正 22.07.2019]　
    一种云平台或服务器，其特征在于，包括如权利要求8至15中任一项所述的电力负荷的预测装置。
  17. 计算机可读存储介质，其上存储有计算机程序；其特征在于，所述计算机程序能够被一处理器执行并实现如权利要求1至7中任一项所述的电力负荷的预测方法。
  18. 计算机程序产品，其特征在于，存储在计算机可读存储介质上，包括计算机程序指令，被执行时能够使至少一处理器执行如权利要求1至7中任一项所述的电力负荷的预测方法。
PCT/CN2019/089102 2019-05-29 2019-05-29 电力负荷的预测方法、装置及存储介质 WO2020237539A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
EP19931378.4A EP3965027A4 (en) 2019-05-29 2019-05-29 METHOD AND APPARATUS FOR POWER LOAD PREDICTION, AND BEARING INFORMATION
US17/614,401 US11740603B2 (en) 2019-05-29 2019-05-29 Power load prediction method and apparatus, and storage medium
PCT/CN2019/089102 WO2020237539A1 (zh) 2019-05-29 2019-05-29 电力负荷的预测方法、装置及存储介质
CN201980096604.1A CN113853623A (zh) 2019-05-29 2019-05-29 电力负荷的预测方法、装置及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/089102 WO2020237539A1 (zh) 2019-05-29 2019-05-29 电力负荷的预测方法、装置及存储介质

Publications (1)

Publication Number Publication Date
WO2020237539A1 true WO2020237539A1 (zh) 2020-12-03

Family

ID=73552455

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/089102 WO2020237539A1 (zh) 2019-05-29 2019-05-29 电力负荷的预测方法、装置及存储介质

Country Status (4)

Country Link
US (1) US11740603B2 (zh)
EP (1) EP3965027A4 (zh)
CN (1) CN113853623A (zh)
WO (1) WO2020237539A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115630831A (zh) * 2022-12-06 2023-01-20 北京华联电力工程监理有限公司 一种电力需求侧管理辅助决策支持系统
CN117638950A (zh) * 2023-10-19 2024-03-01 河北思极科技有限公司 用电策略推荐方法、装置、电子设备及存储介质
CN117951555A (zh) * 2024-03-27 2024-04-30 山东德源电力科技股份有限公司 一种用于负控安装测试装置的策略制定方法
CN118316037A (zh) * 2024-06-07 2024-07-09 北京智芯微电子科技有限公司 智能台区电力负荷预测方法、装置及存储介质

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113853629A (zh) * 2019-05-29 2021-12-28 西门子股份公司 电网用户分类方法、装置和计算机可读存储介质
US11979023B2 (en) * 2020-01-28 2024-05-07 Enphase Energy, Inc. Method and apparatus for relative value generation of resources based on knapsack algorithms
CN114336651B (zh) * 2022-01-04 2023-08-01 国网四川省电力公司营销服务中心 一种基于削峰潜力的电力调度方法及装置
CN115423209B (zh) * 2022-09-27 2023-07-04 中国长江三峡集团有限公司 一种电力负荷需求预测方法、装置、电子设备及存储介质
CN115663807B (zh) * 2022-11-07 2024-09-27 国网宁夏电力有限公司经济技术研究院 基于预测负荷与分布式电源出力的坚强配电网方法
CN116257745B (zh) * 2023-05-10 2023-08-15 杭州致成电子科技有限公司 一种负荷电流极端异常数据处理方法和装置
CN116258355B (zh) * 2023-05-15 2023-08-11 国网浙江省电力有限公司永康市供电公司 适用于多点功率估计的配电台区负荷曲线分解方法及装置
CN116307944B (zh) * 2023-05-18 2023-08-04 山东欣瑞电气有限公司 基于人工智能与物联网的配电箱远程监测系统
CN116307295B (zh) * 2023-05-22 2023-08-04 南京宝能科技有限公司 应用于云平台的智慧能源数字化管理系统及方法
CN116914781B (zh) * 2023-09-12 2023-12-01 中国三峡新能源(集团)股份有限公司辽宁分公司 一种新能源快速频率响应系统及方法
CN117787652A (zh) * 2024-01-11 2024-03-29 湖北华中电力科技开发有限责任公司 园区电力需求预测方法、相关装置及存储介质
CN117977576B (zh) * 2024-03-28 2024-07-16 国网四川省电力公司电力科学研究院 一种基于多尺度历史负荷数据的台区负荷预测方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140152463A1 (en) * 2012-12-03 2014-06-05 Institute For Information Industry Power monitoring device and power analysis method
CN104036356A (zh) * 2014-06-12 2014-09-10 国家电网公司 一种利用分形算法对电网未来运行状态进行预测的方法
CN108074015A (zh) * 2017-12-25 2018-05-25 中国电力科学研究院有限公司 一种风电功率超短期预测方法及系统
CN109190856A (zh) * 2018-10-30 2019-01-11 深圳供电局有限公司 一种区分待预测日的城市电网短期负荷预测方法及系统
CN109308541A (zh) * 2018-08-02 2019-02-05 华南理工大学 一种基于场景分析的电动汽车充电负荷预测方法
CN109508830A (zh) * 2018-11-15 2019-03-22 云南电网有限责任公司 一种电动汽车时空动态负荷预测的方法

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6577962B1 (en) * 2000-09-28 2003-06-10 Silicon Energy, Inc. System and method for forecasting energy usage load
US20030101009A1 (en) * 2001-10-30 2003-05-29 Johnson Controls Technology Company Apparatus and method for determining days of the week with similar utility consumption profiles
US7216021B2 (en) * 2003-10-30 2007-05-08 Hitachi, Ltd. Method, system and computer program for managing energy consumption
US9129231B2 (en) * 2009-04-24 2015-09-08 Rockwell Automation Technologies, Inc. Real time energy consumption analysis and reporting
US9286582B2 (en) * 2009-06-22 2016-03-15 Johnson Controls Technology Company Systems and methods for detecting changes in energy usage in a building
US20120150707A1 (en) * 2010-09-20 2012-06-14 Johnson Controls Technology Company Systems and methods for providing energy efficient building equipment and services
US8738334B2 (en) * 2011-04-29 2014-05-27 International Business Machines Corporation Anomaly detection, forecasting and root cause analysis of energy consumption for a portfolio of buildings using multi-step statistical modeling
US8620632B2 (en) * 2011-06-24 2013-12-31 International Business Machines Corporation Estimating building thermal properties by integrating heat transfer inversion model with clustering and regression techniques for a portfolio of existing buildings
JP6019566B2 (ja) * 2011-11-17 2016-11-02 ソニー株式会社 電力管理装置および電力管理方法
US9568519B2 (en) * 2014-05-15 2017-02-14 International Business Machines Corporation Building energy consumption forecasting procedure using ambient temperature, enthalpy, bias corrected weather forecast and outlier corrected sensor data
US10515308B2 (en) 2015-10-30 2019-12-24 Global Design Corporation Ltd. System, method and cloud-based platform for predicting energy consumption
CN106685674B (zh) * 2015-11-05 2020-01-10 华为技术有限公司 网络事件预测以及建立网络事件预测模型的方法和装置
US20170337644A1 (en) * 2016-05-23 2017-11-23 General Electric Company Data driven invocation of realtime wind market forecasting analytics

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140152463A1 (en) * 2012-12-03 2014-06-05 Institute For Information Industry Power monitoring device and power analysis method
CN104036356A (zh) * 2014-06-12 2014-09-10 国家电网公司 一种利用分形算法对电网未来运行状态进行预测的方法
CN108074015A (zh) * 2017-12-25 2018-05-25 中国电力科学研究院有限公司 一种风电功率超短期预测方法及系统
CN109308541A (zh) * 2018-08-02 2019-02-05 华南理工大学 一种基于场景分析的电动汽车充电负荷预测方法
CN109190856A (zh) * 2018-10-30 2019-01-11 深圳供电局有限公司 一种区分待预测日的城市电网短期负荷预测方法及系统
CN109508830A (zh) * 2018-11-15 2019-03-22 云南电网有限责任公司 一种电动汽车时空动态负荷预测的方法

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115630831A (zh) * 2022-12-06 2023-01-20 北京华联电力工程监理有限公司 一种电力需求侧管理辅助决策支持系统
CN117638950A (zh) * 2023-10-19 2024-03-01 河北思极科技有限公司 用电策略推荐方法、装置、电子设备及存储介质
CN117638950B (zh) * 2023-10-19 2024-05-31 河北思极科技有限公司 用电策略推荐方法、装置、电子设备及存储介质
CN117951555A (zh) * 2024-03-27 2024-04-30 山东德源电力科技股份有限公司 一种用于负控安装测试装置的策略制定方法
CN118316037A (zh) * 2024-06-07 2024-07-09 北京智芯微电子科技有限公司 智能台区电力负荷预测方法、装置及存储介质

Also Published As

Publication number Publication date
EP3965027A4 (en) 2023-01-25
CN113853623A (zh) 2021-12-28
US11740603B2 (en) 2023-08-29
US20220214655A1 (en) 2022-07-07
EP3965027A1 (en) 2022-03-09

Similar Documents

Publication Publication Date Title
WO2020237539A1 (zh) 电力负荷的预测方法、装置及存储介质
CN108399564B (zh) 信用评分方法及装置
WO2016165378A1 (zh) 一种储能电站海量数据清洗方法及系统
CN112560468B (zh) 气象预警文本处理方法、相关装置及计算机程序产品
WO2018059302A1 (zh) 文本识别方法、装置及存储介质
CN113312578B (zh) 一种数据指标的波动归因方法、装置、设备、及介质
CN110147470B (zh) 一种跨机房数据比对系统及方法
CN102426590A (zh) 一种质量评价的方法和装置
CN111353828A (zh) 网点到店人数预测方法及装置
CN110019420A (zh) 一种数据序列预测方法及计算设备
CN113705936A (zh) 基于碳循环的能源配置方法、装置及存储介质
CN116167581A (zh) 电池需求量预估方法、装置、调度方法和计算机设备
CN112328688A (zh) 数据存储方法、装置、计算机设备及存储介质
CN116756522B (zh) 概率预报方法、装置、存储介质及电子设备
CN117458484A (zh) 高容配比光伏短期功率预测方法、装置、设备及存储介质
CN113468385A (zh) 基于边缘处理端的能源梯度确定方法、装置及存储介质
CN112100177A (zh) 数据存储方法、装置、计算机设备及存储介质
Zhou Specific versus diverse computing in media cloud
CN115130788B (zh) 基于大数据分析的用电预测系统
CN114996930A (zh) 一种建模方法、装置、电子设备及存储介质
CN110570136B (zh) 配送范围确定方法、装置、电子设备以及存储介质
CN116449456B (zh) 阵风预报输出方法、装置、计算机设备和存储介质
CN106934489B (zh) 一种面向复杂网络的时序链路预测方法
CN112491971B (zh) 一种计算集群节点调度方法、装置、设备、产品
CN115878964B (zh) 常住人口确定方法、装置、设备和存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19931378

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019931378

Country of ref document: EP

Effective date: 20211129