Prepurchased electric quantity prediction method considering meteorological factors
By building a machine learning-based pre-purchase power prediction model, combining meteorological factors and user types, the problem of predicting electricity usage behavior of the power grid in extreme meteorological environments is solved, the prediction accuracy and adaptability are improved, and the operating risks and costs of power supply companies are reduced. .
Patent Information
- Application Number
- CN202411142156.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-12-20
AI Technical Summary
The existing technology is difficult to effectively predict the electricity use behavior of users of the power grid in extremely harsh meteorological environments, which leads to power supply companies facing operating risks and increased electricity bill recovery costs.
Through the steps of historical data retrieval, data preprocessing, feature extraction, prediction model construction, model training, meteorological factor analysis and model optimization, a machine learning algorithm is used to build a pre-purchase power prediction model, considering different user types and meteorological factors, and optimizing model parameters to improve prediction accuracy.
It realizes a refined prediction of the pre-purchase power of users, improves the accuracy and adaptability of the prediction results, and reduces the operating risks and electricity bill recycling and operational costs of power supply companies.
Smart Images
Figure CN119168119A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of pre-purchased electricity of an electric power system, and in particular relates to a pre-purchased electricity prediction method taking meteorological factors into account. Background Art
[0002] In recent years, with the rapid development of my country's economy and science and technology, the demand for electricity in people's lives and production has continued to grow rapidly. As a hub platform for energy conversion, utilization and transmission, the power grid connects energy production and consumption. Its safe and stable operation directly affects the development of social economy, people's lives and social stability. With the implementation of the new open power system and pre-purchase power policy, the electricity consumption behavior and electricity demands of new power users are also constantly iterating and upgrading. In this process, the role of the power grid will become more prominent, and it will also face higher service requirements.
[0003] However, for a long time, extreme and severe meteorological disasters have occurred frequently in various places, and the safe operation of power grid equipment and its ancillary facilities has been severely challenged. According to the 2020 disaster report released by the Henan Provincial Emergency Management Department, since the flood season in 2020, there have been 13 large-scale heavy rainfall processes in Henan Province, with an average rainfall of 275 mm, and 16 wind and hail disasters. 220kV and above lines have tripped more than 20 times due to heavy rainfall and secondary disasters. Frequent extreme weather and secondary disasters will pose a great threat to the normal operation of power grid equipment, and will also seriously affect the power supply reliability of power grid enterprises and the power experience of customers. Therefore, in order to adapt to the profound changes in the power grid operation model and improve the customer's power experience, the study of pre-purchased electricity forecasting technology taking into account meteorological factors is a hot issue in the current smart grid field.
[0004] At present, there are relatively few studies on the pre-purchased electricity forecasting technology at home and abroad. With the implementation of the open new power system and the pre-purchased electricity policy, facing the frequent extremely severe weather environment, how to analyze the upcoming new power users as the main service objects of the power grid, and the difficulties in charging or even refusal to pay fees, as well as the high operating risks and high operating costs of electricity fee recovery brought to power supply companies, establish a pre-purchased electricity forecasting model taking into account meteorological factors, so as to improve the overall level of power grid services, has become an urgent issue currently faced by the smart power system.
[0005] In order to solve the above problems, it is necessary to develop a pre-purchase electricity forecasting method taking meteorological factors into account. Summary of the invention
[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide a method for predicting the pre-purchased electricity volume with accurate prediction results and good adaptability taking into account meteorological factors.
[0007] The object of the present invention is achieved by: a method for predicting pre-purchased electricity quantity taking into account meteorological factors, comprising the following steps: (1) Historical data retrieval: retrieve the historical electricity consumption data, user type, electricity consumption date, meteorological data and other relevant data of a large number of users; (2) Data preprocessing: cleaning and sorting the retrieved historical data by user type and electricity consumption date to remove outliers and noise; (3) Feature extraction: extracting features from the pre-processed historical electricity consumption data and meteorological data, including the seasonal characteristics and trend characteristics of electricity consumption and the impact of meteorological factors on electricity consumption; (4) Prediction model construction: Based on the machine learning algorithm, a pre-purchase electricity prediction model is constructed by user type, with the extracted features as input and the pre-purchase electricity as output; (5) Model training: Use pre-processed historical data to repeatedly train the pre-purchase electricity prediction model according to user type and adjust the model parameters; (6) Meteorological factor analysis: analyze the impact of different meteorological factors on electricity consumption, determine the impact weights of different meteorological factors, and add the impact weights to the pre-purchase electricity forecast model; (7) Model optimization: using the pre-processed historical data, re-train the pre-purchase electricity prediction model with the added influence weights and optimize the model parameters; (8) Pre-purchased electricity forecast: Based on the forecasted meteorological data, other relevant data and historical electricity consumption in the same period, the trained pre-purchased electricity forecast model is used to predict the pre-purchased electricity of each user.
[0008] Preferably, the user types are residential users, corporate users and power supply transfer entities. When the user type is a power supply transfer entity, its special events are comprehensively considered.
[0009] Preferably, the meteorological data includes temperature, humidity, air pressure, wind speed, sunshine duration and rainfall, wherein the temperature includes the average temperature, the maximum temperature and the minimum temperature.
[0010] Preferably, the other related data include date types and holiday information, the date types are hour, day, week, month, quarter and year, and the holiday information is weekdays, weekends and holidays.
[0011] Preferably, the machine learning algorithm is a combination of one or more of a support vector machine, a random forest and a neural network.
[0012] Preferably, during the model training process, cross-validation and network search methods are used to optimize model parameters.
[0013] Preferably, during the data preprocessing process, a mean filtering method is used to preprocess obviously abnormal high values or low values.
[0014] Preferably, in the feature extraction process, a time series analysis method is used for feature extraction.
[0015] Preferably, in the meteorological factor analysis process, characteristic importance calculation or correlation analysis method is used to determine the influence weights of different meteorological factors on electricity consumption.
[0016] Due to the adoption of the above technical scheme, the beneficial effects of the present invention are as follows: the present invention adopts eight steps of historical data retrieval, data preprocessing, feature extraction, prediction model construction, model training, meteorological factor analysis, model optimization and pre-purchased electricity prediction to complete the user's pre-purchased electricity prediction, first pre-processing the massive historical data, and then extracting features from the massive historical data of massive users as the input of the prediction model, and taking the obtained pre-purchased electricity as the output of the prediction model, and then using the massive historical data to train the model, adding the weights of meteorological factors affecting electricity consumption to the model, and then using the massive historical data to optimize the model, so that the prediction model is more complete, and then the user's pre-purchased electricity taking into account meteorological factors is refinedly predicted, so that the prediction result is more accurate; the present invention adopts different types of users for separate prediction, which can further improve the prediction result and adapt to the use of users with different electricity consumption behaviors; in general, the present invention has the advantages of accurate prediction results and good adaptability. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0018] It should be understood that the pre-purchase of electricity policy is a new model of electricity supply management, which aims to optimize the allocation and use of grid resources through the method of "purchasing electricity first, then using it". Specifically, it requires users to purchase a certain amount of electricity in advance before using it. This will help power supply companies to regulate the overall grid load supply and balance the supply and demand relationship.
[0019] On this basis, new electricity users generally refer to power supply entities, such as commercial complexes, industrial parks, etc. As the intermediary link, the power supply entity is responsible for collecting prepaid electricity charges from end users within its jurisdiction, and clearing and refunding them based on actual electricity consumption within the specified period.
[0020] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.
[0021] like Figure 1 As shown, the present invention provides a method for predicting pre-purchased electricity volume taking into account meteorological factors, comprising the following steps: Step 1: Retrieve historical data: retrieve massive historical electricity consumption data, user types, electricity consumption dates, meteorological data and other relevant data of massive users. It is easy to find patterns from massive data through big data processing technology, and then reasonably refer to pre-purchase electricity forecast.
[0022] User types can be mainly divided into residential users, corporate users and power supply transfer entities. Different types of users have different electricity consumption behaviors. For example, residential users' electricity consumption is seasonal and mainly related to life. The distribution is relatively stable during the same season and is closely related to temperature and meteorological parameters. For example, corporate users' electricity consumption is mainly production activities, which is closely related to the sales characteristics of their products, and the electricity consumption is higher during the peak production season. For example, the electricity consumption of power supply transfer entities is centralized and large in scale, but the situation is more complicated, and more surplus should be considered relative to the pre-purchased electricity.
[0023] When the user type is a power supply transfer entity, due to its own particularity, its special events, such as hosting large-scale events, are comprehensively considered to ensure the smooth holding of large-scale events.
[0024] Meteorological data include temperature, humidity, air pressure, wind speed, sunshine duration and rainfall, among which temperature includes average temperature, maximum temperature and minimum temperature.
[0025] Other relevant data include date type and holiday information. The date types are hour, day, week, month, quarter and year. Pre-purchase electricity forecasts are made for different time periods according to the requirements of regional power supply companies. Generally speaking, pre-purchase electricity forecasts based on days are the most appropriate and common.
[0026] When the date type is day, the holiday information is weekdays, weekends, and holidays. The purpose of different holiday information is to facilitate the pre-processing of historical data.
[0027] Step 2: Data preprocessing: Clean and organize the retrieved historical data by user type and electricity consumption date to remove outliers and noise. Data preprocessing is mainly for electricity consumption data.
[0028] When preprocessing data, try to distinguish and process the causes of abnormal electricity consumption data, and perform manual experience intervention when necessary; generally, for obviously abnormal high or low values in electricity consumption data, the mean filtering method can be used for preprocessing; but special cases, such as anomalies caused by equipment failure, can be processed through interpolation and other methods.
[0029] There are many methods for data preprocessing. Mean filtering is a typical linear filtering algorithm. It has a good effect on smoothing data and removing noise from data. Abnormal high or low values will deviate from the linear filtering law, so mean filtering is applicable. Interpolation is to use known function values to construct a simple function so that the function is equal to the known function value at a given discrete data point, and then use this constructed function to estimate the function value at other unknown points. Equipment failure generally causes power outages, which may cause low or even zero power consumption, so it is suitable to use interpolation for data preprocessing.
[0030] Step 3, feature extraction: extract features from the pre-processed historical electricity consumption data and meteorological data, including the seasonal characteristics and trend characteristics of electricity consumption and the impact of meteorological factors on electricity consumption.
[0031] Seasonal characteristics, trend characteristics, etc. can be extracted from massive historical electricity consumption data. For example, the seasonal cycle and trend changes of electricity consumption can be extracted using time series analysis. At the same time, the influence of factors such as temperature, humidity, and wind speed on electricity consumption can be extracted from meteorological data. For example, the mean and variance of electricity consumption in different temperature ranges can be calculated as the characteristics of temperature factors. In addition, relevant features can be extracted from information such as production plans and equipment operating status.
[0032] Step 4: Prediction model construction: Based on the machine learning algorithm, a pre-purchase electricity prediction model is constructed by user type, with the extracted features as input and the pre-purchase electricity as output. The machine learning algorithm can be a combination of one or more of the existing support vector machine, random forest and neural network.
[0033] Machine learning algorithm is a method that enables computers to automatically learn rules and patterns from data. The support vector machine algorithm is preferred. The support vector machine algorithm finds an optimal hyperplane in high-dimensional space for classification. It has strong generalization ability, is applicable to a variety of data types, and has good interpretability.
[0034] The random forest algorithm builds multiple decision trees and combines the results of these decision trees to make the final prediction. When building each decision tree, samples are randomly extracted from the original data set for row sampling, and some features are randomly selected for column sampling. Each tree grows independently and does not affect each other. It has high accuracy, is insensitive to noise and outliers, can process high-dimensional data, is easy to parallelize, and improves computing efficiency.
[0035] Among them, the neural network algorithm is a computational model that simulates the structure and function of biological neural networks. The neural network is composed of a large number of neurons connected together. Each neuron receives input signals from other neurons, and after processing through a specific activation function, generates output signals and transmits them to other neurons. It has good adaptability and can adapt to different problems and data distributions. It has strong nonlinear processing capabilities, strong modeling capabilities, parallel computing, high computing efficiency, and good generalization capabilities. After learning enough samples, it can accurately predict and classify new input data; the commonly used ReLU activation function can be used.
[0036] Step 5: Model training: Use the pre-processed historical data to repeatedly train the pre-purchase electricity prediction model according to user type, adjust the model parameters, and improve the accuracy of the model parameters through repeated training.
[0037] During the model training process, methods such as cross-validation and network search can also be used to optimize model parameters.
[0038] Step 6: Meteorological factors analysis: Analyze the impact of different meteorological factors on electricity consumption, determine the impact weights of different meteorological factors, and add the impact weights to the pre-purchase electricity forecast model.
[0039] In the process of meteorological factor analysis, the characteristic importance calculation or correlation analysis method is used to determine the weights of the impact of different meteorological factors on electricity consumption.
[0040] Step 7, model optimization: Use the pre-processed historical data to re-train the pre-purchase electricity prediction model with influencing weights and optimize the model parameters.
[0041] Step 8: Pre-purchased electricity forecast: Based on the forecasted meteorological data, other relevant data and historical electricity consumption in the same period, use the trained pre-purchased electricity forecast model to predict the pre-purchased electricity of each user.
[0042] When the present invention is put into practical use, different types of users are charged in advance based on the pre-purchased electricity amount and combined with the latest pre-purchased electricity policy in the region.
[0043] In summary, based on massive historical data, the present invention uses a pre-purchased electricity prediction model to predict the electricity consumption in the region in the future. The power supply company makes an electricity supply plan in advance and applies it to actual work. By purchasing electricity first and then using it, the existing problem of difficult charging is effectively solved, manpower and material resources are saved, the operating risks and electricity fee recovery operating costs of the power supply company are reduced, and the stability and reliability of the power grid are improved.
[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.
Claims
1. A method for predicting pre-purchased electricity quantity taking into account meteorological factors, characterized in that: The steps include: (1) Historical data retrieval: retrieve the historical electricity consumption data, user type, electricity consumption date, meteorological data and other relevant data of a large number of users; (2) Data preprocessing: cleaning and sorting the retrieved historical data by user type and electricity consumption date to remove outliers and noise; (3) Feature extraction: extracting features from the pre-processed historical electricity consumption data and meteorological data, including the seasonal characteristics and trend characteristics of electricity consumption and the impact of meteorological factors on electricity consumption; (4) Prediction model construction: Based on the machine learning algorithm, a pre-purchase electricity prediction model is constructed by user type, with the extracted features as input and the pre-purchase electricity as output; (5) Model training: Use pre-processed historical data to repeatedly train the pre-purchase electricity prediction model according to user type and adjust the model parameters; (6) Meteorological factor analysis: analyze the impact of different meteorological factors on electricity consumption, determine the impact weights of different meteorological factors, and add the impact weights to the pre-purchase electricity forecast model; (7) Model optimization: using the pre-processed historical data, re-train the pre-purchase electricity prediction model with the added influence weights and optimize the model parameters; (8) Pre-purchased electricity forecast: Based on the forecasted meteorological data, other relevant data and historical electricity consumption in the same period, the trained pre-purchased electricity forecast model is used to predict the pre-purchased electricity of each user.
2. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1 is characterized in that: The user types are residential users, corporate users and power supply transfer entities. When the user type is a power supply transfer entity, its special events are comprehensively considered.
3. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1 is characterized in that: The meteorological data include temperature, humidity, air pressure, wind speed, sunshine duration and rainfall, wherein the temperature includes the average temperature, the maximum temperature and the minimum temperature.
4. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1 is characterized in that: The other related data include date type and holiday information, the date type is hour, day, week, month, quarter and year, when the date type is day, the holiday information is working day, weekend and holiday.
5. The method for predicting the pre-purchased electricity quantity taking into account meteorological factors according to claim 1 is characterized in that: The machine learning algorithm is one or more combinations of support vector machines, random forests and neural networks.
6. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1, characterized in that: During the model training process, cross-validation and network search methods are used to optimize model parameters.
7. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1, characterized in that: During the data preprocessing process, the mean filtering method is used to preprocess the obviously abnormal high or low values.
8. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1, characterized in that: In the feature extraction process, a time series analysis method is used to extract features.
9. The method for predicting pre-purchased electricity quantity taking into account meteorological factors according to claim 1, characterized in that: In the meteorological factor analysis process, the influence weights of different meteorological factors on electricity consumption are determined by calculating the characteristic importance or correlation analysis method.