CN111668829A - A prediction method of low-voltage users in distribution network based on meteorological characteristic factors - Google Patents

Abstract

The invention relates to a method for predicting the number of low-voltage users of a distribution network based on meteorological characteristic factors. The historical meteorological monitoring data of all automatic weather stations in the area; 2. Perform dimension matching processing on the number of low-voltage users with different temporal and spatial resolutions and the meteorological monitoring data to obtain the set of meteorological feature factors with the same temporal and spatial resolution and The number of low-voltage users, construct a training sample set; 3. Based on the training sample set, establish a prediction model for the number of low-voltage users with the meteorological feature factor set as the input and the number of low-voltage users as the output; 4. Use the obtained prediction model to predict The number of low-voltage users of the distribution network in the area to be predicted. This method is beneficial to predict the number of low-voltage users in the distribution network simply, efficiently and accurately.

Description

A prediction method of low-voltage users in distribution network based on meteorological characteristic factors

technical field

The invention belongs to the field of electric power distribution network, in particular to a method for predicting the number of low-voltage users of a distribution network based on meteorological characteristic factors.

Background technique

The distribution network is an important part of the power system, and it is a power network that directly supplies power to users. The safe, reliable and stable operation of the distribution network is related to whether the power system can provide users with high-quality power supply and service quality. The phenomenon of low voltage in the distribution network is a manifestation of the insufficient power supply capacity of the distribution network, which is prone to occur in rural areas. With the increasing number of household appliances and agricultural equipment in rural areas, low voltage is more likely to occur during peak hours of electricity consumption. Predicting the low voltage problem of the distribution network is of great significance for improving and optimizing the structure of the distribution network, improving the power supply capacity and quality of the distribution network, and solving the low voltage problem of the distribution network.

Large electricity load, large power supply radius, weak voltage regulation capability, unbalanced three-phase load and insufficient reactive power compensation are the main causes of low voltage in distribution networks. At present, the few distribution network low voltage prediction methods based on support vector machine or D-S evidence theory mainly extract the influencing indicators or factors from the above causes. However, in addition to electricity load, other impact indicators such as power supply radius are difficult to predict effectively. The electricity load is mainly forecasted by meteorological factors such as temperature and rainfall. Therefore, existing methods are based on indirect prediction of partial feature quantities.

With the continuous accumulation of low-voltage data and the substantial increase of regional automatic weather stations, as well as the commercialization of gridded and refined weather forecasting in townships, based on big data and artificial intelligence methods, the results of weather forecasting can be directly used to predict the low voltage of the distribution network. The number of voltage users becomes possible. There is no technical report on this aspect yet.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for predicting the number of low-voltage users of a distribution network based on meteorological characteristic factors, which is conducive to simply, efficiently and accurately predicting the number of low-voltage users of a distribution network.

In order to achieve the above-mentioned purpose, the technical scheme adopted in the present invention is: a method for predicting the number of low-voltage users of a distribution network based on meteorological characteristic factors, comprising the following steps:

Step 1: Acquire the historical low-voltage users of the distribution network in the area to be predicted according to the set time and spatial resolution, and simultaneously acquire the historical meteorological monitoring data of all automatic weather stations in the area;

Step 2: Perform dimension matching processing on the number of low-voltage users with different temporal and spatial resolutions and the meteorological monitoring data to obtain the meteorological feature factor set and the number of low-voltage users with the same temporal and spatial resolution, and construct a training sample set;

Step 3: Establish a prediction model for the number of low-voltage users based on the training sample set, which takes the meteorological feature factor set as the input and the number of low-voltage users as the output;

Step 4, using the obtained prediction model for the number of low-voltage users to predict the number of low-voltage users of the distribution network in the area to be predicted.

Further, the low-voltage user refers to the voltage level of 10 kV and below, the deviation of the three-phase power supply voltage of 10 kV and below is lower than 7% of the nominal voltage, and the deviation of the single-phase power supply voltage is lower than 10% of the nominal voltage. User.

Further, the meteorological characteristic factors include the highest temperature, the lowest temperature, the average temperature, the maximum rainfall, the accumulated rainfall, the maximum wind speed and the average wind speed.

Further, the temporal resolution is months, weeks, days or hours, and the spatial resolution is counties (regions), townships or Taiwan districts.

Further, in the step 2, the dimension matching process is to match the meteorological characteristic factors according to the time and space resolutions of the number of low-voltage users, so that the space and time resolutions of the two are consistent.

Further, according to the required time resolution, the meteorological characteristic factor and the number of low-voltage users are matched in the time dimension; according to the required spatial resolution, the meteorological data of the nearest meteorological station or the meteorological data of multiple meteorological stations in the area are matched. The weighted average of meteorological data is used as the meteorological characteristic value of the area to achieve the matching of spatial dimensions.

Further, in the step 2, the meteorological feature factor set is subjected to dimensionality reduction processing and normalization, and used as the input of the sample training.

Further, the dimensionality reduction processing method is to use a correlation analysis method or a principal component analysis method to reduce the dimensionality of the meteorological characteristic factors in the meteorological characteristic factor set.

Further, in the step 3, the constructed low-voltage user number prediction model is a support vector machine model or a neural network model.

Further, when the constructed low-voltage user number prediction model is a support vector machine model, the penalty factor c and the kernel function parameter g of the support vector machine model are optimized, specifically:

Use coarse-fine grid search method, genetic algorithm or particle swarm algorithm to take different penalty factor c and kernel function parameter g , and use k -fold cross-validation method to obtain different prediction results, and take the penalty factor c and the best prediction effect. The kernel function parameter g is used as the optimal parameter to obtain the optimized support vector machine model.

Compared with the prior art, the solution of the present invention has the following beneficial effects:

1. Predicting the number of low-voltage users based on meteorological characteristic factors that can be refined and forecasting is a direct prediction;

2. It avoids the complex physical process of low voltage, and only needs to train the historical data of meteorological factors and low-voltage users. The operation is simple and the accuracy is high, and it is suitable for distribution networks in any area in any time period;

3. With scalability, other predictable characteristic factors that affect the low voltage of the distribution network can be added to the prediction model to further improve the prediction accuracy and facilitate engineering applications.

4. The obtained prediction results can provide reference for the formulation of low-voltage early warning and prevention measures for distribution network.

Description of drawings

FIG. 1 is a flow chart of a method implementation according to an embodiment of the present invention.

FIG. 2 is a prediction result of the number of low-voltage users in XX province in January 2016 according to an embodiment of the present invention.

FIG. 3 is a prediction result of the number of low-voltage users in XX County in 2016 in an embodiment of the present invention.

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

Referring to FIG. 1, the present invention provides a method for predicting the number of low-voltage users of a distribution network based on meteorological characteristic factors, including the following steps:

Step 1: Acquire the historical number of low-voltage users of the distribution network in the area to be predicted according to a certain time and spatial resolution, and simultaneously acquire the historical meteorological monitoring data of all automatic weather stations in the area.

Among them, the low-voltage user refers to the voltage level of 10 kV and below, the three-phase power supply voltage deviation of 10 kV and below is lower than 7% of the nominal voltage, and the single-phase power supply voltage deviation is lower than 10% of the nominal voltage. user.

In this embodiment, the temporal resolution is month, week, day, or hour, and the spatial resolution is county (region), township, or station area. According to the time resolution of months, weeks, days or hours and the spatial resolution of counties (regions), townships or stations, export a certain business system from the distribution network production management system (PMS), smart distribution network system (SMD) and other business systems The low-voltage user data of the distribution network in the area to be forecasted within a time period; the hourly weather monitoring data of all automatic weather stations in the area during the time period are obtained from the meteorological department.

Use the existing historical meteorological data to determine the daily meteorological characteristic factors of a certain area, including the highest temperature (unit: °C), the lowest temperature (unit: °C), the average temperature (unit: °C), and the maximum rainfall (unit: mm/h) , cumulative rainfall (unit: mm/h), maximum wind speed (unit: m/s) and average wind speed (unit: m/s).

Step 2: Perform dimension matching processing on the number of low-voltage users with different temporal and spatial resolutions and the meteorological monitoring data to obtain the meteorological feature factor set and the number of low-voltage users with the same temporal and spatial resolution, and construct a training sample set.

The dimension matching process is to match the meteorological characteristic factors according to the time and space resolutions of the number of low-voltage users, so that the space and time resolutions of the two are consistent. Specifically:

The meteorological characterization factor is time-dimensionally matched with the number of low-voltage users according to the desired time resolution.

According to the required spatial resolution, the meteorological data of the nearest meteorological station (the shortest distance method) or the weighted average of the meteorological data of multiple meteorological stations in the area (weighted average method) are taken as the meteorological characteristic value of the area, with To achieve the matching of spatial dimensions.

In order to reduce the computational complexity and improve the computational efficiency, the meteorological feature factor set is subjected to dimensionality reduction processing in this embodiment, and the specific dimensionality reduction processing method may adopt one of the following methods according to actual needs: correlation analysis method or principal component analysis method is adopted. The method is used to reduce the dimension of the meteorological feature factors in the meteorological feature factor set. The meteorological feature factor set after dimensionality reduction is normalized and used as the input of sample training.

Correlation analysis and principal component analysis are mature dimensionality reduction processing methods in this technical field. The correlation analysis method is a method of constructing feature subsets. It adopts the statistical correlation method to select the feature quantities with strong correlation with the output, remove the feature quantities with weak correlation with the output, and eliminate the feature quantities with strong correlation between the feature quantities. The principal component analysis method is a transforming feature space method, and the principal components are selected according to the cumulative variance contribution rate greater than 85-95% or the eigenvalue greater than 1.

The normalization process is specifically: define the original data as X , the maximum value of the original data set is X _max , the minimum value of the original data set is X _min , and the normalized data X _norm is calculated as follows:

Step 3: Establish a low-voltage user number prediction model based on the training sample set. The model takes the meteorological feature factor set as the input and the low-voltage user number as the output.

A support vector machine model or a neural network model can be used for the constructed prediction model of the number of low-voltage users. In this embodiment, a support vector machine regression method is used to establish the prediction model.

Select the support vector machine model type according to the meteorological feature factor set, and choose SVR, LIBSVM or LSSVM toolbox. In this embodiment, the LIBSVM toolbox is selected because the LIBSVM toolbox can effectively solve the classification problem and cross-validation parameter selection.

The cross-validation idea and coarse-fine grid search method are used to optimize the parameters of the support vector machine model: penalty factor c and kernel function parameter g . In this embodiment, the coarse-fine grid search method is used to obtain different penalty factors c and kernel function parameters g , and the k -fold cross-validation method is used to obtain different prediction results, and the parameter with the best prediction effect, that is, the parameter with the smallest error, is taken as The optimal value is obtained to obtain the optimized support vector machine model. In this embodiment, k is 3. In addition to the coarse-fine grid search method, the genetic algorithm or particle swarm algorithm can also be used to search and optimize the parameters of the support vector machine model.

Step 4, using the obtained prediction model for the number of low-voltage users to predict the future number of low-voltage users of the distribution network in the area to be predicted.

The support vector machine regression prediction model of distribution network low-voltage users is obtained by using the training sample set and the optimal parameter training. The meteorological characteristic factors of the area to be predicted are subjected to dimensionality reduction and normalization processing as the input of the support vector machine regression prediction model. Run the prediction model and output the prediction result of the number of low-voltage users in the area to be predicted.

The following takes the prediction of the number of low-voltage users in XX Province in January 2016 as an example to further illustrate the beneficial effects of the present invention.

1) Obtain the daily meteorological characteristic factors and the number of low-voltage users in XX Province for several years. In this embodiment, the data from 2012 to 2015 is taken as an example, which is used as a training sample set after normalization;

2) Use the support vector machine regression method to establish the prediction model, and use the grid search method to obtain the optimized parameters;

3) Use the training sample set and optimal parameters to establish a support vector machine regression prediction model;

4) Input the meteorological characteristic factor of XX province in January 2016, and normalize it;

5) Run the prediction model and get the prediction results of the number of low-voltage users in various regions in XX Province in January 2016, as shown in Figure 2.

In this embodiment, the daily meteorological characteristic factors and the number of low-voltage users in the whole province of XX are obtained, and they are predicted as a whole. It is also possible to predict the daily meteorological characteristic factors and the number of low-voltage users in each region (accurate to the district/county) in XX Province for several years, and sum up the results of the number of low-voltage users in each region by date to obtain the year of XX Province in 2016. The forecast results of the number of low-voltage users in January.

The following takes the prediction of the number of low-voltage users in each month of XX County in 2016 as an example to further illustrate the beneficial effects of the present invention.

1) Obtain the daily meteorological characteristic factors and the number of low-voltage users in XX County for several years. In this embodiment, the data from 2012 to 2015 is taken as an example, which is used as a training sample set after normalization;

2) Use the support vector machine regression method to establish the prediction model, and use the grid search method to obtain the optimized parameters;

3) Use the training sample set and optimal parameters to establish a support vector machine regression prediction model;

4) Input the daily meteorological characteristic factors of XX County in 2016 and normalize them;

5) Run the prediction model to get the forecast results of the daily low-voltage users in XX County in 2016. Sum the number of low-voltage users by month to get the forecast results of the number of low-voltage users in each month of XX County in 2016, as shown in Figure 3 .

The above are the preferred embodiments of the present invention, all changes made according to the technical solutions of the present invention, when the resulting functional effects do not exceed the scope of the technical solutions of the present invention, belong to the protection scope of the present invention.

Claims (10)

1. A power distribution network low-voltage user number prediction method based on meteorological characteristic factors is characterized by comprising the following steps:

step 1: acquiring the number of historical low-voltage users of the power distribution network in an area to be predicted according to set time and spatial resolution, and acquiring historical meteorological monitoring data of all automatic meteorological stations in the area;

step 2: carrying out dimension matching processing on the low-voltage user number and meteorological monitoring data with different time and spatial resolutions to obtain a meteorological characteristic factor set and a low-voltage user number with the same time and spatial resolution, and constructing a training sample set;

and step 3: establishing a low-voltage user number prediction model based on a training sample set, wherein the model takes a meteorological characteristic factor set as an input quantity and a low-voltage user number as an output quantity;

and 4, predicting the number of the low-voltage users of the power distribution network in the area to be predicted by using the obtained low-voltage user number prediction model.

2. The method for predicting the number of users of the power distribution network with low voltage based on the meteorological characteristic factors as claimed in claim 1, wherein the users with low voltage are users with voltage levels of 10 kv and below, three-phase power supply voltage deviation of 10 kv and below is lower than 7% of the nominal voltage, and single-phase power supply voltage deviation is lower than 10% of the nominal voltage.

3. The method as claimed in claim 1, wherein the meteorological characteristic factors include maximum temperature, minimum temperature, average temperature, maximum rainfall, cumulative rainfall, maximum wind speed, and average wind speed.

4. The method for predicting the number of the low-voltage users of the power distribution network based on the meteorological characteristic factors as claimed in claim 1, wherein the time resolution is month, week, day or hour, and the spatial resolution is county (region), township or platform area.

5. The method as claimed in claim 1, wherein in step 2, the dimension matching process is performed by matching the meteorological characteristic factors according to the time and spatial resolutions of the number of low-voltage users, so that the time and spatial resolutions of the two are consistent.

6. The method for predicting the number of the low-voltage users of the power distribution network based on the meteorological characteristic factors as claimed in claim 5, wherein the meteorological characteristic factors are matched with the number of the low-voltage users in a time dimension according to the required time resolution; according to the required spatial resolution, the meteorological data of the meteorological site closest to the meteorological site or the weighted average of the meteorological data of a plurality of meteorological sites in the area is used as the meteorological characteristic value of the area, so that the matching of spatial dimensions is achieved.

7. The method for predicting the number of the low-voltage users of the power distribution network based on the meteorological characteristic factors as claimed in claim 1, wherein in the step 2, the meteorological characteristic factor set is subjected to dimensionality reduction and normalization and then used as an input of sample training.

8. The method for predicting the number of the low-voltage users of the power distribution network based on the meteorological characteristic factors as claimed in claim 7, wherein the dimension reduction processing method is to perform dimension reduction on the meteorological characteristic factors in the meteorological characteristic factor set by using a correlation analysis method or a principal component analysis method.

9. The method for predicting the number of the low-voltage users of the power distribution network based on the meteorological characteristic factors as claimed in claim 1, wherein the low-voltage user number prediction model constructed in the step 3 is a support vector machine model or a neural network model.

10. The method for predicting the number of the low-voltage users of the power distribution network based on the meteorological characteristic factors as claimed in claim 9, wherein when the constructed low-voltage user number prediction model is a support vector machine model, a penalty factor for the support vector machine modelcAnd kernel function parametersgOptimizing, specifically:

adopting coarse-fine grid search method, genetic algorithm or particle swarm algorithm to obtain different penalty factorscAnd kernel function parametersgBy usingkThe cross-folding verification method obtains different prediction results and takes the punishment factor with the best prediction effectcAnd kernel function parametersgAnd the optimal parameters are used for obtaining the optimized support vector machine model.

Images