Power load prediction method and device, electronic equipment and computer readable storage medium

Abstract

The invention discloses a power load prediction method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring feature data related to a power load variable in a historical time period; based on weather forecast data in the feature data, a target scene type corresponding to the current load prediction scene is identified, and the target scene type is at least a daily load prediction scene type or a cold-wave weather load prediction scene type; a target power load prediction model corresponding to the target scene type is called to predict the feature data to obtain a prediction result, and the target power load prediction model at least comprises a daily power load prediction model corresponding to the daily load prediction scene type and a daily power load prediction model corresponding to the daily load prediction scene type. Or a cold-wave weather power load prediction model corresponding to the cold-wave weather load prediction scene type, wherein a prediction result is used for at least representing a power load value in a future period of time. The technical problem of low accuracy of power load prediction is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric load forecasting, and in particular, to an electric load forecasting method, device, electronic device, and computer-readable storage medium. Background Art

[0002] Currently, with the transformation of the energy structure and the improvement of environmental awareness, more and more users have switched from traditional coal heating to electric heating. However, the frequent occurrence of global climate change and extreme weather events, especially extreme weather such as cold snaps, has had a significant impact on the load demand of "coal-to-electricity" conversion. Cold snaps not only greatly increase the electricity demand but also may cause grid instability, posing a threat to the safe operation of the power system. Therefore, accurately predicting the electric load of "coal-to-electricity" users under cold snap conditions is crucial for power dispatching, supply-demand balance, and stable operation of the power grid.

[0003] In the related art, electric load forecasting mainly relies on the analysis of factors such as historical load data, meteorological data, and holiday effects, and predicts by establishing a statistical model or a machine learning model. However, when facing extreme weather such as cold snaps, this method often lacks sufficient dynamic adaptability and accuracy to accurately capture the sharp changes in electric load, resulting in a large deviation between the prediction result and the actual demand. Therefore, there is still a technical problem of low accuracy in electric load forecasting.

[0004] For the above technical problem of low accuracy in electric load forecasting, no effective solution has been proposed yet. Summary of the Invention

[0005] Embodiments of the present invention provide an electric load forecasting method, device, electronic device, and computer-readable storage medium to at least solve the technical problem of low accuracy in electric load forecasting.

[0006] According to one aspect of the embodiments of the present invention, an electric load forecasting method is provided. The method may include: obtaining feature data related to an electric load variable within a historical time period; identifying a target scenario type corresponding to the current load forecasting scenario based on the meteorological forecast data in the feature data, where the target scenario type is at least a daily load forecasting scenario type or a cold snap weather load forecasting scenario type; calling a target electric load forecasting model corresponding to the target scenario type to predict the feature data to obtain a prediction result, where the target electric load forecasting model is at least: a daily electric load forecasting model corresponding to the daily load forecasting scenario type, or a cold snap weather electric load forecasting model corresponding to the cold snap weather load forecasting scenario type, and the prediction result is used to at least represent the electric load value within a future period of time.

[0007] Optionally, the feature data at least includes: power load data within a historical time period, weather forecast data, object type data, and date data. Among them, the power load data is used to represent the power consumption at different time points within the historical time period, the weather forecast data is used to represent the weather conditions within the historical time period and the weather conditions within the predicted time period, the object type data is used to represent the attributes and types of power consumption objects, and the date data is used to represent the date information corresponding to the historical time period.

[0008] Optionally, based on the weather forecast data in the feature data, identify the target scenario type corresponding to the current load prediction scenario, including: based on the weather forecast data, determine the extreme weather index for a period of time in the future; in response to the extreme weather index exceeding the preset threshold, identify the target scenario type corresponding to the current load prediction scenario as the cold wave weather load prediction scenario type; in response to the extreme weather index not exceeding the preset threshold, identify the target scenario type corresponding to the current load prediction scenario as the daily load prediction scenario type.

[0009] Optionally, the method further includes: in response to the target scenario type being the daily load prediction scenario type, determine the target power load prediction model as the daily load prediction model corresponding to the daily load prediction scenario type; in response to the target scenario type being the cold wave weather load prediction scenario type, determine the target power load prediction model as the cold wave weather power load prediction model corresponding to the cold wave weather load prediction scenario type.

[0010] Optionally, call the target power load prediction model corresponding to the target scenario type to predict the feature data, and obtain the prediction result, including: according to the preset data cleaning rules, clean the feature data to obtain the cleaned feature data; identify the variables related to the power load variable from the cleaned feature data; perform normalization processing on the feature data corresponding to the identified variables to obtain the preprocessed feature data; call the target power load prediction model corresponding to the target scenario type to predict the preprocessed feature data to obtain the prediction result.

[0011] Optionally, the method further includes: obtain a historical feature data sample set, where the historical feature data sample set at least includes daily load data samples and cold wave weather load data samples; respectively preprocess the daily load data samples and cold wave weather load data samples in the historical feature data sample set to obtain the preprocessed daily load data samples and preprocessed cold wave weather load data samples; based on the preprocessed daily load data samples, train the gated recurrent unit model to obtain a daily power load prediction model, and based on the preprocessed cold wave weather load data samples, train the gated recurrent unit model to obtain a cold wave weather power load prediction model.

[0012] Optionally, preprocess the daily load data samples and cold snap weather load data samples in the historical feature data sample set respectively to obtain the preprocessed daily load data samples and preprocessed cold snap weather load data samples, including: cleaning the daily load data samples and cold snap weather load data samples respectively according to preset data cleaning rules to obtain the cleaned daily load data samples and cold snap weather load data samples; identifying variable samples related to the power load variable from the cleaned daily load data samples and cold snap weather load data samples respectively; and normalizing the data samples corresponding to the identified variable samples to obtain the preprocessed daily load data samples and preprocessed cold snap weather load data samples.

[0013] Optionally, train the gated recurrent unit model based on the preprocessed daily load data samples to obtain a daily power load prediction model, and train the gated recurrent unit model based on the preprocessed cold snap weather load data samples to obtain a cold snap weather power load prediction model, including: dividing the daily load prediction data samples into a first training set and a first test set according to a preset ratio, and dividing the cold snap weather load prediction data samples into a second training set and a second test set according to a preset ratio; training the gated recurrent unit model using the first training set to obtain a trained first gated recurrent unit model, and training the gated recurrent unit model using the second training set to obtain a trained second gated recurrent unit model; verifying the first prediction accuracy of the first gated recurrent unit model using the first test set, and verifying the second prediction accuracy of the second gated recurrent unit model using the second test set; determining the first gated recurrent unit model as the daily power load prediction model in response to the first prediction accuracy being greater than the prediction accuracy threshold; and determining the second gated recurrent unit model as the cold snap weather power load prediction model in response to the second prediction accuracy being greater than the prediction accuracy threshold.

[0014] Optionally, the model architecture of the daily power load prediction model is two layers, and the model architecture of the cold snap weather power load prediction model is three layers.

[0015] According to another aspect of the embodiments of the present invention, there is also provided a power load prediction device. The device may include: an acquisition unit configured to acquire feature data related to power load changes within a historical time period; an identification unit configured to identify a target scenario type corresponding to the current load prediction scenario based on the weather forecast data in the feature data, where the target scenario type is at least a daily load prediction scenario type or a cold snap weather load prediction scenario type; a call unit configured to call a target power load prediction model corresponding to the target scenario type to predict the feature data and obtain a prediction result, where the target power load prediction model is at least: a daily power load prediction model corresponding to the daily load prediction scenario type, or a cold snap weather power load prediction model corresponding to the cold snap weather load prediction scenario type, and the prediction result is used to at least represent the power load value within a future period of time.

[0016] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a memory storing an executable program; a processor configured to run the program, where when the program runs, it executes the power load prediction method in each embodiment of the present invention.

[0017] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, where the computer-readable storage medium includes a stored executable program, and when the executable program runs, it controls the device where the computer-readable storage medium is located to execute the power load prediction method in each embodiment of the present invention.

[0018] According to another aspect of the embodiments of the present invention, there is also provided a computer program product, including a computer program, where when the computer program is executed by a processor, it implements the power load prediction method in each embodiment of the present invention.

[0019] According to another aspect of the embodiments of the present invention, there is also provided a computer program product, including a non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the power load prediction method in each embodiment of the present invention.

[0020] According to another aspect of the embodiments of the present invention, there is also provided a computer program, where when the computer program is executed by a processor, it implements the power load prediction method in each embodiment of the present invention.

[0021] In an embodiment of the present invention, characteristic data related to power load variables within a historical time period is obtained; based on the meteorological forecast data in the characteristic data, the target scenario type corresponding to the current load forecasting scenario is identified, where the target scenario type is at least a daily load forecasting scenario type or a cold snap weather forecasting load forecasting scenario type; the target power load forecasting model corresponding to the target scenario type is called to perform a forecast on the characteristic data to obtain a forecast result, where the target power load forecasting model is at least: a daily power load forecasting model corresponding to the daily load forecasting scenario type, or a cold snap weather power load forecasting model corresponding to the cold snap weather load forecasting scenario type. That is to say, in the embodiment of the present invention, the load forecasting scenario is divided into a daily load forecasting scenario and a cold snap weather load forecasting scenario, and different load forecasting scenarios correspond to different load forecasting models, fully considering the impact of extreme weather factors such as cold snaps on power loads. Based on this, through scenario recognition, the power load forecasting model most suitable for the current load forecasting scenario can be selected for power load forecasting, avoiding the limitations of a single model when facing a complex and changing environment, solving the technical problem of low accuracy in power load forecasting, and achieving the technical effect of improving the accuracy of power load forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present invention, form a part of the present invention, and the schematic embodiments and descriptions thereof are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0023] Figure 1 is a flowchart of a power load forecasting method according to an embodiment of the present invention;

[0024] Figure 2 is a schematic diagram of a medium- and long-term load forecasting method for "coal-to-electricity" users considering cold snap factors according to an embodiment of the present invention;

[0025] FIG. 3(a) is a schematic diagram of an update calculation process of a gated recurrent unit model according to an embodiment of the present invention;

[0026] FIG. 3(b) is a schematic diagram of another update calculation process of a gated recurrent unit model according to an embodiment of the present invention;

[0027] Figure 4 is a schematic diagram of the structure of a power load forecasting device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solution in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used in appropriate cases can be interchanged so that the embodiments of the present invention described herein can be implemented in an order different from those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] According to an embodiment of the present invention, an embodiment of a power load forecasting method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0032] Figure 1 is a flowchart of a power load forecasting method according to an embodiment of the present invention. As Figure 1 shown, the method may include the following steps:

[0033] Step S102, obtaining feature data related to the power load variable within a historical time period.

[0034] In the technical solution provided in step S102 of the present invention above, feature data related to the power load variable within a historical time period is obtained. Among them, the feature data may be feature data related to the power load variable of "coal-to-electricity" users. For example, electricity consumption data, meteorological data, date data, etc. during the heating season. Among them, the historical time period may be a period of time adjacent to the current time period. For example, the past week, the past half month, etc. The specific duration is not limited here.

[0035] In this embodiment, the electricity consumption data during the heating season can record the power consumption within a historical time period. For example, the electricity consumption of "coal-to-electricity" users during the heating season, where the electricity consumption can exist in the form of a time series. For example, it can be collected according to cycles such as daily, weekly, and monthly. The meteorological data can be obtained through meteorological stations or meteorological service providers, including weather conditions (sunny, rainy, snowy, etc.), wind force, wind direction, etc. The date data can reflect the periodic changes in the power load. For example, the difference in electricity consumption between weekdays and holidays, and the electricity consumption patterns in different seasons.

[0036] It should be noted that the above methods and specific contents for obtaining the characteristic data are only for illustrative purposes and are not specifically limited here. As long as the characteristic data can be used for power load forecasting and enables the target power load forecasting model to predict the change law of power demand, it is within the protection scope of the embodiments of the present invention.

[0037] Optionally, obtaining the characteristic data related to the power load variable within a historical time period can capture the associations between the power load demand and various factors, provide sufficient information for the target power load forecasting model, and then perform accurate power load forecasting.

[0038] Step S104, based on the meteorological forecast data in the characteristic data, identify the target scenario type corresponding to the current load forecasting scenario.

[0039] In the technical solution provided in step S104 of the present invention above, the target scenario type can be at least a daily load forecasting scenario type or a cold wave weather load forecasting scenario type. The cold wave weather load forecasting scenario type is used to represent extreme weather such as cold waves, for example, extreme meteorological conditions such as a sharp drop in temperature, an increase in wind speed, snowfall, or freezing. Under such weather conditions, due to the sharp drop in temperature, the heating demand will increase significantly, and the power load will be significantly affected, which is particularly obvious among "coal-to-electricity" users and will lead to a sharp increase in the power load. The daily load forecasting scenario type is the load forecasting demand under non-extreme weather conditions such as cold waves. Under this scenario type, the power load is within the normal fluctuation range.

[0040] In this embodiment, after obtaining the characteristic data related to the power load variable within a historical time period, the target scenario type corresponding to the current load forecasting scenario can be identified based on the meteorological forecast data in the characteristic data, realizing the scenario self-adaptability of power load forecasting.

[0041] For example, if the current meteorological forecast data shows that the temperature will drop significantly in the next few days, accompanied by strong winds, and the meteorological department has issued a cold wave warning, then the target scenario type corresponding to the current load forecasting scenario can be identified as the cold wave weather load forecasting scenario type. On the contrary, if the meteorological forecast data shows that the weather will be mild in the next few days and there are no special meteorological events, it can be identified as the daily load forecasting scenario type.

[0042] Step S106: Invoke the target power load forecasting model corresponding to the target scenario type to forecast the feature data, and obtain the forecasting result.

[0043] In the technical solution provided in step S106 of the present invention above, the target power load forecasting model can be at least: the daily power load forecasting model corresponding to the daily load forecasting scenario type, or the cold wave weather power load forecasting model corresponding to the cold wave weather load forecasting scenario type. The forecasting result can be used to at least represent the power load value in a future period of time.

[0044] In this embodiment, power load forecasting models adapted to different scenario types are pre-trained. Among them, the daily power load forecasting model can be used to forecast the power load in a future period of time under non-extreme weather conditions, such as on ordinary weekdays or holidays. The cold wave weather power load forecasting model can be used to forecast the power load during cold wave weather because it fully considers the influence of extreme weather factors such as cold waves on power load forecasting. Based on this, after determining the target scenario type, by invoking the target power load forecasting model corresponding to the target scenario type to forecast the feature data, the change law of the power load in this scenario can be accurately captured and reflected, thereby improving the pertinence and accuracy of the forecasting.

[0045] Optionally, in daily power load forecasting, the daily power load forecasting model can analyze user behavior patterns, the difference in power load between weekdays and holidays, and seasonal electricity consumption trends, so as to provide a more practical power load forecasting. In cold wave weather load forecasting, the cold wave weather power load forecasting model can take into account the influence of extreme weather conditions (such as cold wave weather) on power load, such as the sharp increase in heating power load caused by a sudden drop in temperature, and possible equipment failures or reduced energy efficiency during cold waves, so as to obtain a more accurate forecasting result and reflect the actual power load demand under cold wave conditions.

[0046] In steps S102 to S106 of the present invention above, by collecting feature data related to power load in a historical time period, screening out meteorological forecast data related to meteorology from it, and then using the meteorological forecast data to identify different scenario types of power load forecasting, so as to call the corresponding power load forecasting models in the daily load forecasting scenario type and the cold wave weather load forecasting scenario type for power load forecasting, avoiding the limitations of a single power load forecasting model when facing a complex and changeable environment, solving the technical problem of low accuracy of power load forecasting, and achieving the technical effect of improving the accuracy of power load forecasting.

[0047] The above method of this embodiment will be further introduced below.

[0048] As an alternative embodiment, the feature data at least includes: power load data within a historical time period, weather forecast data, object type data, and date data. Among them, the power load data is used to represent the power consumption at different time points within the historical time period, the weather forecast data is used to represent the weather conditions within the historical time period and the weather conditions within the predicted time period, the object type data is used to represent the attributes and types of power consumption objects, and the date data is used to represent the date information corresponding to the historical time period.

[0049] In this embodiment, the feature data may at least include: power load data within a historical time period, weather forecast data, object type data, and date data. Among them, the power load data can be used to represent the power consumption at different time points within the historical time period. For example, the power consumption at different time points (such as every hour, daily maximum and minimum values) within the historical time period. The weather forecast data can be used to represent the weather conditions within the historical time period and the weather conditions within the predicted time period. For example, whether it is a cold snap weather. The object type data can be used to represent the attributes and types of power consumption objects. For example, user profile data. The date data is used to represent the date information corresponding to the historical time period. For example, whether it is a holiday.

[0050] Optionally, the power load data, weather forecast data, object type data, and date data in the feature data together constitute the training data of the target power load prediction model, providing a data basis for the target power load prediction model.

[0051] Optionally, the weather forecast data not only includes the weather conditions within the historical time period, such as temperature, humidity, wind speed, weather type (sunny, rainy, snowy, etc.), but also can include the weather forecast within the predicted time period, such as temperature prediction, snowfall prediction, wind power prediction, etc. The weather forecast data is an important external factor affecting the power load. Especially under extreme weather conditions, such as cold snap weather, it is particularly crucial for the prediction of heating power load. The object type data can include the specific classification of power consumption users, such as residential, commercial, and industrial users, the climate type of the area where the users are located, and the types of power equipment used by the users. Different types of users and equipment have different power demand and response patterns. For example, industrial users have a high power load demand during specific time periods on weekdays, while residential users have a sharp increase in heating power load demand during cold snap weather. The date data can provide the corresponding date information within the historical time period, including specific year, month, day, week, etc. It can reflect the power load differences between weekdays and rest days, seasonal power load changes, and power load peaks on specific dates (such as holidays, large event days).

[0052] As an alternative embodiment, in step S104, based on the weather forecast data in the feature data, identifying the target scenario type corresponding to the current load forecasting scenario includes: based on the weather forecast data, determining the extreme weather indicators for a period of time in the future; in response to the extreme weather indicators exceeding a preset threshold, identifying the target scenario type corresponding to the current load forecasting scenario as the cold snap weather load forecasting scenario type; in response to the extreme weather indicators not exceeding the preset threshold, identifying the target scenario type corresponding to the current load forecasting scenario as the daily load forecasting scenario type.

[0053] In this embodiment, during the process of identifying the target scenario type corresponding to the current load forecasting scenario based on the weather forecast data in the feature data, the extreme weather indicators for a period of time in the future can be determined based on the weather forecast data. For example, the minimum temperature, snowfall, wind speed, etc. After determining the extreme weather indicators for a period of time in the future, if the extreme weather indicators for a period of time in the future exceed the preset threshold, then the target scenario type corresponding to the current load forecasting scenario is identified as the cold snap weather load forecasting scenario type. If the extreme weather indicators for a period of time in the future do not exceed the preset threshold, then the target scenario type corresponding to the current load forecasting scenario is identified as the daily load forecasting scenario type.

[0054] Optionally, a preset threshold is set for each extreme weather indicator, and the preset threshold can be set based on the empirical values of the significant impact on the power load under historical cold snap weather conditions. For example, for the minimum temperature, the preset threshold may be set to a temperature value below zero degrees; for the wind speed, the preset threshold can be set to the strong wind threshold common under cold snap conditions; for the snowfall, the preset threshold can be set to the snowfall value that may cause an obvious impact on the power facilities. This is only for illustrative purposes and no specific limitations are made here.

[0055] Optionally, by comparing the extreme weather indicators with the preset threshold, the target scenario type corresponding to the current load forecasting scenario can be determined. For example, if the minimum temperature is lower than the preset threshold for the cold snap temperature, the wind speed is higher than the preset threshold for the cold snap wind speed, or the snowfall exceeds the preset threshold for the cold snap snowfall, then the target scenario type corresponding to the current load forecasting scenario can be automatically identified as the cold snap weather load forecasting scenario type. If the extreme weather indicators do not exceed the preset threshold, that is, the extreme weather indicators are all within the normal range, then the target scenario type corresponding to the current load forecasting scenario can be identified as the daily load forecasting scenario type. Through the above steps, the target scenario type corresponding to the current load forecasting scenario can be dynamically identified based on the weather forecast data, providing an important basis for selecting the corresponding power load forecasting model subsequently and ensuring the accuracy of the load forecasting.

[0056] As an alternative embodiment, the method further includes: in response to the target scenario type being the daily load forecasting scenario type, determining that the target power load forecasting model is the daily load forecasting model corresponding to the daily load forecasting scenario type; in response to the target scenario type being the cold snap weather load forecasting scenario type, determining that the target power load forecasting model is the cold snap weather power load forecasting model corresponding to the cold snap weather load forecasting scenario type.

[0057] In this embodiment, if the target scenario type is the daily load forecasting scenario type, then it can be determined that the target power load forecasting model is the daily load forecasting model corresponding to the daily load forecasting scenario type. If the target scenario type is the cold snap weather load forecasting scenario type, then it can be determined that the target power load forecasting model is the cold snap weather power load forecasting model corresponding to the cold snap weather load forecasting scenario type.

[0058] Optionally, if it is determined according to the meteorological forecast data in the feature data that the extreme weather index within a period of time in the future does not exceed the preset threshold, it means that there is no expected impact of cold snaps or other extreme weather events. Based on this, it can be identified that the target scenario type corresponding to the current load forecasting scenario is the daily load forecasting scenario type. In this scenario type, a daily power load forecasting model corresponding to the daily load forecasting scenario type can be selected for power load forecasting. On the contrary, if the extreme weather index within a period of time in the future exceeds the preset threshold, it can be identified that the target scenario type corresponding to the current load forecasting scenario is the cold snap weather load forecasting scenario type. In this scenario type, a cold snap weather power load forecasting model corresponding to the cold snap weather load forecasting scenario type can be selected for power load forecasting.

[0059] As an alternative embodiment, step S106, calling the target power load forecasting model corresponding to the target scenario type to predict the feature data to obtain a prediction result, includes: cleaning the feature data according to the preset data cleaning rules to obtain the cleaned feature data; identifying the variables related to the power load variable from the cleaned feature data; normalizing the feature data corresponding to the identified variables to obtain the preprocessed feature data; calling the target power load forecasting model corresponding to the target scenario type to predict the preprocessed feature data to obtain a prediction result.

[0060] In this embodiment, when calling the target power load forecasting model corresponding to the target scenario type to forecast the feature data, the feature data can be first cleaned according to the preset data cleaning rules to obtain the cleaned feature data. Then, variables highly correlated with the power load variable are identified from the cleaned feature data. Then, the feature data corresponding to the identified variables is normalized to unify different feature scales of the data into a standard range, obtaining the preprocessed feature data. Then, the target power load forecasting model corresponding to the target scenario type is called to forecast the preprocessed feature data, obtaining the forecasting result.

[0061] Optionally, cleaning the feature data according to the preset data cleaning rules may include: deleting data records containing missing values, filling missing values with the mean or median, identifying and replacing or deleting data outliers, unifying the data format, etc., to ensure that the cleaned feature data is complete and accurate. This is only an exemplary example here and does not limit the data cleaning process.

[0062] Optionally, after cleaning the feature data according to the preset data cleaning rules, meteorological variables, time variables, load variables, etc. strongly correlated with the power load variable can be identified from the cleaned feature data. Among them, meteorological variables may include: daily average temperature, maximum temperature, minimum temperature, weather, wind force, wind direction, etc., time variables may include: year, month, date, week, period, working day / holiday, etc., and load variables may include power consumption.

[0063] In this embodiment, according to the identified target scenario type, the corresponding power load forecasting model is automatically selected and called. When facing daily load forecasting or load forecasting in cold snap weather, an adapted power load forecasting model can be dynamically called for power load forecasting, avoiding the problem of poor performance of a single model in different scenarios, thus solving the technical problem of low accuracy of power load forecasting and achieving the technical effect of improving the accuracy of power load forecasting.

[0064] Next, the training process of the target power load forecasting model will be further introduced.

[0065] As an alternative embodiment, the method further includes: obtaining a historical feature data sample set, where the historical feature data sample set at least includes daily load data samples and cold snap weather load data samples; respectively preprocessing the daily load data samples and the cold snap weather load data samples in the historical feature data sample set to obtain preprocessed daily load data samples and preprocessed cold snap weather load data samples; training a gated recurrent unit model based on the preprocessed daily load data samples to obtain a daily power load prediction model, and training the gated recurrent unit model based on the preprocessed cold snap weather load data samples to obtain a cold snap weather power load prediction model.

[0066] In this embodiment, different scenario types (daily weather prediction scenario type and cold snap weather prediction scenario type) are pre-divided. Based on this, when predicting using a gated recurrent unit (GRU) model under different scenario types, a historical feature data sample set can be obtained, where the historical feature data sample set at least includes daily load data samples and cold snap weather load data samples. Among them, the daily load data samples are mainly for predicting the power load under daily weather conditions, and the cold snap weather load data samples are mainly for predicting the power load under extreme weather conditions, especially cold snap weather.

[0067] Optionally, after obtaining the historical feature data sample set, the daily load data samples and the cold snap weather load data samples in the historical feature data sample set can be respectively preprocessed to obtain preprocessed daily load data samples and preprocessed cold snap weather load data samples. Then, based on the preprocessed daily load data samples, the GRU model is trained to obtain a daily power load prediction model, and based on the preprocessed cold snap weather load data samples, the GRU model is trained to obtain a cold snap weather power load prediction model.

[0068] For example, when obtaining the historical feature data sample set, data such as the electricity consumption information collection system data, meteorological data, and date data for three heating seasons can be collected. Then, the daily load data samples and the cold snap weather load data samples are extracted from the collected feature data.

[0069] Optionally, the daily power load prediction model is applicable to predicting the power load under normal weather conditions, while the cold snap weather power load prediction model can provide a more accurate power load prediction under extreme weather conditions. The construction of different power load prediction models under the above different scenario types can improve the accuracy of power load prediction and can more effectively cope with the sharp increase in power load brought by cold snap weather among "coal-to-electricity" users.

[0070] As an alternative embodiment, the daily load data samples and cold snap weather load data samples in the historical feature data sample set are preprocessed respectively to obtain the preprocessed daily load data samples and the preprocessed cold snap weather load data samples, including: according to the preset data cleaning rules, the daily load data samples and the cold snap weather load data samples are respectively cleaned to obtain the cleaned daily load data samples and the cold snap weather load data samples; respectively from the cleaned daily load data samples and the cold snap weather load data samples, for example, meteorological variables, time variables, user type variables, load variables, etc., variable samples related to the power load variable are identified, such as strongly correlated meteorological variables, time variables, load variables; the data samples corresponding to the identified variable samples are normalized to obtain the preprocessed daily load data samples and the preprocessed cold snap weather load data samples.

[0071] In this embodiment, when preprocessing the daily load data samples and the cold snap weather load data samples in the historical feature data sample set, the string-type data can be first tagged as numerical data recognizable by the model, and then, according to the preset data cleaning rules, the daily load data samples and the cold snap weather load data samples are respectively cleaned. For example, data missing, redundant, inaccurate, non-standard, invalid, etc. are cleaned according to the rules to obtain the cleaned daily load data samples and the cold snap weather load data samples.

[0072] For example, the standardization process of the feature data can be carried out through the following formula:

[0073]

[0074] where x ij (t k ) can be used to represent the original value of the j-th index value in the i-th region at time t k , can be used to represent the value after standardization, M j = max{x ij (t k )} can be used to represent the maximum value of the j-th index, m j = min{x ij (t k )} can be used to represent the minimum value of the j-th index, and λ j can be used to represent the centered target value of the j-th index.

[0075] Optionally, after cleaning the daily load data samples and cold snap weather load data samples, feature extraction can be performed on the cleaned daily load data samples and cold snap weather load data samples. For example, variable samples related to the power load variable are identified from the cleaned daily load data samples and cold snap weather load data samples respectively. For example, meteorological variable samples, time variable samples, load variable samples, etc. This is only an exemplary example and does not limit the specific content of the variable samples.

[0076] Optionally, after identifying the variable samples related to the power load variable, the data samples corresponding to the identified variable samples can be normalized to unify different feature scales of the feature data into a standard range, thereby obtaining the preprocessed daily load data samples and the preprocessed cold snap weather load data samples.

[0077] As an optional embodiment, based on the preprocessed daily load data samples, the gated recurrent unit model is trained to obtain a daily power load prediction model, and based on the preprocessed cold snap weather load data samples, the gated recurrent unit model is trained to obtain a cold snap weather power load prediction model, including: dividing the daily load prediction data samples into a first training set and a first test set according to a preset ratio, and dividing the cold snap weather load prediction data samples into a second training set and a second test set according to a preset ratio; training the gated recurrent unit model using the first training set to obtain a trained first gated recurrent unit model, and training the gated recurrent unit model using the second training set to obtain a trained second gated recurrent unit model; verifying the first prediction accuracy of the first gated recurrent unit model using the first test set, and verifying the second prediction accuracy of the second gated recurrent unit model using the second test set; in response to the first prediction accuracy being greater than the prediction accuracy threshold, determining the first gated recurrent unit model as the daily power load prediction model; in response to the second prediction accuracy being greater than the prediction accuracy threshold, determining the second gated recurrent unit model as the cold snap weather power load prediction model.

[0078] In this embodiment, in the process of training the gated recurrent unit model based on the preprocessed daily load data samples to obtain a daily power load prediction model, and training the gated recurrent unit model based on the preprocessed cold wave weather load data samples to obtain a cold wave weather power load prediction model, the daily load prediction data samples can be first divided into a first training set and a first test set according to a preset ratio, and the cold wave weather load prediction data samples can be divided into a second training set and a second test set according to a preset ratio. The first training set can be used to train the gated recurrent unit model to obtain a trained first gated recurrent unit model, and the second training set can be used to train the gated recurrent unit model to obtain a trained second gated recurrent unit model. Then, the first test set is used to verify the first prediction accuracy of the first gated recurrent unit model, and the second test set is used to verify the second prediction accuracy of the second gated recurrent unit model.

[0079] Optionally, after obtaining the first prediction accuracy and the second prediction accuracy, it can be further determined whether the trained first gated recurrent unit model meets the standard based on the first prediction accuracy, and whether the trained second gated recurrent unit model meets the standard based on the second prediction accuracy. For example, when the first prediction accuracy is greater than the prediction accuracy threshold, the trained first gated recurrent unit model is determined as the daily power load prediction model. When the second prediction accuracy is greater than the prediction accuracy threshold, the trained second gated recurrent unit model is determined as the cold wave weather power load prediction model.

[0080] Optionally, the gated recurrent unit model can be used to capture long-term dependencies when processing time series data. In the "coal-to-electricity" load prediction, the load is usually significantly affected by factors such as weather conditions and user behavior patterns, and the above factors form long-term dependencies in the time series.

[0081] Optionally, two scenarios of "coal-to-electricity" user heating season daily load prediction and cold wave weather load prediction are divided. Key temperature differences and weather characteristic conditions of meteorological changes can be set based on the scenarios. Model training can be carried out according to the two scenarios to continuously optimize and adjust the parameters, compare the model prediction results with the actual results, and calculate the prediction accuracy.

[0082] Optionally, in the process of training the gated recurrent unit model, the parameters of the gated recurrent unit model can be continuously optimized and adjusted through the backpropagation algorithm to minimize the prediction error, and the prediction results of the gated recurrent unit model are evaluated and optimized.

[0083] As an optional embodiment, the model architecture of the daily power load prediction model is two-layer, and the model architecture of the cold wave weather power load prediction model is three-layer.

[0084] In this embodiment, in daily scenarios, according to the differences between input data and output data, the model architecture of the daily power load forecasting model can be two - layer. In the cold wave weather forecasting scenario, according to the differences in data, the model architecture of the cold wave weather power load forecasting model can be three - layer.

[0085] Optionally, in addition to the model architecture of the power load forecasting model, training data can be used to train the gated recurrent unit (GRU) model. Finally, it is confirmed that the parameter step of the GRU model in daily scenarios can be 7. According to the differences between input data and output data, a two - layer GRU model is selected. The number of neurons in the input layer can be 32, the number of neurons in the output layer can be 64, the batch size can be 16, and the number of training epochs of the Adam optimizer for adaptive matrix estimation can be 100. The GRU model in the cold wave scenario can be adjusted as follows: the batch size can be 32, and the number of training epochs of the Adam optimizer can be 300.

[0086] Optionally, the evaluation of the gated recurrent unit model can be carried out by combining data algorithms and actual load comparison. The trained gated recurrent unit model can be evaluated. The evaluation metrics include mean squared error (MSE), coefficient of determination, and the anti - normalized predicted values and true values. The prediction accuracy can be verified by comparing the prediction results with the actual load values.

[0087] In the embodiment of the present invention, by collecting feature data related to power load in a historical time period, screening out meteorological forecast data related to meteorology from it, using the meteorological forecast data to identify different scene types of power load forecasting, and then calling the corresponding power load forecasting models in the daily load forecasting scene type and the cold wave weather load forecasting scene type for power load forecasting, the limitations of a single model in the face of a complex and changeable environment are avoided, the technical problem of low accuracy of power load forecasting is solved, and the technical effect of improving the accuracy of power load forecasting is achieved.

[0088] Next, the technical solutions of the embodiments of the present invention will be illustrated by way of preferred embodiments.

[0089] Currently, with the improvement of energy conservation and efficiency throughout society and the promotion of replacing coal with electricity, the load ratio of "coal - to - electricity" in winter has been increasing year by year. Due to the frequent occurrence of global climate change and extreme weather events, especially extreme climate phenomena such as cold waves, which have had a significant impact on the load demand of "coal - to - electricity", the medium - and long - term load forecasting of "coal - to - electricity" users has become particularly complex.

[0090] In related technologies, medium- and long-term load forecasting methods mainly rely on factors such as historical load data, temperature change trends, and holiday effects, and conduct forecasting by establishing statistical models or machine learning models. However, when faced with extreme weather such as cold snaps, these related methods often struggle to accurately capture the sharp changes in load due to a lack of sufficient dynamic adaptability and precision, resulting in a large deviation between the forecasting results and the actual demand. Although a small number of studies have attempted to improve the accuracy of medium- and long-term load forecasting by constructing physical-statistical hybrid models, due to the need for detailed building information and meteorological data support, they cannot be applied to actual production.

[0091] In summary, although there are currently some studies on "coal-to-electricity" load forecasting methods, there is still much room for improvement in the field of medium- and long-term load forecasting, especially in comprehensively considering the complexity, dynamics, and regional differences of cold snap factors. In addition, the existing technologies have the following defects and deficiencies:

[0092] I. The influencing factors are not comprehensively considered. In particular, the significant cold snap factor is not fully considered, and the model is single, resulting in inaccurate forecasting results.

[0093] When conducting medium- and long-term load forecasting, the existing technologies often ignore the important meteorological factor of cold snap, do not conduct detailed scenario division, and only use the same idea for forecasting. There is a large deviation between the forecasting results and the actual load demand. Especially under extreme weather conditions, this deviation may be more significant.

[0094] Inaccurate forecasting may lead to unreasonable power resource scheduling and allocation, which may not only cause energy waste but also fail to meet the heating needs of users in cold weather, affecting the user experience.

[0095] II. The model complexity is high and the calculation efficiency is low.

[0096] Some studies use complex mathematical models for forecasting. Although the accuracy is improved to a certain extent, it has too high a dependence on data granularity and cannot be realized in application. Moreover, the complex algorithms lead to low calculation efficiency and are difficult to meet the requirements of real-time and rapid response.

[0097] High-complexity models may result in a long time-consuming forecasting process and cannot provide effective decision support in a timely manner.

[0098] III. Poor adaptability and difficulty in coping with emergencies.

[0099] The forecasting models of the existing technologies are often constructed based on historical data, and have poor model generalization ability in scenarios such as emergencies (such as extreme weather, equipment failures, etc.).

[0100] In case of emergencies, the forecasting model may not be able to adjust in time, resulting in the heating system being unable to respond in time and affecting the user heating experience.

[0101] However, the present invention proposes a medium- and long-term load forecasting method and device for "coal to electricity" users taking into account cold wave factors, fully integrating the cold wave weather in winter that has a greater impact on heating load, and incorporating sudden weather changes, increased temperature differences, equipment failures and other sudden factors during the cold wave into the forecasting model. According to the user needs of "coal to electricity" load forecasting, the constructed medium- and long-term load forecasting model automatically identifies the daily load forecasting scenarios in the heating season and the cold wave meteorological load forecasting scenarios in the heating season, automatically matches the forecasting parameters, and realizes the accurate forecasting of the "coal to electricity" load in the heating season. A set of medium- and long-term load forecasting devices for "coal to electricity" users taking into account cold wave factors developed based on the big data architecture using programming languages ​​can realize rapid data integration and processing, governance and periodic medium- and long-term load forecasting, rolling output of the daily maximum and minimum load forecast results for the next 7 days, and support the latest data for automatic verification, which is innovative, universal and has great protection value. Thereby solving the problem of large forecasting errors caused by frequent extreme weather in the heating season, complex influence of sudden factors, and high model complexity. In addition, the technical effect of improving the accuracy of power load forecasting is achieved, and the technical problem of low accuracy of power load forecasting is solved.

[0102] The method is further described below.

[0103] Figure 2 is a schematic diagram of a method for predicting medium- and long-term loads of "coal-to-electricity" users taking into account cold wave factors according to an embodiment of the present invention. Figure 2 As shown, the method comprises the following steps:

[0104] Step S201, obtaining data.

[0105] In the technical solution provided in the above step S201 of the present invention, data related to the task can be collected, including the electricity consumption information collection system, weather, date and other data for the three heating seasons of 2021.11-2022.03, 2022.11-2023.03, and 2023.11-2024.03. This is only an illustrative example.

[0106] Step S202, cleaning the acquired data.

[0107] In the technical solution provided in the above step S202 of the present invention, data such as the electricity consumption information collection system, weather, and date can be extracted, and string type data can be labeled as numerical data that can be recognized by the model. On this basis, data cleaning rules are formulated to clean data according to the rules for missing, redundant, inaccurate, irregular, invalid, etc. data. The data can be analyzed to remove seasonality and periodicity, and the data can be standardized to obtain a data set.

[0108] For example, the data can be standardized using the above formulas (1), (2), and (3).

[0109] Step S203, perform feature extraction.

[0110] In the technical solution provided in the above step S203 of the present invention, after the data is standardized, feature extraction can be further performed to extract feature data and perform correlation analysis between features. Among them, the features here mainly include: strongly correlated meteorological variables (such as average daily temperature, maximum temperature, minimum temperature, weather, wind force, wind direction), time variables (such as year, month, date, week, cycle, working day and holiday) and load variables. After the feature data is extracted, the feature data can be unified from different feature scales into a standard range to facilitate comparison or fusion of variables between different feature data, that is, data normalization, and then the processed data is used as model training data to improve the convergence speed and accuracy of the algorithm.

[0111] Step S204, performing data division.

[0112] In the technical solution provided in the above step S204 of the present invention, after the data is feature extracted and normalized, it can be divided into a training set and a test set, and then most of the data is used as a training set for training the model; the remaining data is used as a test set for evaluating the performance of the model.

[0113] Step S205, constructing a model.

[0114] In the technical solution provided in step S205 of the present invention, a GRU model can be selected to construct a load forecasting model. The GRU model can capture long-term dependencies when processing time series data. In the "coal to electricity" load forecasting, the power load is usually significantly affected by factors such as weather conditions and user behavior patterns, which form long-term dependencies in the time series.

[0115] Step S206, training the model.

[0116] In the technical solution provided in step S206 of the present invention, two scenarios can be divided: daily load forecasting during the heating season for users who switch from coal to electricity and load forecasting under cold wave weather. Temperature difference and weather characteristic conditions that are key to meteorological changes can be set based on the scenarios. Model training can be performed based on the two scenarios to achieve continuous optimization and adjustment of parameters, and the model prediction results can be compared with the actual results to calculate the accuracy.

[0117] Optionally, during the training process, the model parameters can be continuously optimized and adjusted through the back propagation algorithm to minimize the prediction error, evaluate and optimize the model results. The training set data can be used for model training, and it is finally confirmed that the parameter step size of the GRU model in daily scenarios can be 7. A two-layer GRU model is selected according to the input data and output data. The number of neurons in the input layer can be 32, the number of neurons in the output layer can be 64, the batch size can be 16, and the number of training rounds of the adaptive matrix estimation Adam optimizer can be 100. The GRU model in the cold wave scenario can be adjusted as follows: a three-layer GRU model can be selected according to the data, the batch size can be 32, and the number of training rounds of the Adam optimizer can be 300. Thus, a medium- and long-term load forecasting model for "coal-to-electricity" users taking into account cold wave factors is constructed, which automatically switches the "coal-to-electricity" load forecast based on historical loads, meteorological forecast data, and other scenarios.

[0118] Step S207: perform model evaluation.

[0119] In the technical solution provided in step S207 of the present invention, the model effectiveness evaluation can be carried out by combining data algorithm and actual load comparison. First, the trained model is evaluated using the validation set data, and the evaluation indicators include mean square error, determination coefficient, and denormalized predicted value and true value. Second, the prediction accuracy is verified by comparing the prediction result with the actual load value.

[0120] Step S208, performing model optimization.

[0121] In the technical solution provided in the above step S208 of the present invention, a GRU load prediction model with higher prediction accuracy can be obtained by adjusting model parameters and adjusting hyperparameters.

[0122] Step S209: perform load forecasting to obtain forecasting results.

[0123] In the technical solution provided in step S209 of the present invention, the constructed medium- and long-term load forecasting model for "coal-to-electricity" users taking into account the cold wave factor can be used to forecast the daily load in the next 7 days to obtain the daily maximum and minimum load values. During the forecasting process, the features can be input into the model to obtain the forecast results.

[0124] Optionally, in the cold wave scenario, the model prediction results can be applied to the load forecast from December 18, 2023 to December 24, 2023. The minimum load prediction accuracy reached 97.81% at the highest, and the average accuracy was 95.84%; the maximum load prediction accuracy reached 97.57% at the highest, and the average accuracy was 96.27%, and the model prediction results were good. In the daily scenario, the model prediction results can be applied to the load forecast from December 25, 2023 to March 8, 2024. The daily minimum load prediction accuracy reached 99.89% at the highest, and the average accuracy was 92.87%; the maximum load prediction accuracy reached 99.98% at the highest, and the average accuracy was 92.94%, and the model prediction results were good.

[0125] Step S210: Visualize the prediction results.

[0126] In the technical solution provided in step S210 of the present invention, the prediction results can be visualized in the final stage of data analysis and model prediction. The prediction curve, actual curve and accuracy curve can be drawn using the graphics library in the professional visualization tool python for analysis and comparison.

[0127] Optionally, by building digital applications, a medium- and long-term load forecast visualization device for "coal-to-electricity" users taking into account cold wave factors can be built. Business integration can be performed on data from the power system, electricity consumption information system, and meteorological environment. User profiles, electricity consumption information, and environmental data can be associated and matched. Based on the big data base, Python language programs can be developed to achieve data integration processing and load forecasting, and finally a medium- and long-term load forecast device for "coal-to-electricity" users taking into account cold wave factors is formed, which outputs the detailed results of the daily maximum and minimum load forecasts for "coal-to-electricity" users in the heating season in a rolling 7-day cycle.

[0128] FIG. 3( a ) is a schematic diagram of a gated cycle unit model update calculation process according to an embodiment of the present invention. As shown in FIG. 3( a ), x t Indicates the input at the current moment, h t-1 represents the hidden state at the previous moment, r t represents the reset gate, z t represents the update gate, represents the candidate hidden state, h t represents the final hidden state, σ can be used to represent the sigmoid activation function, and tanh can be used to represent the hyperbolic tangent function.

[0129] FIG3( b ) is a schematic diagram of another gated cycle unit model update calculation process according to an embodiment of the present invention. As shown in FIG3( b ), x t Indicates the input at the current moment, h t-1represents the hidden state of the previous moment, ⊙ represents the dot product operation, h t represents the final hidden state, r represents the reset gate (reset), z represents the update gate (update), h' represents the candidate hidden state, and W can be used to represent the weight matrix of the candidate hidden state, W r Can be used to represent the weight matrix of the reset gate, W z The weight matrix σ can be used to represent the update gate. σ can be used to represent the sigmoid activation function. tanh can be used to represent the hyperbolic tangent function. t Used to represent load forecast results, where t is used to represent time.

[0130] The following introduces the calculation formula and principle of the GRU model.

[0131] The update gate z(t) in the GRU model can be used to control how much information of the hidden state of the previous moment needs to be retained to the current moment, which can be described by the following formula:

[0132] z(t)=σ(Wz·[h(t-1),x(t)]+bz)

[0133] Among them, σ can be used to represent the sigmoid activation function, Wz can be used to represent the weight matrix of the update gate, bz can be used to represent the bias term of the update gate, and [h(t-1),x(t)] can be used to represent the concatenation of the hidden state h(t-1) at the previous moment and the input x(t) at the current moment.

[0134] The reset gate r(t) in the GRU model can be used to control how much information of the hidden state at the previous moment needs to be ignored, which can be described by the following formula:

[0135] r(t)=σ(Wr·[h(t-1),x(t)]+br)

[0136] Among them, Wr can be used to represent the weight matrix of the reset gate, and br can be used to represent the bias term of the reset gate.

[0137] Optionally, the candidate hidden state can be obtained by calculating the hidden state after the reset gate and the current input The obtained candidate hidden state can be described by the following formula:

[0138]

[0139] Among them, Wh can be used to represent the weight matrix of the candidate hidden state, and bh can be used to represent the bias term of the candidate hidden state.

[0140] Using the update gate and the candidate hidden state and the hidden state of the previous moment for calculation, the final hidden state h(t) can be obtained, which can be described by the following formula:

[0141]

[0142] In this embodiment, in load forecasting, the input of the GRU model can be historical load data, meteorological data (including cold wave information) and other relevant features. The model predicts the trend and volatility of future loads by learning the time dependency and periodic changes of these input data. Finally, by adjusting the model parameters and optimizing the model structure, a GRU load forecasting model with higher prediction accuracy can be obtained.

[0143] According to an embodiment of the present invention, a power load prediction device is also provided. Figure 4 is a schematic diagram of the structure of a power load prediction device according to an embodiment of the present invention. Figure 4 As shown, the power load prediction device 400 includes: an acquisition unit 402 , an identification unit 404 and a calling unit 406 .

[0144] The acquisition unit 402 is used to acquire characteristic data related to the change of power load in a historical time period.

[0145] The identification unit 404 is used to identify the target scenario type corresponding to the current load forecast scenario based on the meteorological forecast data in the feature data, wherein the target scenario type is at least a daily load forecast scenario type or a cold wave weather load forecast scenario type.

[0146] The calling unit 406 is used to call the target power load prediction model corresponding to the target scenario type to predict the characteristic data and obtain a prediction result, wherein the target power load prediction model is at least: a daily power load prediction model corresponding to the daily load prediction scenario type, or a cold wave weather power load prediction model corresponding to the cold wave weather load prediction scenario type, and the prediction result is used to at least represent the power load value in a future period of time.

[0147] Optionally, the identification unit 404 may include: a determination module for determining extreme weather indicators in a future period of time based on meteorological forecast data; a first identification module for identifying that the target scenario type corresponding to the current load forecast scenario is a cold wave weather load forecast scenario type in response to the extreme weather indicators exceeding a preset threshold; and a second identification module for identifying that the target scenario type corresponding to the current load forecast scenario is a daily load forecast scenario type in response to the extreme weather indicators not exceeding a preset threshold.

[0148] Optionally, the power load forecasting device is also used to: in response to the target scenario type being a daily load forecasting scenario type, determine the target power load forecasting model as a daily load forecasting model corresponding to the daily load forecasting scenario type; in response to the target scenario type being a cold wave weather load forecasting scenario type, determine the target power load forecasting model as a cold wave weather power load forecasting model corresponding to the cold wave weather load forecasting scenario type.

[0149] Optionally, the power load prediction device is also used to: call the target power load prediction model corresponding to the target scenario type to predict the feature data to obtain a prediction result, including: according to a preset data cleaning rule, performing data cleaning on the feature data to obtain cleaned feature data; identifying variables related to the power load variables from the cleaned feature data; normalizing the feature data corresponding to the identified variables to obtain pre-processed feature data; calling the target power load prediction model corresponding to the target scenario type to predict the pre-processed feature data to obtain a prediction result.

[0150] Optionally, the power load prediction device is also used to: obtain a historical characteristic data sample set, wherein the historical characteristic data sample set includes at least daily load data samples and cold wave weather load data samples; preprocess the daily load data samples and cold wave weather load data samples in the historical characteristic data sample set, respectively, to obtain preprocessed daily load data samples and preprocessed cold wave weather load data samples; train a gated cyclic unit model based on the preprocessed daily load data samples to obtain a daily power load prediction model, and train a gated cyclic unit model based on the preprocessed cold wave weather load data samples to obtain a cold wave weather power load prediction model.

[0151] Optionally, the power load forecasting device is also used to: preprocess the daily load data samples and cold wave weather load data samples in the historical characteristic data sample set, respectively, to obtain preprocessed daily load data samples and preprocessed cold wave weather load data samples, including: according to preset data cleaning rules, respectively, performing data cleaning on the daily load data samples and the cold wave weather load data samples, to obtain cleaned daily load data samples and cold wave weather load data samples; respectively, identifying variable samples related to the power load variables from the cleaned daily load data samples and cold wave weather load data samples; normalizing the data samples corresponding to the identified variable samples, to obtain preprocessed daily load data samples and preprocessed cold wave weather load data samples.

[0152] Optionally, the power load prediction device is also used to: train the gated cyclic unit model based on the preprocessed daily load data samples to obtain the daily power load prediction model, and train the gated cyclic unit model based on the preprocessed cold wave weather load data samples to obtain the cold wave weather power load prediction model, including: dividing the daily load prediction data samples into a first training set and a first test set according to a preset ratio, and dividing the cold wave weather load prediction data samples into a second training set and a second test set according to a preset ratio; training the gated cyclic unit model using the first training set to obtain the trained first gated cyclic unit model, and training the gated cyclic unit model using the second training set to obtain the trained second gated cyclic unit model; verifying the first prediction accuracy of the first gated cyclic unit model using the first test set, and verifying the second prediction accuracy of the second gated cyclic unit model using the second test set; in response to the first prediction accuracy being greater than the prediction accuracy threshold, determining the first gated cyclic unit model as the daily power load prediction model; in response to the second prediction accuracy being greater than the prediction accuracy threshold, determining the second gated cyclic unit model as the cold wave weather power load prediction model.

[0153] An embodiment of the present invention further provides an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the power load prediction method in each embodiment of the present invention is executed when the program is running.

[0154] An embodiment of the present invention further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program runs, the device where the computer-readable storage medium is located is controlled to execute the power load forecasting method in each embodiment of the present invention.

[0155] An embodiment of the present invention further provides a computer program product, including a computer program, which implements the power load prediction method in each embodiment of the present invention when executed by a processor.

[0156] An embodiment of the present invention further provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the power load forecasting method in each embodiment of the present invention is implemented.

[0157] An embodiment of the present invention further provides a computer program, which, when executed by a processor, implements the power load prediction method in each of the above-mentioned embodiments of the present invention.

[0158] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0159] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0160] In the several embodiments provided by the present invention, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units can be a logical function division, and there can be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0161] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0162] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0163] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0164] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting power load, characterized in that: include: Obtain characteristic data related to power load variables over a historical period; Based on the meteorological forecast data in the characteristic data, identifying the target scenario type corresponding to the current load forecast scenario, wherein the target scenario type is at least a daily load forecast scenario type or a cold wave weather load forecast scenario type; The target power load prediction model corresponding to the target scenario type is called to predict the characteristic data to obtain a prediction result, wherein the target power load prediction model is at least: a daily power load prediction model corresponding to the daily load prediction scenario type, or a cold wave weather power load prediction model corresponding to the cold wave weather load prediction scenario type, and the prediction result is used to at least represent the power load value in a future period of time.

2. The method according to claim 1, characterized in that The characteristic data at least includes: power load data within the historical time period, the weather forecast data, object type data, and date data, wherein the power load data is used to indicate the power consumption at different time points within the historical time period, the weather forecast data is used to indicate the weather conditions within the historical time period and the weather conditions within the predicted time period, the object type data is used to indicate the attributes and types of power-consuming objects, and the date data is used to indicate the date information corresponding to the historical time period.

3. The method according to claim 1, characterized in that Based on the weather forecast data in the characteristic data, identifying the target scenario type corresponding to the current load forecast scenario includes: Based on the weather forecast data, determine extreme weather indicators within a period of time in the future; In response to the extreme weather index exceeding a preset threshold, identifying that the target scenario type corresponding to the current load forecast scenario is the cold wave weather load forecast scenario type; In response to the extreme weather index not exceeding the preset threshold, the target scenario type corresponding to the current load forecast scenario is identified as the daily load forecast scenario type.

4. The method according to claim 3, characterized in that The method further comprises: In response to the target scenario type being the daily load forecasting scenario type, determining the target power load forecasting model to be the daily load forecasting model corresponding to the daily load forecasting scenario type; In response to the target scenario type being the cold wave weather load forecasting scenario type, the target power load forecasting model is determined to be the cold wave weather power load forecasting model corresponding to the cold wave weather load forecasting scenario type.

5. The method according to claim 1, characterized in that The target power load prediction model corresponding to the target scenario type is called to predict the characteristic data to obtain a prediction result, including: According to a preset data cleaning rule, the characteristic data is cleaned to obtain the cleaned characteristic data; identifying variables related to power load variables from the cleaned feature data; Normalizing the feature data corresponding to the identified variable to obtain the preprocessed feature data; The target power load prediction model corresponding to the target scenario type is called to predict the preprocessed feature data to obtain the prediction result.

6. The method according to claim 1, characterized in that The method further comprises: Acquire a historical characteristic data sample set, wherein the historical characteristic data sample set includes at least daily load data samples and the cold wave weather load data samples; Preprocessing the daily load data samples and the cold wave weather load data samples in the historical characteristic data sample set respectively to obtain the preprocessed daily load data samples and the preprocessed cold wave weather load data samples; Based on the preprocessed daily load data samples, the gated cyclic unit model is trained to obtain the daily power load prediction model, and based on the preprocessed cold wave weather load data samples, the gated cyclic unit model is trained to obtain the cold wave weather power load prediction model.

7. The method according to claim 1, characterized in that Preprocessing the daily load data samples and the cold wave weather load data samples in the historical characteristic data sample set respectively to obtain the preprocessed daily load data samples and the preprocessed cold wave weather load data samples, including: According to the preset data cleaning rule, the daily load data samples and the cold wave weather load data samples are respectively cleaned to obtain the cleaned daily load data samples and the cold wave weather load data samples; identifying variable samples related to power load variables from the cleaned daily load data samples and the cold wave weather load data samples respectively; The data samples corresponding to the identified variable samples are normalized to obtain the pre-processed daily load data samples and the pre-processed cold wave weather load data samples.

8. The method according to claim 6, characterized in that Based on the preprocessed daily load data samples, the gated cyclic unit model is trained to obtain the daily power load forecasting model, and based on the preprocessed cold wave weather load data samples, the gated cyclic unit model is trained to obtain the cold wave weather power load forecasting model, including: Dividing the daily load forecast data samples into a first training set and a first test set according to a preset ratio, and dividing the cold wave weather load forecast data samples into a second training set and a second test set according to the preset ratio; Using the first training set to train the gated recurrent unit model to obtain a trained first gated recurrent unit model, and using the second training set to train the gated recurrent unit model to obtain a trained second gated recurrent unit model; Verifying a first prediction accuracy of the first gated recurrent unit model using the first test set, and verifying a second prediction accuracy of the second gated recurrent unit model using the second test set; In response to the first prediction accuracy being greater than a prediction accuracy threshold, determining the first gated cycle unit model as the daily power load prediction model; In response to the second prediction accuracy being greater than the prediction accuracy threshold, the second gated cycle unit model is determined as the cold wave weather power load prediction model.

9. The method according to claim 8, characterized in that The model architecture of the daily power load forecasting model is two-layered, and the model architecture of the cold wave weather power load forecasting model is three-layered.

10. A power load forecasting device, characterized in that: include: An acquisition unit, used to acquire characteristic data related to power load changes in a historical time period; an identification unit, configured to identify a target scenario type corresponding to a current load forecast scenario based on the meteorological forecast data in the characteristic data, wherein the target scenario type is at least a daily load forecast scenario type or a cold wave weather load forecast scenario type; A calling unit is used to call the target power load prediction model corresponding to the target scenario type to predict the characteristic data and obtain a prediction result, wherein the target power load prediction model is at least: a daily power load prediction model corresponding to the daily load prediction scenario type, or a cold wave weather power load prediction model corresponding to the cold wave weather load prediction scenario type, and the prediction result is used to at least represent the power load value in a future period of time.

11. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 9 when running.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 9.

13. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images