Power dynamic prediction system and prediction method using generative time sequence model
Through the generative timing model combined with the analysis of historical electricity consumption data and meteorological data, the problem of existing power prediction systems ignoring seasonal temperature changes is solved, and higher precision power prediction and resource optimization are achieved.
Patent Information
- Application Number
- CN202510155229.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-12
- Publication Date
- 2025-03-14
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing power forecasting system ignores the direct impact of seasonal temperature changes on electricity demand, resulting in low power forecasting accuracy, affecting the efficiency of power system scheduling and resource allocation.
The power dynamic prediction system using a generative timing model is used to collect historical electricity consumption data and meteorological data, analyze and correct the electricity consumption impact to generate a predicted electricity consumption data sequence.
The accuracy of power forecasting is improved, seasonal temperature changes and multi-dimensional influence are taken into account, and the allocation of power resources is optimized.
Smart Images
Figure CN119627908A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of power systems, and in particular to a power dynamic prediction system and prediction method using a generative time series model. Background Art
[0002] In modern society, changes in electricity consumption are affected by many factors, among which seasonal temperature fluctuations are one of the most significant factors. Whether it is residential electricity consumption or industrial production electricity consumption, temperature changes will directly affect the fluctuation of electricity demand. When the temperature is high in summer or low in winter, the use of air conditioners and heating equipment increases significantly, resulting in a surge in electricity demand; while in spring and autumn when the temperature is moderate, the demand for electricity is relatively low. Although the existing power dispatching and forecasting system can provide certain forecasting capabilities based on historical electricity consumption data, there are still many deficiencies in the accurate analysis and application of seasonal temperature fluctuations.
[0003] At present, most of the existing power forecasting systems perform simple trend analysis based on historical power consumption data, ignoring the direct impact of seasonal temperature changes on power demand. Although some systems take temperature factors into account, they usually use static, single models to handle them. For example, some traditional power forecasting models use simple linear regression models to evaluate the impact of temperature on power consumption. Although this method is simple to calculate, it is difficult to capture the nonlinear relationship between temperature and power demand. In fact, the impact of temperature on power demand is often multi-dimensional, and the trend of power demand changes in different temperature ranges is not the same, and traditional models often cannot fully cope with this complexity.
[0004] In summary, the prior art has a technical problem that the direct impact of seasonal temperature changes on electricity demand is ignored, resulting in low accuracy of power forecasting, which further affects the efficiency of power system scheduling and resource allocation. Summary of the invention
[0005] The purpose of this application is to provide a power dynamic prediction system and prediction method using a generative time series model to solve the technical problem in the prior art that the direct impact of seasonal temperature changes on electricity demand is ignored, resulting in low power prediction accuracy, which further affects the efficiency of power system scheduling and resource allocation.
[0006] In view of the above problems, the present application provides a power dynamic prediction system and prediction method using a generative time series model.
[0007] In the first aspect, the present application provides a power dynamic prediction system using a generative time series model, including: a power consumption prediction unit, which is used to collect historical power consumption data sequences within a preset time range in the past of a target area, and input them into a power prediction channel pre-trained based on a generative time series model to predict a basic power consumption data sequence within a preset time range in the future; an accuracy analysis unit, which is used to collect meteorological data sequences predicted within the preset time range in the future of the target area, and extract weather data sequences and temperature data sequences, and analyze weather forecast accuracy parameters and temperature forecast accuracy parameters of the weather data sequences and temperature data sequences; a power consumption impact analysis unit, which is used to perform power consumption impact analysis based on the weather data sequences and temperature data sequences, and correct them through the weather forecast accuracy parameters and temperature forecast accuracy parameters to obtain weather power consumption impact parameter sequences and temperature power consumption impact parameter sequences; a power consumption compensation unit, which is used to compensate the basic power consumption data sequence using the weather power consumption impact parameter sequence and the temperature power consumption impact parameter sequence to obtain a predicted power consumption data sequence as a power dynamic prediction result.
[0008] In the second aspect, the present application also provides a method for dynamic power prediction using a generative time series model, including: collecting historical power consumption data sequences within a preset time range in the past for a target area, and inputting them into a power prediction channel pre-trained based on a generative time series model to predict a basic power consumption data sequence within a preset time range in the future; collecting meteorological data sequences forecasted within the preset time range in the target area in the future, and extracting weather data sequences and temperature data sequences, and analyzing weather forecast accuracy parameters and temperature forecast accuracy parameters of the weather data sequences and temperature data sequences; performing power consumption impact analysis based on the weather data sequences and temperature data sequences, and correcting them using the weather forecast accuracy parameters and temperature forecast accuracy parameters to obtain weather power consumption impact parameter sequences and temperature power consumption impact parameter sequences; using the weather power consumption impact parameter sequences and temperature power consumption impact parameter sequences to compensate for the basic power consumption data sequence, and obtaining a predicted power consumption data sequence as a result of dynamic power prediction.
[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: By collecting the historical electricity consumption data sequence of the target area within the preset time range in the past, and inputting it into the power prediction channel pre-trained based on the generative time series model, the basic electricity consumption data sequence within the preset time range in the future is predicted; the meteorological data sequence predicted within the preset time range in the future in the target area is collected, and the weather data sequence and the temperature data sequence are extracted, and the weather forecast accuracy parameters and the temperature forecast accuracy parameters of the weather data sequence and the temperature data sequence are analyzed; according to the weather data sequence and the temperature data sequence, the power consumption impact analysis is carried out, and the weather forecast accuracy parameters and the temperature forecast accuracy parameters are used for correction to obtain the weather power consumption impact parameter sequence and the temperature power consumption impact parameter sequence; the weather power consumption impact parameter sequence and the temperature power consumption impact parameter sequence are used to compensate the basic power consumption data sequence to obtain the predicted power consumption data sequence as the power dynamic prediction result, so as to realize the power prediction technical goal considering seasonal temperature changes and multi-dimensional influences, and achieve the technical effect of improving power prediction accuracy and optimizing power resource allocation.
[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0012] Figure 1 This is a schematic diagram of the structure of the power dynamic prediction system using the generative time series model in this application; Figure 2 This is a flow chart of a method for power dynamic prediction using a generative time series model in this application.
[0013] Description of reference numerals: The power consumption prediction unit 11 , the accuracy analysis unit 12 , the power consumption impact analysis unit 13 , and the power consumption compensation unit 14 . DETAILED DESCRIPTION
[0014] This application provides a power dynamic prediction system and prediction method using a generative time series model, which solves the technical problem in the prior art that the direct impact of seasonal temperature changes on electricity demand is ignored, resulting in low power prediction accuracy, which further affects the efficiency of power system scheduling and resource allocation. The technical goal of power prediction that takes into account seasonal temperature changes and multi-dimensional impacts is achieved, achieving the technical effect of improving power prediction accuracy and optimizing power resource allocation.
[0015] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.
[0016] For example, please refer to the attached Figure 1 , the present application provides a power dynamic prediction system using a generative time series model, specifically including: The power consumption prediction unit 11 is used to collect the historical power consumption data sequence of the target area within the past preset time range, and input it into the power prediction channel pre-trained based on the generative time series model to predict and obtain the basic power consumption data sequence within the future preset time range.
[0017] Specifically, the target area refers to the area to be subjected to power dynamic forecasting. The preset time range in the past is the historical power consumption time of the target area, which can be the past month, such as the past month. The historical power consumption data of the target area in the past period of time is collected. Assuming that the past time range is a month, the daily power consumption in the past month is collected.
[0018] Then, the power forecast channel is trained through historical data in the past time range, and the historical power consumption data sequence is input into the trained power forecast channel. For example, the power demand in the next time range is predicted based on the power consumption in September, such as predicting the daily power consumption in October.
[0019] The accuracy analysis unit 12 is used to collect the meteorological data sequence predicted within the future preset time range in the target area, extract the weather data sequence and the temperature data sequence, and analyze the weather forecast accuracy parameters and temperature forecast accuracy parameters of the weather data sequence and the temperature data sequence.
[0020] Specifically, weather and temperature will affect electricity consumption data. For example, high or low temperatures will lead to the use of air conditioning and heating equipment. Weather includes sunny, cloudy, overcast, rainy, etc. Weather will affect the use of lighting equipment, which in turn affects electricity consumption data.
[0021] Collect meteorological data sequences within the preset time range in the target area in the future and collect weather forecast data for a period of time in the future. For example, if the preset time range in the future is next month, the meteorological data may include weather information such as sunny, cloudy, and rainy days, as well as the highest and lowest temperatures of each day, such as 25 degrees Celsius and 15 degrees Celsius.
[0022] Next, the weather and temperature related parts are extracted from the meteorological data to form weather data series and temperature data series. The weather data series includes the weather conditions of each day, such as the forecast of sunny days, cloudy days, etc. for consecutive days; the temperature data series records the forecast temperature of each day, such as the highest and lowest temperatures of each day.
[0023] Then, the weather data series and temperature data series are analyzed, and the accuracy parameters of the weather forecast and temperature forecast are calculated respectively. The accuracy of the forecast is obtained by comparing the weather forecasts of similar time periods in history with the actual weather conditions. For example, if the sunny weather forecast in the historical forecast actually occurs 80% of the time, then the accuracy parameter of the weather forecast is 0.8. Similarly, if the error between the temperature forecast and the actual temperature remains within 2 degrees Celsius, then the accuracy of the temperature forecast may be 0.9.
[0024] The electricity consumption impact analysis unit 13 is used to perform electricity consumption impact analysis based on the weather data sequence and the temperature data sequence, and to calibrate through the weather forecast accuracy parameter and the temperature forecast accuracy parameter to obtain a weather electricity consumption impact parameter sequence and a temperature electricity consumption impact parameter sequence.
[0025] Specifically, the impact of meteorological factors on electricity consumption is analyzed based on the weather data series and the temperature data series. For example, when the weather is sunny, electricity consumption may increase due to the increased use of air conditioners or other equipment; when the temperature is 25 degrees Celsius, electricity consumption may be relatively stable, with no significant increase or decrease. Through analysis, the basic weather electricity consumption impact parameter series and the basic temperature electricity consumption impact parameter series are obtained. For example, the electricity consumption impact parameter on a sunny day may be 1.1, while the temperature impact parameter of 25 degrees Celsius is 1.05.
[0026] Next, the basic parameters are corrected using the accuracy parameters of the weather forecast and the accuracy parameters of the temperature forecast. For example, if the accuracy of the weather forecast is 90%, then the sunny day impact parameter 1.1 will be adjusted to 1.1×0.9=0.99. Similarly, when the temperature forecast accuracy is 85%, the temperature impact parameter 1.05 will be adjusted to 1.05×0.85=0.8925. Through correction, a more accurate sequence of weather electricity consumption impact parameters and temperature electricity consumption impact parameter sequences are obtained, which can more accurately predict future electricity demand.
[0027] The power consumption compensation unit 14 is used to compensate the basic power consumption data sequence by using the weather power consumption influence parameter sequence and the temperature power consumption influence parameter sequence to obtain a predicted power consumption data sequence as a power dynamic prediction result.
[0028] Specifically, each weather-related electricity consumption parameter and temperature-related electricity consumption parameter are combined according to a certain weight to comprehensively consider the overall impact of weather and temperature on electricity consumption. The calculated electricity consumption parameter sequence is used to perform compensation calculation on the preliminary predicted basic electricity consumption data sequence. The compensation calculation method is to add 1 to the sum of the electricity consumption parameters, and then multiply it by the basic electricity consumption data to obtain the predicted electricity consumption data sequence as the power dynamic prediction result.
[0029] The power dynamic prediction system using the generative time series model can achieve the technical goal of power prediction taking into account seasonal temperature changes and multi-dimensional influences, and achieve the technical effect of improving power prediction accuracy and optimizing power resource allocation.
[0030] Furthermore, the present application also includes: Collect historical electricity consumption data sequences of the target area within a preset time range in the past; pre-train a power prediction channel based on a generative time series model, wherein the power prediction channel includes power prediction branches for multiple preset time ranges; input the historical electricity consumption data sequences into the power prediction branches corresponding to the preset time range in the future, and predict and output a basic electricity consumption data sequence within the preset time range in the future.
[0031] Specifically, the preset time range in the past is the historical electricity consumption time of the target area, which can be the past month, such as the past month. The historical electricity consumption data sequence of the target area within the preset time range in the past is collected, such as the historical electricity consumption data sequence including the daily electricity consumption in the past month.
[0032] Next, the power forecast channel is used to process power data, including forecast branches for multiple time ranges, each of which corresponds to a different forecast time period. The power forecast branch is pre-trained with power consumption data for the corresponding time period to predict future power consumption trends.
[0033] Then, the historical electricity consumption data sequence is input into the power forecast branch corresponding to the preset time range in the future. For example, each month from January to December corresponds to a power forecast branch, and the power demand for the next 12 months is predicted through the power forecast branch as the basic electricity consumption data sequence.
[0034] By collecting past electricity consumption data in the target area and predicting future electricity consumption through a pre-trained model.
[0035] Furthermore, the present application also includes: According to the multiple preset time ranges, multiple sample historical electricity consumption data sequence sets and multiple sample future electricity consumption data sequence sets are collected in the historical electricity consumption data of the target area respectively; based on the generative time series model, multiple power prediction branches in the power prediction channel are constructed, wherein each power prediction branch includes a power prediction generator and a power prediction discriminator; the multiple sample historical electricity consumption data sequence sets and the multiple sample future electricity consumption data sequence sets are respectively used to perform generative adversarial supervised training on the multiple power prediction branches until convergence, so as to obtain multiple trained power prediction branches; the multiple power prediction branches are combined to obtain a power prediction channel.
[0036] Specifically, within different preset time ranges, sample data is collected from the historical electricity consumption data of each month, including multiple sample historical electricity consumption data series sets and multiple sample future electricity consumption data series sets. For example, twelve months are used as multiple preset time ranges, and the historical electricity consumption data from the past twelve months to the past month are collected as training samples, and the actual electricity consumption data of the next month is used as the corresponding future electricity consumption data series.
[0037] Next, the entire power prediction channel is divided into multiple power prediction branches, each of which corresponds to a different time period prediction. Each power prediction branch includes a power prediction generator and a power prediction discriminator. The power prediction generator is responsible for generating predicted power consumption data, while the power prediction discriminator is used to evaluate whether the generated data is close to the actual power consumption data, thereby being able to more flexibly process power consumption data predictions in different time ranges. For example, each power prediction branch corresponds to a preset time range.
[0038] Then, the adversarial training between the power forecast generator and the power forecast discriminator is continuously optimized until the model's prediction results are close to the actual power consumption data. For example, the power consumption of the next month is predicted using the historical data of the past month until the model's prediction of the next month is close enough to the actual data. The same steps are performed in all preset time ranges until all power forecast branches are fully trained.
[0039] Finally, multiple power forecast branches are combined to obtain a power forecast channel. By combining multiple power forecast branches, a complete power forecast channel is constructed, which can simultaneously process power demand forecasts for multiple different time ranges. For example, each month from January to December corresponds to a power forecast branch. The power forecast branches are combined together to simultaneously predict the power demand for the next 12 months.
[0040] By collecting historical and future electricity consumption data within different time ranges, constructing multiple electricity prediction branches and performing generative adversarial training, we will eventually form an electricity prediction channel that can accurately predict multiple time periods.
[0041] Furthermore, the present application also includes: Collect the meteorological data sequence predicted within the future preset time range in the target area, and extract the weather data sequence and the temperature data sequence; analyze the weather forecast accuracy parameters of the weather data sequence based on the historical forecast weather data in the target area; analyze the temperature forecast accuracy parameters of the temperature data sequence based on the historical forecast temperature data in the target area.
[0042] Specifically, the weather forecast data within the preset time range in the future in the target area is collected to extract the weather-related data series and the temperature-related data series. For example, the preset time range in the future is the next month, and the daily weather conditions, such as sunny days and rainy days, and the daily temperature information, such as 25 degrees Celsius during the day and 15 degrees Celsius at night, are extracted from the weather forecast to form the weather data series and the temperature data series respectively.
[0043] Next, the accuracy of the weather forecast is analyzed based on the historical weather forecast data in the target area. Past weather forecasts are compared with actual weather conditions to derive a parameter for forecast accuracy. For example, if the weather forecast for the past month is 80% accurate, then the accuracy parameter is 0.8.
[0044] Then, based on the historical temperature forecast data, the accuracy of the temperature forecast is analyzed. The accuracy is calculated by comparing the past temperature forecast values with the actual temperature. For example, if the deviation between the temperature forecast and the actual temperature in the past month is within 2 degrees Celsius, then the accuracy of the temperature forecast is 0.9.
[0045] By collecting future meteorological data and extracting weather and temperature separately, the accuracy parameters of weather and temperature forecasts are calculated by analyzing past weather and temperature forecast data, respectively, to further calibrate future forecasts.
[0046] Furthermore, the present application also includes: According to the future preset time range, multiple preset time ranges in the same period of history in the target area are determined; multiple historical forecast weather data sequences within the multiple preset time ranges in the same period are collected, and multiple actual weather data sequences within the multiple preset time ranges in the same period of history that are actually monitored are collected to calculate multiple historical weather accuracy parameters; according to the multiple historical weather accuracy parameters, an average historical weather accuracy parameter is calculated to obtain an average historical weather accuracy parameter; a number of historical weather accuracy parameters are randomly selected from the multiple historical weather accuracy parameters, the deviation from the average historical weather accuracy parameter is calculated, and the mean of the deviation is calculated, and the average historical weather accuracy parameter is compensated and corrected to obtain a weather forecast accuracy parameter, wherein the difference between 1 and the mean of the deviation is multiplied by the average historical weather accuracy parameter to complete the compensation correction and obtain the weather forecast accuracy parameter.
[0047] Specifically, multiple time ranges in the same period in the historical time in the target area are determined based on the future preset time range. For example, if the future preset time range is October, then historical time periods that are also October in the past few years are selected as multiple preset time ranges in the same period to ensure that weather data is analyzed and compared under similar seasonal conditions.
[0048] Next, collect historical weather forecast data and actual monitored weather data within multiple preset time ranges of the same period. For example, in October of the past few years, the daily forecast may be sunny, cloudy, etc., while the actual weather data records the actual conditions of these days. By comparing the forecast data with the actual weather data, looking at the error between the actual weather data and the forecasted weather data, calculate the weather forecast accuracy parameters for each preset time range of the same period, and avoid reducing the accuracy due to the forecast of sunny days but cloudy days in reality.
[0049] Then, based on multiple historical weather accuracy parameters, an average historical weather accuracy parameter is calculated. The larger the deviation, the more unstable the weather forecast accuracy, and the actual weather accuracy parameter may be lower. For example, if the weather forecast accuracy for October in the past five years was 0.85, 0.80, 0.90, 0.75, and 0.88, the average accuracy calculated is (0.85+0.80+0.90+0.75+0.88) / 5=0.836.
[0050] Next, randomly select several historical weather accuracy parameters and calculate their deviations from the average accuracy parameter. For example, after extracting 0.80 and 0.88, the deviations from 0.836 are -0.036 and 0.044 respectively. Then, calculate the mean of all deviations and compensate the average accuracy parameter based on the mean of the deviations. The specific method is to subtract the mean of the deviations from 1 and multiply it by the average historical weather accuracy parameter to obtain the compensated weather forecast accuracy parameter. Assuming the mean of the deviations is 0.01, the corrected accuracy parameter is (1-0.01)×0.836=0.82764.
[0051] By determining historical data from the preset time range in the future, analyzing the historical accuracy of weather forecasts and making corrections, more accurate accuracy parameters for future weather forecasts can be obtained. After correction, accuracy parameters that are more in line with current conditions can be obtained.
[0052] Furthermore, the present application also includes: Collect a set of sample weather data and a set of sample temperature data of the target area, and collect the amplitude of change of the electricity consumption data and the average electricity consumption data of the target area under different sample weather data or different sample temperature data, and mark and obtain a set of sample basic weather electricity consumption impact parameters and a set of sample basic temperature electricity consumption impact parameters; use the set of sample weather data and the set of sample basic weather electricity consumption impact parameters to construct a weather electricity consumption impact classifier; input multiple weather data in the weather data sequence into the weather electricity consumption classifier, and classify and obtain a basic weather electricity consumption impact parameter sequence; use the weather forecast accuracy parameter to perform compensation calculation on the basic weather electricity consumption impact parameter sequence to obtain a weather electricity consumption impact parameter sequence, wherein the compensation calculation is performed by multiplying each basic weather electricity consumption impact parameter by the weather forecast accuracy parameter; collect a set of sample temperature data and a set of sample basic temperature electricity consumption impact parameters, construct a temperature electricity consumption classifier, classify multiple temperature data in the temperature data sequence, and obtain a basic temperature electricity consumption impact parameter sequence; use the temperature forecast accuracy parameter to perform compensation calculation on the basic temperature electricity consumption impact parameter sequence to obtain a temperature electricity consumption impact parameter sequence.
[0053] Specifically, the sample weather data set and sample temperature data set of the target area are collected, and the change range of electricity consumption data and average electricity consumption data under different weather or temperature conditions is collected, that is, the change of electricity consumption in the target area under different weather conditions (such as sunny, cloudy, rainy, etc.) and different temperature conditions is recorded. For example, the electricity consumption on sunny days is 10,000 kWh, and the average electricity consumption is 9,000 kWh, then the change range is 1,000 kWh. Through the change range, the basic impact parameters of each weather and temperature condition on electricity consumption are marked. Among them, the marked change range includes positive and negative real numbers, that is, the weather electricity consumption impact parameter is positive or negative, such as 37℃ high temperature weather electricity consumption data is 10% higher than the average electricity consumption data of the target area, and cloudy electricity consumption is 5% higher; then, the sample weather data set and the basic weather electricity consumption impact parameter set are used to construct a mapping relationship as a weather electricity consumption impact classifier. The role of the weather electricity consumption impact classifier is to classify and predict the impact of electricity consumption based on different weather data. For example, the weather-related electricity consumption classifier predicts that electricity consumption will be 10% higher than the average on sunny days.
[0054] Then, multiple weather data in the weather data sequence are input into the weather power consumption impact classifier to obtain a basic weather power consumption impact parameter sequence. In other words, the weather data for a period of time in the future (such as sunny days, cloudy days) are input into the classifier to obtain the power consumption impact parameters under each weather condition, such as the impact parameter of sunny days is 1.1 and that of cloudy days is 0.9.
[0055] Next, the accuracy parameter of the weather forecast is used to compensate the basic weather electricity consumption parameter. For example, if the accuracy of the weather forecast is 80%, the electricity consumption parameter for sunny days will be corrected to 1.1×0.8=0.88, making the forecast closer to the actual situation.
[0056] Similarly, a sample temperature data set and a sample basic temperature power consumption impact parameter set are collected to construct a mapping relationship as a temperature power consumption impact classifier. The impact of temperature on power consumption is predicted based on the temperature data classification. For example, a temperature of 25 degrees Celsius corresponds to a basic impact parameter of 1.05.
[0057] Then, the temperature data is input into the temperature-electricity impact classifier to obtain the basic temperature-electricity impact parameter sequence. After that, it is compensated by the accuracy parameter of the temperature forecast. For example, if the accuracy of the temperature forecast is 90%, then the impact parameter of 25 degrees Celsius, 1.05, will be corrected to 1.05×0.9=0.945.
[0058] By collecting electricity consumption change data under different weather and temperature conditions, we build a classifier for the impact of weather and temperature on electricity consumption, and use the classifier to predict electricity consumption changes under future weather and temperature conditions. Then, we revise the prediction based on the accuracy of the weather and temperature forecast.
[0059] Furthermore, the present application also includes: A weighted calculation is performed on multiple weather electricity impact parameters and multiple temperature electricity impact parameters in the weather electricity impact parameter sequence and the temperature electricity impact parameter sequence to obtain an electricity impact parameter sequence; and the multiple electricity impact parameters in the electricity impact parameter sequence are used to perform compensation calculation on the basic electricity data sequence to obtain a predicted electricity data sequence, wherein the compensation calculation is performed by multiplying the basic electricity data by the sum of 1 and the electricity impact parameter.
[0060] Specifically, multiple parameters in the weather-related electricity consumption parameter sequence and the temperature-related electricity consumption parameter sequence are weighted and combined according to certain weights to comprehensively consider the overall impact of weather and temperature on electricity consumption. For example, the electricity consumption impact parameter on a sunny day is 1.2, and the electricity consumption impact parameter on a temperature of 25 degrees Celsius is 1.1. After weighted calculation, a comprehensive impact parameter of 1.15 may be obtained.
[0061] Next, use the electricity consumption impact parameter sequence just calculated to perform compensation calculation on the basic electricity consumption data sequence. The compensation calculation method is to add 1 to the sum of the electricity consumption impact parameters, and then multiply it by the basic electricity consumption data. Among them, the electricity consumption impact parameters include positive and negative numbers, and the sum of 1 and the electricity consumption impact parameters may be greater than 1 or less than 1, that is, it affects the improvement of electricity consumption data or affects the reduction of electricity consumption data. For example, if the basic electricity consumption data is 10,000 kWh, and the sum of the electricity consumption impact parameters is 0.15, then the compensated electricity consumption is 10,000×(1+0.15)=11,500 kWh, that is, under the influence of weather and temperature, the actual electricity consumption will be 15% higher than the basic data.
[0062] By obtaining a series of predicted electricity consumption data, the basic electricity demand is taken into account, and the impact of weather and temperature is combined to ensure a more accurate forecast.
[0063] In summary, the power dynamic prediction system using the generative time series model provided by the present application has the following technical effects: By collecting the historical electricity consumption data sequence of the target area within the preset time range in the past, and inputting it into the power prediction channel pre-trained based on the generative time series model, the basic electricity consumption data sequence within the preset time range in the future is predicted; the meteorological data sequence predicted within the preset time range in the future in the target area is collected, and the weather data sequence and the temperature data sequence are extracted, and the weather forecast accuracy parameters and the temperature forecast accuracy parameters of the weather data sequence and the temperature data sequence are analyzed; according to the weather data sequence and the temperature data sequence, the power consumption impact analysis is carried out, and the weather forecast accuracy parameters and the temperature forecast accuracy parameters are used for correction to obtain the weather power consumption impact parameter sequence and the temperature power consumption impact parameter sequence; the weather power consumption impact parameter sequence and the temperature power consumption impact parameter sequence are used to compensate the basic power consumption data sequence to obtain the predicted power consumption data sequence as the power dynamic prediction result, so as to realize the power prediction technical goal considering seasonal temperature changes and multi-dimensional influences, and achieve the technical effect of improving power prediction accuracy and optimizing power resource allocation.
[0064] Embodiment 2: Based on the power dynamic prediction system using the generative time series model in the previous embodiment, the present application also provides a power dynamic prediction method using the generative time series model, please refer to the attached Figure 2 ,include: Collect the historical electricity consumption data sequence of the target area within the preset time range in the past, and input it into the power prediction channel pre-trained based on the generative time series model to predict and obtain the basic electricity consumption data sequence within the preset time range in the future; collect the meteorological data sequence predicted within the preset time range in the target area in the future, extract and obtain the weather data sequence and the temperature data sequence, and analyze the weather forecast accuracy parameters and the temperature forecast accuracy parameters of the weather data sequence and the temperature data sequence; perform electricity consumption impact analysis based on the weather data sequence and the temperature data sequence, and calibrate them through the weather forecast accuracy parameters and the temperature forecast accuracy parameters to obtain the weather electricity consumption impact parameter sequence and the temperature electricity consumption impact parameter sequence; use the weather electricity consumption impact parameter sequence and the temperature electricity consumption impact parameter sequence to compensate the basic electricity consumption data sequence to obtain the predicted electricity consumption data sequence as the power dynamic prediction result.
[0065] Furthermore, the power dynamic prediction method using the generative time series model collects the historical power consumption data sequence of the target area within the past preset time range, and inputs it into the power prediction channel pre-trained based on the generative time series model to predict and obtain the basic power consumption data sequence within the future preset time range, including: Collect historical electricity consumption data sequences of the target area within a preset time range in the past; pre-train a power prediction channel based on a generative time series model, wherein the power prediction channel includes power prediction branches for multiple preset time ranges; input the historical electricity consumption data sequences into the power prediction branches corresponding to the preset time range in the future, and predict and output a basic electricity consumption data sequence within the preset time range in the future.
[0066] Furthermore, the power dynamic prediction method using the generative time series model pre-trains the power prediction channel based on the generative time series model, including: According to the multiple preset time ranges, multiple sample historical electricity consumption data sequence sets and multiple sample future electricity consumption data sequence sets are collected in the historical electricity consumption data of the target area respectively; based on the generative time series model, multiple power prediction branches in the power prediction channel are constructed, wherein each power prediction branch includes a power prediction generator and a power prediction discriminator; the multiple sample historical electricity consumption data sequence sets and the multiple sample future electricity consumption data sequence sets are respectively used to perform generative adversarial supervised training on the multiple power prediction branches until convergence, so as to obtain multiple trained power prediction branches; the multiple power prediction branches are combined to obtain a power prediction channel.
[0067] Furthermore, the method for dynamic power prediction using a generative time series model collects a meteorological data sequence forecasted within the future preset time range in the target area, extracts a weather data sequence and a temperature data sequence, and analyzes the weather forecast accuracy parameters and the temperature forecast accuracy parameters of the weather data sequence and the temperature data sequence, including: Collect the meteorological data sequence predicted within the future preset time range in the target area, and extract the weather data sequence and the temperature data sequence; analyze the weather forecast accuracy parameters of the weather data sequence based on the historical forecast weather data in the target area; analyze the temperature forecast accuracy parameters of the temperature data sequence based on the historical forecast temperature data in the target area.
[0068] Furthermore, the method for dynamic power prediction using a generative time series model includes analyzing the weather forecast accuracy parameters of the weather data sequence according to the historical forecast weather data in the target area, including: According to the future preset time range, multiple preset time ranges in the same period of history in the target area are determined; multiple historical forecast weather data sequences within the multiple preset time ranges in the same period are collected, and multiple actual weather data sequences within the multiple preset time ranges in the same period of history that are actually monitored are collected to calculate multiple historical weather accuracy parameters; according to the multiple historical weather accuracy parameters, an average historical weather accuracy parameter is calculated to obtain an average historical weather accuracy parameter; a number of historical weather accuracy parameters are randomly selected from the multiple historical weather accuracy parameters, the deviation from the average historical weather accuracy parameter is calculated, and the mean of the deviation is calculated, and the average historical weather accuracy parameter is compensated and corrected to obtain a weather forecast accuracy parameter, wherein the difference between 1 and the mean of the deviation is multiplied by the average historical weather accuracy parameter to complete the compensation correction and obtain the weather forecast accuracy parameter.
[0069] Furthermore, the power dynamic prediction method using the generative time series model performs power consumption impact analysis based on the weather data sequence and the temperature data sequence, and performs correction through the weather forecast accuracy parameter and the temperature forecast accuracy parameter to obtain the weather power consumption impact parameter sequence and the temperature power consumption impact parameter sequence, including: Collect a set of sample weather data and a set of sample temperature data of the target area, and collect the amplitude of change of the electricity consumption data and the average electricity consumption data of the target area under different sample weather data or different sample temperature data, and mark and obtain a set of sample basic weather electricity consumption impact parameters and a set of sample basic temperature electricity consumption impact parameters; use the set of sample weather data and the set of sample basic weather electricity consumption impact parameters to construct a weather electricity consumption impact classifier; input multiple weather data in the weather data sequence into the weather electricity consumption classifier, and classify and obtain a basic weather electricity consumption impact parameter sequence; use the weather forecast accuracy parameter to perform compensation calculation on the basic weather electricity consumption impact parameter sequence to obtain a weather electricity consumption impact parameter sequence, wherein the compensation calculation is performed by multiplying each basic weather electricity consumption impact parameter by the weather forecast accuracy parameter; collect a set of sample temperature data and a set of sample basic temperature electricity consumption impact parameters, construct a temperature electricity consumption classifier, classify multiple temperature data in the temperature data sequence, and obtain a basic temperature electricity consumption impact parameter sequence; use the temperature forecast accuracy parameter to perform compensation calculation on the basic temperature electricity consumption impact parameter sequence to obtain a temperature electricity consumption impact parameter sequence.
[0070] Furthermore, the method for dynamic power forecasting using a generative time series model uses the weather power consumption influence parameter sequence and the temperature power consumption influence parameter sequence to compensate the basic power consumption data sequence to obtain a predicted power consumption data sequence, including: A weighted calculation is performed on multiple weather electricity impact parameters and multiple temperature electricity impact parameters in the weather electricity impact parameter sequence and the temperature electricity impact parameter sequence to obtain an electricity impact parameter sequence; and the multiple electricity impact parameters in the electricity impact parameter sequence are used to perform compensation calculation on the basic electricity data sequence to obtain a predicted electricity data sequence, wherein the compensation calculation is performed by multiplying the basic electricity data by the sum of 1 and the electricity impact parameter.
[0071] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The power dynamic prediction system using a generative time series model and the specific examples in the aforementioned embodiment 1 are also applicable to the power dynamic prediction method using a generative time series model in this embodiment. Through the aforementioned detailed description of the power dynamic prediction system using a generative time series model, those skilled in the art can clearly understand the power dynamic prediction method using a generative time series model in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here.
[0072] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
[0073] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.
Claims
1. A power dynamic prediction system using a generative time series model, characterized in that: include: The power consumption prediction unit is used to collect the historical power consumption data sequence of the target area within the preset time range in the past, and input it into the power prediction channel pre-trained based on the generative time series model to predict the basic power consumption data sequence within the preset time range in the future; an accuracy analysis unit, for collecting a meteorological data sequence predicted within the future preset time range in the target area, extracting a weather data sequence and a temperature data sequence, and analyzing a weather forecast accuracy parameter and a temperature forecast accuracy parameter of the weather data sequence and the temperature data sequence; An electricity consumption impact analysis unit, configured to perform electricity consumption impact analysis based on the weather data sequence and the temperature data sequence, and to perform corrections using the weather forecast accuracy parameter and the temperature forecast accuracy parameter to obtain a weather electricity consumption impact parameter sequence and a temperature electricity consumption impact parameter sequence; The electricity consumption compensation unit is used to use the weather electricity consumption influence parameter sequence and the temperature electricity consumption influence parameter sequence to compensate the basic electricity consumption data sequence to obtain a predicted electricity consumption data sequence as a dynamic power prediction result.
2. The power dynamic prediction system using a generative time series model according to claim 1, characterized in that: Collect the historical electricity consumption data sequence of the target area within the preset time range in the past, and input it into the power prediction channel based on the pre-training of the generative time series model to predict the basic electricity consumption data sequence within the preset time range in the future, including: Collect historical electricity consumption data series of the target area within a preset time range in the past; Based on the generative time series model, pre-training a power prediction channel, wherein the power prediction channel includes a plurality of power prediction branches within a preset time range; The historical electricity consumption data sequence is input into the power prediction branch corresponding to the future preset time range, and the prediction output obtains the basic electricity consumption data sequence within the future preset time range.
3. The power dynamic prediction system using a generative time series model according to claim 2, characterized in that: Based on the generative time series model, the power forecast channel is pre-trained, including: According to the plurality of preset time ranges, respectively collecting a plurality of sample historical electricity consumption data series sets and a plurality of sample future electricity consumption data series sets in the historical electricity consumption data of the target area; Based on the generative time series model, a plurality of power prediction branches in the power prediction channel are constructed, wherein each power prediction branch includes a power prediction generator and a power prediction discriminator; Using the multiple sample historical power consumption data sequence sets and the multiple sample future power consumption data sequence sets respectively, the multiple power prediction branches are subjected to generative adversarial supervised training until convergence, thereby obtaining multiple trained power prediction branches; The multiple power prediction branches are combined to obtain a power prediction channel.
4. The power dynamic prediction system using a generative time series model according to claim 1, characterized in that: Collecting a meteorological data sequence predicted within the future preset time range in the target area, extracting a weather data sequence and a temperature data sequence, and analyzing weather forecast accuracy parameters and temperature forecast accuracy parameters of the weather data sequence and the temperature data sequence, including: Collecting a meteorological data sequence predicted within the future preset time range in the target area, and extracting a weather data sequence and a temperature data sequence; Analyzing weather forecast accuracy parameters of the weather data sequence based on historical forecast weather data within the target area; The temperature forecast accuracy parameters of the temperature data sequence are analyzed based on the historical forecast temperature data in the target area.
5. The power dynamic prediction system using a generative time series model according to claim 4, characterized in that: Analyzing the weather forecast accuracy parameters of the weather data sequence according to the historical forecast weather data in the target area includes: According to the future preset time range, determine a plurality of preset time ranges of the same period in the historical time in the target area; Collecting multiple historical forecast weather data sequences within the multiple preset time ranges of the same period, and collecting multiple actual weather data sequences actually monitored within the multiple preset time ranges of the same period, and calculating and obtaining multiple historical weather accuracy parameters; Calculate and obtain an average historical weather accuracy parameter based on the multiple historical weather accuracy parameters; A number of historical weather accuracy parameters are randomly selected from the multiple historical weather accuracy parameters, the deviation from the average historical weather accuracy parameter is calculated, and the mean of the deviation is calculated, and compensation correction is performed on the average historical weather accuracy parameter to obtain a weather forecast accuracy parameter, wherein the compensation correction is completed by subtracting the difference between 1 and the mean of the deviation and multiplying it by the average historical weather accuracy parameter to obtain the weather forecast accuracy parameter.
6. The power dynamic prediction system using a generative time series model according to claim 1, characterized in that: According to the weather data sequence and the temperature data sequence, the electricity consumption impact analysis is performed, and the weather forecast accuracy parameter and the temperature forecast accuracy parameter are used for correction to obtain the weather electricity consumption impact parameter sequence and the temperature electricity consumption impact parameter sequence, including: Collect a set of sample weather data and a set of sample temperature data of the target area, and collect the variation range of the power consumption data and the average power consumption data of the target area under different sample weather data or different sample temperature data, and mark and obtain a set of sample basic weather power consumption impact parameters and a set of sample basic temperature power consumption impact parameters; Using the sample weather data set and the sample basic weather electricity consumption impact parameter set, construct a weather electricity consumption impact classifier; Inputting a plurality of weather data in the weather data sequence into the weather power consumption impact classifier, and classifying to obtain a basic weather power consumption impact parameter sequence; Using the weather forecast accuracy parameter, a compensation calculation is performed on the basic weather electricity consumption impact parameter sequence to obtain a weather electricity consumption impact parameter sequence, wherein the compensation calculation is performed by multiplying each basic weather electricity consumption impact parameter by the weather forecast accuracy parameter; Collecting a sample temperature data set and a sample basic temperature power consumption impact parameter set, constructing a temperature power consumption impact classifier, classifying a plurality of temperature data in the temperature data sequence, and obtaining a basic temperature power consumption impact parameter sequence; The temperature forecast accuracy parameter is used to perform compensation calculation on the basic temperature power consumption influence parameter sequence to obtain the temperature power consumption influence parameter sequence.
7. The power dynamic prediction system using a generative time series model according to claim 1, characterized in that: The weather electricity consumption influence parameter sequence and the temperature electricity consumption influence parameter sequence are used to compensate the basic electricity consumption data sequence to obtain a predicted electricity consumption data sequence, including: Performing weighted calculation on a plurality of weather electricity consumption influence parameters and a plurality of temperature electricity consumption influence parameters in the weather electricity consumption influence parameter sequence and the temperature electricity consumption influence parameter sequence to obtain an electricity consumption influence parameter sequence; The basic electricity consumption data sequence is compensated by using a plurality of electricity consumption influencing parameters in the electricity consumption influencing parameter sequence to obtain a predicted electricity consumption data sequence, wherein the compensation calculation is performed by multiplying the basic electricity consumption data by the sum of 1 and the electricity consumption influencing parameter.
8. A method for predicting power dynamics using a generative time series model, characterized in that: The method is performed by the power dynamic prediction system using the generative time series model according to any one of claims 1 to 7, wherein the power dynamic prediction method using the generative time series model comprises: Collect the historical electricity consumption data sequence of the target area within the preset time range in the past, and input it into the power prediction channel based on the pre-training of the generative time series model to predict the basic electricity consumption data sequence within the preset time range in the future; Collecting a meteorological data sequence predicted within the future preset time range in the target area, extracting a weather data sequence and a temperature data sequence, and analyzing a weather forecast accuracy parameter and a temperature forecast accuracy parameter of the weather data sequence and the temperature data sequence; Conducting electricity consumption impact analysis based on the weather data sequence and the temperature data sequence, and performing corrections using the weather forecast accuracy parameter and the temperature forecast accuracy parameter to obtain a weather electricity consumption impact parameter sequence and a temperature electricity consumption impact parameter sequence; The weather electricity consumption influence parameter sequence and the temperature electricity consumption influence parameter sequence are used to compensate the basic electricity consumption data sequence to obtain a predicted electricity consumption data sequence as a dynamic power prediction result.