Whole-province electric power prediction method based on fusion of multiple meteorological sources and multiple models

Abstract

The invention belongs to the technical field of power prediction. The invention provides a whole-province electric power prediction method based on multi-meteorological-source and multi-model fusion, and the method comprises the steps: obtaining meteorological data of a plurality of future time periods at different time scales from a plurality of meteorological sources, and carrying out the preprocessing of the meteorological data; obtaining enterprise power consumption data and resident power consumption data of the target province; dividing the plurality of prediction models into model groups, testing each model group by using real power load power and meteorological data of the whole province to obtain an evaluation score of each model group, screening based on the evaluation scores to obtain a target model group, and integrating the prediction models in the target model group to obtain an integrated model; and inputting the preprocessed meteorological data, enterprise power consumption data and resident power consumption data into the integrated model, and outputting the predicted power load power of the whole province by the integrated model. According to the method, the power load power of the whole province is accurately predicted by comprehensively adopting a multi-meteorological-source and multi-model fusion technology.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power forecasting, and in particular to a province-wide electric power forecasting method based on the fusion of multiple meteorological sources and multiple models. Background Art

[0002] Against the backdrop of today's booming power industry, accurate power forecasting plays a vital role in ensuring the stable operation of power systems, optimizing power resource allocation, and improving the economic benefits of power companies. Traditional power forecasting methods often rely on single meteorological source data, such as EC, GFS, or CMC, and a single forecasting model. However, single meteorological source data has limitations and cannot fully and accurately reflect complex and changing meteorological conditions, which in turn affects the accuracy of power forecasts. At the same time, existing power forecasts are all based on a single forecasting model. However, due to its own algorithmic characteristics and limited scope of application, single forecasting models are unable to fully explore the potential patterns and complex relationships in the data, resulting in large deviations in the forecast results.

[0003] Therefore, how to further improve the accuracy of electric power forecasting is a technical problem that needs to be solved urgently. Summary of the Invention

[0004] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a province-wide electric power prediction method, electronic equipment, computer storage medium and computer program product based on the fusion of multiple meteorological sources and multiple models.

[0005] The present invention provides a province-wide electric power forecasting method based on the fusion of multiple meteorological sources and multiple models, comprising the following steps: Acquire meteorological data for multiple future time periods at different time scales from multiple meteorological sources and preprocess the meteorological data; and obtain enterprise electricity consumption data and residential electricity consumption data for the target province, the enterprise electricity consumption data including first long-term electricity consumption historical data and first recent electricity consumption historical data, and the residential electricity consumption data including second long-term electricity consumption historical data and second recent electricity consumption historical data; Divide several prediction models into model groups, test each model group using real provincial power load power and meteorological data, obtain an evaluation score for each model group, screen a target model group based on the evaluation score, and integrate the prediction models in the target model group to obtain an integrated model; The pre-processed meteorological data, the enterprise-type electricity consumption data and the residential-type electricity consumption data are input into the integrated model, and the integrated model outputs its predicted power load of the entire province.

[0006] In some embodiments, obtaining meteorological data for multiple future time periods at different time scales from multiple meteorological sources includes: Obtaining the long-term historical weather forecast data for the province from each weather source, calculating the time difference between a certain future period and the current moment, and obtaining the weather forecast accuracy of each weather source corresponding to the time difference based on the long-term historical weather forecast data; Determine several meteorological sources whose meteorological forecast accuracy is higher than an accuracy threshold as target meteorological sources corresponding to the future period, and predict meteorological data corresponding to the future period based on the meteorological forecast recent historical data of each target meteorological source; Repeat the above steps to obtain meteorological data corresponding to each of the future time periods at different time scales.

[0007] In some embodiments, obtaining enterprise electricity consumption data and residential electricity consumption data of the target province includes: Calculating a deviation between meteorological data of a certain future period and average meteorological data of previous years, and determining a first coefficient and a second coefficient based on the deviation; wherein the first coefficient is less than the second coefficient; Obtaining enterprise electricity consumption data and residential electricity consumption data for a target province, adjusting the first recent historical electricity consumption data in the enterprise electricity consumption data using the first coefficient to obtain third recent historical electricity consumption data; and adjusting the second recent historical electricity consumption data in the residential electricity consumption data using the second coefficient to obtain fourth recent historical electricity consumption data; Finally, the enterprise electricity consumption data includes the first long-term electricity consumption data and the third recent electricity consumption data, and the residential electricity consumption data includes the second long-term electricity consumption data and the fourth recent electricity consumption data.

[0008] In some embodiments, the grouping of the plurality of prediction models into model groups comprises: Determining a first quantity of target meteorological sources that predict meteorological data for a certain future time period, and obtaining a second quantity based on the first quantity of target meteorological sources; At least the second number of prediction models are screened and divided into the same group to form the model group.

[0009] In some embodiments, each model group is tested using real provincial power load power and meteorological data to obtain an evaluation score for each model group. A target model group is obtained based on the evaluation score, and each prediction model in the target model group is integrated to obtain an integrated model, including: Using real provincial electric power load power and meteorological data to test a certain model group, and obtaining a first evaluation score of the model group; Obtaining a third coefficient based on the deviation value matching, and using the third coefficient to fine-tune the first evaluation score to a second evaluation score; Repeat the above steps to obtain a second evaluation score for each model group; The model group with the largest or smallest evaluation score is determined as the target model group, and the prediction models in the target model group are integrated to obtain the integrated model.

[0010] In some embodiments, integrating the prediction models in the target model group to obtain the integrated model includes: Any one of Stacking, Boosting and Bagging is used to integrate the prediction models in the target model group to obtain the integrated model.

[0011] In some embodiments, the integrated model outputs its predicted provincial electric load power, including: The integrated model outputs its predicted provincial electric load power and its confidence level.

[0012] The present invention also provides an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute any of the methods described above.

[0013] The present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is run by a processor, the method described in any of the above items is executed.

[0014] The present invention also provides a computer program product, which includes a computer program stored in a computer storage medium, and when the computer program is executed by a processor of an electronic device, implements the method as described in any of the above items.

[0015] The beneficial effects of this invention lie in its comprehensive use of multiple meteorological sources and multi-model fusion technology to predict provincial power load power. The use of multiple meteorological sources significantly improves the availability and accuracy of meteorological data, while multi-model fusion effectively reduces the bias caused by the algorithmic characteristics and applicability limitations of a single model, significantly improving prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure. Figure 1 It is a flow chart of a method for predicting province-wide electric power based on the fusion of multiple meteorological sources and multiple models disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0018] like Figure 1 As shown, the embodiment of the present invention discloses a province-wide electric power prediction method based on the fusion of multiple meteorological sources and multiple models, comprising the following steps: S10, obtaining meteorological data of multiple future time periods at different time scales from multiple meteorological sources, and preprocessing the meteorological data; and obtaining enterprise electricity consumption data and residential electricity consumption data of the target province, wherein the enterprise electricity consumption data includes first long-term electricity consumption data and first recent electricity consumption data, and the residential electricity consumption data includes second long-term electricity consumption data and second recent electricity consumption data.

[0019] Meteorological data for multiple future time periods at different time scales (e.g., short-term 1-3 days, medium-term 4-10 days, and long-term 10 days or longer) is obtained from multiple meteorological sources, such as the EC, GFS, and CMC mentioned in the technical background section, as well as satellite remote sensing and ground-based meteorological observation stations. This meteorological data includes information such as temperature, air pressure, humidity, and precipitation. Because data from different meteorological sources may vary in format, accuracy, and timestamps, preprocessing is required. This preprocessing includes, but is not limited to, data cleaning to remove outliers and erroneous data; data standardization to unify data of varying magnitudes to the same scale for subsequent analysis; and time synchronization to ensure temporal consistency between data from different sources. At the same time, the historical data of electricity consumption of enterprises and residents in the province are also obtained to predict the power load power of the province in the future period. Electricity consumption data from businesses in target provinces is collected. Long-term historical data (e.g., data from the past few years) can reflect long-term patterns and trends in electricity consumption, such as the patterns of seasonal production. Recent historical data (e.g., data from the past few months or weeks) can reflect the immediate impact of recent changes in business production activities on electricity consumption, such as increased electricity consumption caused by temporary equipment additions to expand production. Residential electricity consumption data is also collected province-wide. The second type of long-term historical data (residential electricity consumption data) can help analyze seasonal and cyclical characteristics of residential electricity use, such as changes in electricity consumption caused by cooling in the summer and heating in the winter. The second type of recent historical data (residential electricity consumption data) can capture the impact of recent changes in residents' living habits and sudden weather changes on electricity consumption, such as the increased frequency of air conditioning use during consecutive high-temperature weather periods.

[0020] S20, divide several prediction models into model groups, use the real province-wide power load power and meteorological data to test each model group, obtain an evaluation score for each model group, screen and obtain a target model group based on the evaluation score, integrate the prediction models in the target model group to obtain an integrated model.

[0021] The present invention adopts a multi-model prediction technology, that is, a variety of models that can be used for power load power prediction, such as neural network models, time series models, regression models, etc., are divided into different model groups. Each model group is tested using real power load power data for the entire province and corresponding meteorological data. During the test, the model group predicts the power load power based on the input data, and then compares and scores the predicted results with the actual power load power. The evaluation score can be calculated based on a variety of indicators, such as the root mean square error (RMSE), which measures the average magnitude of the deviation between the predicted value and the true value; the mean absolute error (MAE), which reflects the average absolute value of the prediction error, etc. The evaluation score of each model group is calculated comprehensively through these indicators. It can be understood that in order to ensure the diversity of the integrated model, the types of prediction models in the model group should be as close as possible to the number of prediction models, that is, the number of prediction models of the same type should be reduced.

[0022] Based on the evaluation scores, the model group with the best prediction results is selected as the target model group. The prediction models in the target model group are integrated to obtain an integrated model with better prediction performance.

[0023] S30, inputting the pre-processed meteorological data, the enterprise-type electricity consumption data and the residential-type electricity consumption data into the integrated model, and the integrated model outputs its predicted power load power for the entire province.

[0024] The preprocessed meteorological data, enterprise- and residential-specific electricity consumption data obtained above are fed into the integrated model as input. The integrated model analyzes and processes the input data. Based on its ability to analyze how meteorological factors influence enterprise and residential electricity consumption, as well as the interrelationships between enterprise and residential electricity consumption patterns, the model predicts and outputs the province's power load. This forecast, which comprehensively considers data from multiple meteorological sources and various electricity consumption data characteristics, offers greater accuracy and reliability than a single model or meteorological source, providing strong support for power system scheduling and planning.

[0025] This method uses a combination of multiple meteorological sources and multi-model fusion technology to predict the province's power load. The use of multiple meteorological sources significantly improves the availability and accuracy of meteorological data, while multi-model fusion effectively reduces the bias caused by the algorithmic characteristics and applicability limitations of a single model, significantly improving prediction accuracy.

[0026] In some embodiments, obtaining meteorological data for multiple future time periods at different time scales from multiple meteorological sources includes: Obtaining the long-term historical weather forecast data for the province from each weather source, calculating the time difference between a certain future period and the current moment, and obtaining the weather forecast accuracy of each weather source corresponding to the time difference based on the long-term historical weather forecast data; Determine several meteorological sources whose meteorological forecast accuracy is higher than an accuracy threshold as target meteorological sources corresponding to the future period, and predict meteorological data corresponding to the future period based on the meteorological forecast recent historical data of each target meteorological source; Repeat the above steps to obtain meteorological data corresponding to each of the future time periods at different time scales.

[0027] In this embodiment, long-term historical weather forecast data for the province is collected from multiple meteorological sources (such as the EC, GFS, CMC, satellite remote sensing, and ground-based meteorological observation stations). The long-term historical data here has the same meaning as previously mentioned, referring to historical data from a longer, earlier period, and will not be further described here. This long-term historical weather forecast data includes weather forecast information from various meteorological sources for different periods in the past and future. For a specific future time period (e.g., the temperature forecast for the third day), the time difference between that future time period and the current moment (i.e., three days) is calculated. Then, based on previously collected historical weather forecast data, each weather source's past forecasts for the same time difference (i.e., forecasts for the next three days) are compared with the actual weather conditions to determine the forecast accuracy for that time difference. For example, if a weather source's past 100 temperature forecasts for the next three days yielded an acceptable error between the forecast results and the actual temperature in 80 of them, then its forecast accuracy for the next three days is 80%. Set an accuracy threshold (e.g., 70%) and identify several meteorological sources with forecast accuracy above this threshold as target meteorological sources for that future time period. For example, if the future time period is 1-3 days, these target meteorological sources are relatively reliable for short-term weather forecasts; if the future time period is 4-10 days, these target meteorological sources are relatively reliable for medium-term weather forecasts; and if the future time period is more than 10 days, these target meteorological sources are relatively reliable for long-term weather forecasts.

[0028] Next, based on the recent historical forecast data from each target meteorological source (e.g., forecasts for similar future time periods within the past week), the corresponding weather data for that future time period is predicted. Because these target meteorological sources have historically performed well in forecasting for this time period (short-term, medium-term, or long-term), a comprehensive analysis of their recent forecast data can more accurately predict the weather conditions for the current future time period of interest. For example, target meteorological sources A, B, and C have a high accuracy rate in forecasting the weather for the next three days. By analyzing the meteorological data they provide for the past two weeks (including actual and forecast data, and the sampling intervals for meteorological data from different target sources vary), the current forecast data for the next three days, such as temperature and humidity, is synthesized. Repeat the above two steps for different future time periods at different time scales. For example, in addition to the three-day period mentioned above, you can also calculate the weather forecast accuracy for different time scales, such as the next week or the next two weeks, determine the corresponding target weather source, and then perform weather data forecasts based on the target weather source's recent historical data. This will obtain weather data for each future time period at different time scales. This weather data, derived from multiple weather sources, provides more comprehensive and accurate basic information for subsequent power forecasts.

[0029] In some embodiments, obtaining enterprise electricity consumption data and residential electricity consumption data of the target province includes: Calculating a deviation between meteorological data of a certain future period and average meteorological data of previous years, and determining a first coefficient and a second coefficient based on the deviation; wherein the first coefficient is less than the second coefficient; Obtaining enterprise electricity consumption data and residential electricity consumption data for a target province, adjusting the first recent historical electricity consumption data in the enterprise electricity consumption data using the first coefficient to obtain third recent historical electricity consumption data; and adjusting the second recent historical electricity consumption data in the residential electricity consumption data using the second coefficient to obtain fourth recent historical electricity consumption data; Finally, the enterprise electricity consumption data includes the first long-term electricity consumption data and the third recent electricity consumption data, and the residential electricity consumption data includes the second long-term electricity consumption data and the fourth recent electricity consumption data.

[0030] In this embodiment, after predicting meteorological data (predicted values) for different future time periods using the aforementioned method, the predicted meteorological data for each future time period is compared with the average meteorological data from previous years. For example, the difference between the temperature on the fifth day in the future (e.g., Qingming Festival) and the average temperature for Qingming Festivals over the past five years is calculated; this difference is the deviation value. Similarly, deviation values ​​for meteorological factors such as humidity and wind speed can also be calculated. The specific analysis factors selected can be freely set. Of course, the deviation values ​​corresponding to multiple factors can also be normalized and then weighted and integrated, but this is not limited to this invention.

[0031] The first coefficient and the second coefficient are determined based on the calculated deviation value.

[0032] Among them, the second coefficient is used to adjust the second electricity consumption recent historical data in the actual residential electricity consumption data. Since residents' daily electricity consumption is usually more susceptible to weather changes, for example, when the temperature changes, residents will use air conditioners, heaters and other equipment more frequently, so the impact of meteorological changes on residents' electricity consumption is relatively large. Correspondingly, the second coefficient is set to a larger value, such as 1.2, which is in line with the characteristics that residents' electricity consumption is more sensitive to meteorological changes. Moreover, by appropriately optimizing and adjusting the residential electricity consumption data that is more sensitive to meteorological changes, the data based on which the subsequent integrated model predicts the province's power load power can be closer to the actual situation, thereby improving the accuracy of the prediction.

[0033] The first coefficient is used to adjust the first recent historical data from the actual enterprise electricity consumption data. While enterprise electricity consumption is also affected by weather, compared to residential electricity consumption, their production activities may not change as quickly as residents' daily lives due to weather changes. Therefore, enterprise electricity consumption is relatively less affected by weather changes. Therefore, the first coefficient is determined to be a smaller value than the second coefficient, for example, 1.05.

[0034] It is understandable that the first and second coefficients can be used to adjust recent historical electricity consumption data by multiplication. The purpose of such adjustments is to more accurately reflect the relationship between recent electricity consumption by businesses and residents and weather changes, while taking into account meteorological factors. This is because weather deviations can have a certain impact on the production and operation activities of businesses and the electricity consumption behavior of residents. Such weather anomalies are known to businesses and residents in advance, which in turn affects electricity consumption in the future period.

[0035] After these adjustments, business electricity consumption data now includes the first long-term historical data (business electricity consumption data from the past few years) and the adjusted third recent historical data; residential electricity consumption data includes the second long-term historical data (residential electricity consumption data from the past few years) and the adjusted fourth recent historical data. This processed electricity consumption data comprehensively considers the different impacts of meteorological factors on business and residential electricity consumption, ensuring that the subsequent electricity consumption data input into the integrated model more accurately reflects the relationship between actual electricity consumption and meteorological changes, thereby improving the accuracy of the province's power load forecast.

[0036] In some embodiments, the grouping of the plurality of prediction models into model groups comprises: Determining a first quantity of target meteorological sources that predict meteorological data for a certain future time period, and obtaining a second quantity based on the first quantity of target meteorological sources; At least the second number of prediction models are screened and divided into the same group to form the model group.

[0037] In this embodiment, the present invention uses an integrated model composed of multiple models to predict the power load power of the entire province to improve the prediction accuracy. However, at the same time, the number of models in the integrated model should be reasonably determined and cannot be too small. Too few models will not achieve the purpose of improving the prediction accuracy, while too many models will lead to reduced prediction efficiency. The present invention adopts the following method, specifically: For a certain future period (e.g., the 7th day in the future), target meteorological sources for predicting meteorological data for this period (i.e., meteorological sources whose meteorological prediction accuracy is higher than an accuracy threshold) have been previously determined. At this time, the number of these target meteorological sources is counted to obtain a first number.

[0038] Based on the first number of target meteorological sources, a second number is derived using a preset matching rule. This preset matching rule may be a pre-set empirical rule, or may be derived based on a specific algorithm or logical deduction, and is not specifically limited in the present invention. For example, if the matching relationship between the number of target meteorological sources and the number of prediction models is set to 1:1, then when the first number is 5, the second number is also 5; or if the number of target meteorological sources is n, the number of prediction models is n+1, and similar rules may be used.

[0039] At least a second number of prediction models are selected from all prediction models (e.g., neural network models, time series models, regression models, and other different types of models). For example, if the second number is 5, then 5 or more prediction models are selected from the plurality of prediction models. These selected prediction models are grouped together, and this group constitutes the model group.

[0040] It is understandable that, given the significant differences in accuracy and granularity of the meteorological data provided by each target meteorological source, the accuracy of meteorological data for a future period based solely on a single target meteorological source is insufficient, while the accuracy of meteorological data for a future period based on multiple target meteorological sources is relatively more reliable. To address this issue, the present invention sets a negative correlation between the second number and the first number of target meteorological sources. For example, a reasonable interval for setting the second number is [3-10]. When the first number is 5, the corresponding second number is set to 5; when the first number is 3, the corresponding second number is set to 7.

[0041] This setup leverages the strengths of multiple forecasting models to comprehensively analyze the relationship between meteorological and electricity usage data, improving the accuracy and reliability of the final forecast results. Different model groups can then be tested and evaluated to select the best performing model for the final integrated model.

[0042] In some embodiments, each model group is tested using real provincial power load power and meteorological data to obtain an evaluation score for each model group. A target model group is obtained based on the evaluation score, and each prediction model in the target model group is integrated to obtain an integrated model, including: Using real provincial electric power load power and meteorological data to test a certain model group, and obtaining a first evaluation score of the model group; Obtaining a third coefficient based on the deviation value matching, and using the third coefficient to fine-tune the first evaluation score to a second evaluation score; Repeat the above steps to obtain a second evaluation score for each model group; The model group with the largest or smallest evaluation score is determined as the target model group, and the prediction models in the target model group are integrated to obtain the integrated model.

[0043] In this embodiment, after the model groups are constructed, their overall prediction performance needs to be tested. This involves using real provincial power load and meteorological data to test each model group, obtaining a corresponding evaluation score, and then determining the model group with the highest evaluation score as the final target model group. Then, a suitable integration method is used to integrate the prediction models in the target model group.

[0044] Specifically, a model group is tested using real provincial power load data and corresponding meteorological data (preprocessed and obtained from multiple meteorological sources). During testing, each prediction model in the group predicts the provincial power load for at least one future time period based on the input meteorological and power load data (including both enterprise and residential power consumption data, and the meaning is equivalent). The predicted results are then compared with the actual provincial power load. A first evaluation score for the model group is calculated using a combination of pre-defined evaluation metrics, such as root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). This score reflects the model group's predictive performance based on the current test data. A higher score (or lower score, depending on the calculation method and the evaluation criteria used, e.g., a lower RMSE indicates a more accurate prediction) indicates a better prediction performance.

[0045] The present invention also provides for fine-tuning the aforementioned first evaluation score to improve its accuracy. Specifically, based on the previously calculated deviation between the meteorological data for a specific future time period and the average meteorological data from previous years (calculated when acquiring electricity consumption data), a third coefficient is derived using a preset matching rule. The preset matching rule can be a function that determines the corresponding third coefficient based on the magnitude of the deviation, with a larger deviation indicating a larger third coefficient. Alternatively, the preset matching rule can be a pre-set table that searches for the corresponding third coefficient based on a range of deviation values. It is understood that a larger deviation indicates more abnormal weather conditions in that province for that future time period, and thus a greater difficulty in predicting the province's power load power by the model group. This is because the historical data has a lower reference value. In this case, the third coefficient can be used to appropriately increase the first evaluation score to more closely align the evaluation of the model group with its actual predictive capability. The third coefficient is positively correlated with the deviation, but its upper limit should be limited, for example, not exceeding 1.2, to avoid distortion caused by excessive fine-tuning.

[0046] Repeat the above two steps, test each model group and fine-tune the evaluation score based on the deviation value, thus obtaining the second evaluation score for each model group. In this way, the prediction performance of all model groups after considering the influence of meteorological data deviation is evaluated.

[0047] The second evaluation scores of all model groups are compared, and the model group with the highest (or lowest, depending on the evaluation criteria) evaluation score is determined as the target model group. This means that this target model group has the best performance in predicting the province's power load, taking into account meteorological data deviations.

[0048] In some embodiments, integrating the prediction models in the target model group to obtain the integrated model includes: Any one of Stacking, Boosting and Bagging is used to integrate the prediction models in the target model group to obtain the integrated model.

[0049] In this embodiment, the individual prediction models in the target model group are integrated. This integration can be performed using a simple weighted average, where each model is assigned a weight based on its performance in the test and the weighted sum of the prediction results is used to obtain the final prediction value. Alternatively, more complex methods such as voting, stacking, boosting, and bagging can be used to further improve prediction performance by combining multiple layers of models, ultimately resulting in an integrated model. This integrated model will be used for subsequent province-wide power load power forecasts, with the hope of improving prediction accuracy and reliability.

[0050] Since the above integration methods are all mature existing technologies, the present invention will not elaborate on them in detail here.

[0051] In some embodiments, the integrated model outputs its predicted provincial electric load power, including: The integrated model outputs its predicted provincial electric load power and its confidence level.

[0052] In this embodiment, the integrated model not only outputs its predicted provincial power load power but also the confidence level associated with that power load power. This confidence level represents the integrated model's self-assessment of the accuracy of its predicted provincial power load power. The confidence level can be used by relevant personnel to analyze and judge the prediction results. For example, if the confidence level is lower than expected, the integrated model can be configured to perform another prediction, which helps improve the accuracy of subsequent power dispatch decisions.

[0053] An embodiment of the present invention further discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute any of the methods described above.

[0054] An embodiment of the present invention further discloses a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any of the above items is executed.

[0055] An embodiment of the present invention further discloses a computer program product, which includes a computer program stored in a computer storage medium. When the computer program is executed by a processor of an electronic device, it implements any of the above methods.

[0056] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0057] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0058] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0059] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions of this disclosure can be achieved. This is not a limitation herein.

[0060] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A provincial electric power forecasting method based on the fusion of multiple meteorological sources and multiple models, characterized in that: The steps include: Acquire meteorological data for multiple future time periods at different time scales from multiple meteorological sources and preprocess the meteorological data; and obtain enterprise electricity consumption data and residential electricity consumption data for the target province, the enterprise electricity consumption data including first long-term electricity consumption historical data and first recent electricity consumption historical data, and the residential electricity consumption data including second long-term electricity consumption historical data and second recent electricity consumption historical data; Divide several prediction models into model groups, test each model group using real provincial power load power and meteorological data, obtain an evaluation score for each model group, screen a target model group based on the evaluation score, and integrate the prediction models in the target model group to obtain an integrated model; The pre-processed meteorological data, the enterprise-type electricity consumption data and the residential-type electricity consumption data are input into the integrated model, and the integrated model outputs its predicted power load of the entire province.

2. The method for predicting provincial electric power based on the fusion of multiple meteorological sources and multiple models according to claim 1 is characterized by: The method of obtaining meteorological data of multiple future time periods at different time scales from multiple meteorological sources includes: Obtaining the long-term historical weather forecast data for the province from each weather source, calculating the time difference between a certain future period and the current moment, and obtaining the weather forecast accuracy of each weather source corresponding to the time difference based on the long-term historical weather forecast data; Determine several meteorological sources whose meteorological forecast accuracy is higher than an accuracy threshold as target meteorological sources corresponding to the future period, and predict meteorological data corresponding to the future period based on the meteorological forecast recent historical data of each target meteorological source; Repeat the above steps to obtain meteorological data corresponding to each of the future time periods at different time scales.

3. The method for predicting provincial electric power based on the fusion of multiple meteorological sources and multiple models according to claim 2 is characterized by: The acquisition of enterprise electricity consumption data and residential electricity consumption data of the target province includes: Calculating a deviation between meteorological data of a certain future period and average meteorological data of previous years, and determining a first coefficient and a second coefficient based on the deviation; wherein the first coefficient is less than the second coefficient; Obtaining enterprise electricity consumption data and residential electricity consumption data for a target province, adjusting the first recent historical electricity consumption data in the enterprise electricity consumption data using the first coefficient to obtain third recent historical electricity consumption data; and adjusting the second recent historical electricity consumption data in the residential electricity consumption data using the second coefficient to obtain fourth recent historical electricity consumption data; Finally, the enterprise electricity consumption data includes the first long-term electricity consumption data and the third recent electricity consumption data, and the residential electricity consumption data includes the second long-term electricity consumption data and the fourth recent electricity consumption data.

4. The method for predicting provincial electric power based on the fusion of multiple meteorological sources and multiple models according to claim 3 is characterized by: The prediction models are divided into model groups, including: Determining a first quantity of target meteorological sources that predict meteorological data for a certain future time period, and obtaining a second quantity based on the first quantity of target meteorological sources; At least the second number of prediction models are screened and divided into the same group to form the model group.

5. The method for predicting provincial electric power based on the fusion of multiple meteorological sources and multiple models according to claim 4 is characterized by: Each model group is tested using real provincial power load power and meteorological data to obtain an evaluation score for each model group. A target model group is obtained based on the evaluation score, and each prediction model in the target model group is integrated to obtain an integrated model, including: Using real provincial electric power load power and meteorological data to test a certain model group, and obtaining a first evaluation score of the model group; Obtaining a third coefficient based on the deviation value matching, and using the third coefficient to fine-tune the first evaluation score to a second evaluation score; Repeat the above steps to obtain a second evaluation score for each model group; The model group with the largest or smallest evaluation score is determined as the target model group, and the prediction models in the target model group are integrated to obtain the integrated model.

6. The method for predicting provincial electric power based on the fusion of multiple meteorological sources and multiple models according to claim 1 is characterized by: The integrating the prediction models in the target model group to obtain the integrated model includes: Any one of Stacking, Boosting and Bagging is used to integrate the prediction models in the target model group to obtain the integrated model.

7. The method for predicting provincial electric power based on the fusion of multiple meteorological sources and multiple models according to claim 1 is characterized by: The integrated model outputs its predicted provincial electric load power, including: The integrated model outputs its predicted provincial electric load power and its confidence level.

Images