Electricity customer load prediction method and system

Abstract

According to the power customer load prediction method and system provided by the invention, personalized prediction is carried out according to power consumption modes of different users or regions, different models are trained for different seasons and special events of power customers, and continuous updating is carried out, so that the prediction personalization is improved, and the flexibility is enhanced. And meanwhile, data from various sources such as historical load data, meteorological data, holidays and festivals, user participation power market regulation information and the like are combined, so that the accuracy of power consumption prediction is improved, the accuracy and real-time performance of customer power load prediction are improved, and a scientific basis is provided for power dispatching, resource allocation and electricity price strategy formulation.

Description

Technical Field

[0001] The present invention relates to the field of information prediction, and in particular to the field of customer-side load prediction in a power grid system, and more specifically to a method and system for predicting load on power customers. Background Art

[0002] With the development of economy and social progress, the demand for electricity, as the core part of modern social infrastructure, continues to grow. Especially in the context of the accelerated process of new urbanization and industrialization, electricity consumption has shown diversity and complexity, which has put forward higher requirements for the planning, operation and service of power grids. Therefore, predicting customer load has become the key for power companies to optimize resource allocation, improve service quality and ensure power supply stability.

[0003] In order to provide matching services to electricity users, it is necessary to effectively predict the electricity consumption information of electricity users, so as to reasonably plan the distribution and configuration of electricity consumption in advance. However, traditional load forecasting methods have limitations in accuracy, flexibility, and adaptability, and cannot meet the needs of power grid planning and pre-configuration. Summary of the invention

[0004] To solve the above problems, the present invention provides a method and system for predicting load for electricity users. By integrating multi-source data and analyzing personalized electricity consumption behavior of users, an intelligent and adaptive power load forecasting system is constructed to provide more accurate and timely load forecasting services for electricity sales companies.

[0005] The first aspect of the present invention discloses a method for predicting load of electricity customers, the method comprising:

[0006] Step S1: collecting daily load data, meteorological data, holidays, social and economic factors, electricity price policies, and special event information data from electricity users within a current predetermined time period;

[0007] Step S2: within the time period, according to different seasonal periods, cluster analysis is performed according to the daily load curve of the electricity users to analyze different electricity consumption patterns of the electricity users;

[0008] Step S3: For electricity users, sensitivity analysis is performed on their different factor independent variables to construct corresponding feature vectors;

[0009] Step S4: training a load forecasting model based on the feature vector.

[0010] According to the method for load forecasting of electricity users described in the first aspect of the present invention, step S2 specifically comprises: selecting typical daily electricity consumption behavior curves of electricity users in different seasonal periods for cluster analysis, using cosin distance to calculate the distance of electricity consumption curve vectors, and dividing users with the same electricity consumption behavior into the same cluster.

[0011] The method for predicting load for electricity users according to the first aspect of the present invention further includes: selecting electricity consumption curves of electricity users during holidays and special event days in different seasons for cluster analysis to obtain electricity consumption behavior data.

[0012] According to the method for predicting load of electricity customers described in the first aspect of the present invention, step S3 specifically includes:

[0013] We selected a variety of independent variables including temperature, wind speed, holidays, and special events, and used the multivariate linear regression model for regression analysis. We combined the PCA dimensionality reduction method to select important independent variables, and constructed a feature vector based on the independent variable factors selected by each electricity customer.

[0014] According to the method for predicting load of electricity users described in the first aspect of the present invention, step S4 specifically includes:

[0015] Based on the acquired feature vectors, the prediction model is trained using the collected historical data.

[0016] According to the method for predicting load for electricity users described in the first aspect of the present invention, the method of using the collected historical data to train the prediction model further includes:

[0017] Temperature, humidity and rainfall are selected as characteristic variables, load sequence is used as main variable, original data is divided into load data, meteorological data and event data, and original load sequence is constructed; original load sequence is converted into multiple frequency domain subsequences through Fourier transform.

[0018] According to the first aspect of the present invention, the method for predicting load for electricity users further includes: performing model training on the multiple frequency domain subsequences using a Bi-LSTM algorithm.

[0019] The second aspect of the present invention discloses an image retrieval system based on a super dictionary, the system comprising a processing unit, the processing unit being configured to: execute steps for implementing the method for predicting load of electricity customers described in the first aspect.

[0020] The third aspect of the present invention discloses an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the program in the memory to implement the power customer load forecasting method described in the first aspect.

[0021] A fourth aspect of the present invention discloses a computer-readable storage medium, wherein the storage medium stores the method for predicting load for electricity customers described in the first aspect.

[0022] In summary, the solution proposed by the present invention has the following technical effects: The present invention provides a method and system for predicting load for electricity users, which performs personalized prediction based on the electricity consumption patterns of different users or regions, trains different models for different seasons and special events of electricity users, and continuously updates and improves the personalization of prediction, thereby enhancing flexibility. At the same time, it combines data from multiple sources such as historical load data, meteorological data, holidays, and information on users' participation in electricity market regulation to improve prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 is a flow chart of a method for predicting load for electricity customers according to Embodiment 1 of the present invention;

[0025] Figure 2 is an overall block diagram of a system according to an embodiment of the present invention;

[0026] Figure 3 is a structural diagram for analyzing a typical daily electricity consumption behavior of an electricity user according to an embodiment of the present invention;

[0027] Figure 4 is a flow chart of a load prediction algorithm for electricity users according to Embodiment 2 of the present invention;

[0028] Figure 5 It is the load forecast result of a certain user on two days obtained according to the electricity customer load forecasting algorithm of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] Figure 1 is a flow chart of a method for predicting load for power customers according to Embodiment 1 of the present invention. Figure 1 As shown, the implementation steps are as follows:

[0031] Step S1: collecting daily load data, meteorological data, holidays, social and economic factors, electricity price policies, and special event information data from electricity users within a current predetermined time period;

[0032] For customers of power sales companies, the daily load data, meteorological data, holidays, power market policies, and special event information data for the past two years are collected.

[0033] Customer daily load data collection: Daily load data is collected every 15 minutes and consists of an array vector of 96 active load values.

[0034] Meteorological data are mainly divided into three categories: temperature, humidity, and wind speed. Each type of data is consistent with the collection step of daily load data and is also an array of 96 values.

[0035] Holiday data is marked with holidays in the past two years. Usually, the electricity consumption pattern during holidays is different from that on weekdays and needs to be trained separately.

[0036] Electricity market policies include regular market electricity prices and special demand response policies. Usually, different electricity price policies will affect users' electricity usage habits and thus affect the load.

[0037] Special events: whether the customers served by the main power sales companies are undertaking large-scale events such as sports events and concerts. Usually, such special events will significantly change the electricity demand in the local area in a short period of time.

[0038] Step S2: within the time period, according to different seasonal periods, cluster analysis is performed based on the daily load curves of the electricity users to analyze different electricity consumption patterns of the electricity users.

[0039] In different seasonal periods, cluster analysis is performed based on the customer's daily load curve to analyze the different electricity consumption patterns of customers. Specifically, by selecting the customer's typical daily electricity consumption curve in different seasonal periods for cluster analysis, the main purpose is to analyze the user's electricity consumption behavior. Therefore, the cosin distance is used to calculate the distance of the electricity consumption curve vector, and users with the same electricity consumption behavior are divided into the same cluster.

[0040] Select customers' electricity consumption curves during holidays and special events in different seasons for cluster analysis to analyze their electricity consumption behaviors under atypical circumstances, such as Figure 3 As shown in FIG. 1 , a typical daily electricity consumption behavior analysis structure diagram of an electricity user is shown. The diagram shows the relevant characteristics of the user's electricity consumption, specifically the electricity consumption behavior curves of four types of electricity users.

[0041] Step S3: For electricity users, sensitivity analysis is performed on their different factor independent variables to construct corresponding feature vectors.

[0042] Among them, for different customers, sensitivity analysis of different independent variables of different factors is carried out, different load forecasting feature engineering is constructed, and multiple independent variables such as temperature, wind speed, holidays, special events, etc. are selected. The multivariate linear regression model is used for regression analysis. The independent variable factors with large coefficients are the factors that affect the user sensitivity. At the same time, combined with PCA and other dimensionality reduction methods, important independent variables are comprehensively selected.

[0043] A feature vector is constructed based on the variable factors selected by each customer. The feature vectors of each customer for different scenarios are different to ensure personalization.

[0044] Step S4: training a load forecasting model based on the feature vector.

[0045] The overall block diagram of this method system is as follows Figure 2 As shown, it mainly includes major functional modules such as data collection, historical data conversion and storage, real-time data processing, user electricity consumption behavior analysis, user load prediction model training, and user load real-time prediction.

[0046] In this embodiment, the daily load data, meteorological data, holidays, socio-economic factors, electricity price policies, and special event information data of the power sales company's customers in the past two years can be collected; 2) In different seasonal periods, cluster analysis is performed based on the customer's daily load curve to analyze the different electricity consumption patterns of customers; 3) For different customers, sensitivity analysis is performed on their different factor independent variables to construct different load forecasting feature projects; 4) Load forecasting model training is performed for different customers.

[0047] Load forecasting model training is carried out for different customers, mainly including customers with different electricity consumption behaviors, and feature vectors constructed based on sensitive factors selected by customers, using collected historical data to train forecasting models.

[0048] like Figure 4 As shown, it is a flow chart of the load prediction method for electricity customers in the second embodiment of the present invention. For users in different groups, temperature, humidity, rainfall, etc. can be selected as characteristic variables, the load sequence can be used as the main variable, and the original data can be divided into load data, meteorological data, event data, etc.

[0049] Converting the original load sequence into multiple frequency domain subsequences through Fourier transform can solve the problems of load fluctuation and nonlinearity and improve the accuracy of load forecasting.

[0050] In order to solve the problem of too many subsequences and long calculation time, it is usually set to retain 6 subsequences in the frequency domain. The Bi-LSTM algorithm is used to train the model for each subsequence. Finally, the load is predicted by a set of trained models, and the multiple predicted subsequences are inverse Fourier transformed to generate the load prediction value in the time domain.

[0051] The prediction model can be a combination of multiple forms, such as the LSTM prediction model and its variants that can directly predict the sequence, or a linear regression prediction model that can be built at each acquisition time point. Through the combination of 96 models, the customer's one-day prediction can be completed. Figure 5 The load forecast result for a certain user on two days.

[0052] The present invention also provides a power customer load prediction system for implementing the above-mentioned power customer load prediction method.

[0053] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the program in the memory to implement the above-mentioned power customer load forecasting method.

[0054] The present invention also provides a computer-readable storage medium, which stores a computer program for implementing the above-mentioned electricity customer load prediction method.

[0055] In summary, the present invention proposes a method and system for predicting the load of electricity users. According to the electricity consumption characteristics of different electricity users, sensitivity analysis is performed on the independent variables of different factors, and different load prediction feature sequences are constructed. By considering the influence of multiple data factors for different electricity users and different time periods, personalized load prediction model training is performed, thereby realizing accurate prediction of the load of electricity users. The accuracy and real-time performance of customer power load prediction is improved, providing a scientific basis for power dispatching, resource allocation, and electricity price strategy formulation, helping power sales companies to operate efficiently and optimize power market transactions.

[0056] Please note that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several variations and improvements can be made, which all belong to the scope of protection of the present application. Therefore, the scope of protection of the patent in this application shall be based on the attached claims.

Claims

1. A method for predicting load of electricity customers, characterized in that: The method comprises: Step S1: collecting daily load data, meteorological data, holidays, social and economic factors, electricity price policies, and special event information data from electricity users within a current predetermined time period; Step S2: within the time period, according to different seasonal periods, cluster analysis is performed according to the daily load curve of the electricity users to analyze different electricity consumption patterns of the electricity users; Step S3: for the electricity user, sensitivity analysis is performed on the independent variables of different factors to construct a corresponding feature vector; Step S4: training a load forecasting model based on the feature vector.

2. The method for predicting power customer load according to claim 1, characterized in that: The step S2 specifically includes: selecting typical daily electricity consumption behavior curves of electricity users in different seasons for cluster analysis, using cosin distance to calculate the distance of electricity consumption curve vectors, and dividing users with the same electricity consumption behavior into the same cluster.

3. The method for predicting power customer load according to claim 2, characterized in that: Also includes: The electricity consumption curves of electricity users during holidays and special event days in different seasons are selected for cluster analysis to obtain electricity consumption behavior data.

4. The method for predicting power customer load according to claim 1, characterized in that: The step S3 specifically includes: We selected a variety of independent variables including temperature, wind speed, holidays, and special events, and used the multivariate linear regression model for regression analysis. We combined the PCA dimensionality reduction method to select important independent variables, and constructed a feature vector based on the independent variable factors selected by each electricity customer.

5. The method for predicting power customer load according to claim 2, characterized in that: Step S4 specifically includes: Based on the acquired feature vectors, the prediction model is trained using the collected historical data.

6. The method for predicting power customer load according to claim 5, characterized in that: The use of the collected historical data to train the prediction model also includes: Temperature, humidity and rainfall are selected as characteristic variables, load sequence is used as main variable, original data is divided into load data, meteorological data and event data, and original load sequence is constructed; original load sequence is converted into multiple frequency domain subsequences through Fourier transform.

7. The method for predicting power customer load according to claim 6, characterized in that: Also includes: The Bi-LSTM algorithm is used to perform model training on the multiple frequency domain subsequences.

8. A power customer load forecasting system, characterized in that: The system comprises a processing unit, and the processing unit is configured to execute steps for implementing the method for predicting load of electricity customers according to any one of claims 1-7.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor is used to execute the program in the memory to implement the power customer load forecasting method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program for implementing the method for predicting load for electricity customers according to any one of claims 1 to 7.

Images