CN114819280A - A kind of distribution network load forecasting method and system - Google Patents

Abstract

The invention discloses a distribution network load forecasting method, belonging to the technical field of power load forecasting. The specific steps of the forecasting method are as follows: step 1: obtaining historical power data; step 2: determining load influencing factors and selecting similar days; step 3: Determine the electric vehicle load data; Step 4: Extract the training set; Step 5: Build a regional short-term load prediction model; Step 6: Load prediction; According to the influence of different loads in different regions and electric vehicle loads, a regional short-term load prediction model is constructed, which is conducive to improving the accuracy of regional short-term power load prediction, and thus is conducive to providing an important reference for the reasonable control of the start and stop of generator sets.

Description

A kind of distribution network load forecasting method and system

technical field

The invention relates to the technical field of power load forecasting, in particular to a method and system for load forecasting of a distribution network.

Background technique

With the closer relationship between electric energy and residents' life, industrial and agricultural production, and social development, whether safe, high-quality, economical and reliable electric energy can be satisfied is more and more important for electricity customers; electric energy itself is difficult to store in large quantities, and it needs to be generated, transmitted, Change, distribution and other links can meet the user's requirements, and unavoidable losses will be formed in each link; therefore, it is necessary to effectively reduce losses, improve the utilization rate of primary energy, and formulate accurate power generation plans and reasonable dispatch plans in advance. It is an important means to improve the stability and economy of the power grid, and accurate load forecasting provides a reasonable reference for the formulation of plans and proposals; currently, it can be divided into ultra-short-term load forecasting, short-term load forecasting, medium There are four types of load forecasting and long-term load forecasting. Among them, short-term load forecasting plays an important role in unit start-stop, water-thermal power coordination, tie line exchange power, load economic distribution, reservoir dispatching and equipment maintenance, so it has become the focus of current research; for short-term load forecasting For forecasting, it is necessary to fully study the law of power grid load changes and analyze the relevant factors of load changes. However, due to the large differences in power load influencing factors between different regions, and the connection of a large number of electric vehicle loads has caused huge short-term load forecasting in the region. It is becoming more and more important to improve the accuracy of regional short-term power load forecasting, reasonably control the start and stop of generator sets, and improve the stability and economy of the power grid; therefore, it is particularly important to invent a method and system for load forecasting of distribution network ;

After searching, Chinese Patent No. CN107491812B discloses a short-term load forecasting method considering real-time electricity price. The invention performs short-term load forecasting based on the relationship between electricity price and load, and has good forecasting performance;

After searching, Chinese Patent No. CN102930356A discloses a short-term load forecasting method based on the sensitivity of meteorological factors. The invention improves the forecasting accuracy of short-term power load by correcting meteorological factors;

Judging from the existing patents that have been applied for, most of the existing short-term load forecasting methods rely on specific load influencing factors (electricity price, time and weather) for short-term load forecasting. Although this type of method has a fast forecasting speed, it does not consider other load impact related Therefore, it is impossible to improve the accuracy of regional short-term power load forecasting, and thus cannot provide an important reference for rationally controlling the start and stop of generator sets. To this end, we propose a distribution network load forecasting method and system. .

SUMMARY OF THE INVENTION

The purpose of the present invention is to propose a method and system for predicting the load of a distribution network in order to solve the defects existing in the prior art.

In order to achieve the above object, the present invention adopts the following technical solutions:

A distribution network load forecasting method, the specific steps of the forecasting method are as follows:

Step 1: Obtain historical power data, and obtain historical power data of the area to be predicted; the historical power data includes historical load impact data and historical load data;

Step 2: determining the load influencing factors and selecting similar days, determining the main influencing factors of the load in the area to be predicted according to the change of the historical load data, and extracting the similar historical days of the prediction days in the area to be predicted according to the similarity measurement model;

Step 3: Determine the load data of electric vehicles, obtain relevant data of electric vehicles, and determine the daily load data of electric vehicles in the area to be predicted according to them;

Step 4: Extracting the training set, extracting the load impact data and load data of the similar historical days and the daily load data of the electric vehicle as a training set, a test set and a verification set;

Step 5: Build a regional short-term load prediction model, construct a deep network model, and use the training set as the input data of the deep network model for training to form a regional short-term load prediction model, and verify it through the verification test set. If the error of the load prediction model is within the threshold, the model will be output; otherwise, skip to step 4 to re-extract the training set;

Step 6: Load forecasting, collecting and inputting the relevant data of the forecast date of the area to be forecasted, and forecasting based on the regional short-term load forecasting model, obtaining and outputting the short-term regional load forecasting result.

Further, the historical load influence data in step 1 includes specific influence factor data and random influence factor data; the specific influence factor data includes electricity price, time and weather; the random influence factor data needs to be specifically determined according to the region to which it belongs.

Further, the specific process for determining the load influencing factors described in step 2 is as follows:

S1: Select a certain historical day of a certain load influencing factor one by one, and extract the historical load data of the certain historical day;

S2: Obtain a historical day that is the same or similar to other influencing factors of a certain historical day and that has a different influencing factor of a certain load, as a comparison day;

S3: Divide the historical load data of the comparison day and the historical load data of a certain historical day to obtain an operation result, and if the operation result is outside 0.5-1.5, determine the influence of the certain load The factor is the main influencing factor of the load of the area to be predicted, otherwise, the certain load influencing factor is excluded.

Further, the selection of the similar days in step 2 is determined according to the main influencing factors of the regional load to be predicted, which is calculated by the similarity measurement model, and its specific formula is as follows:

In the formula: cosθ is the similarity, the interval is [0, 1]; A is the load impact feature vector of the forecast day in the area to be predicted; B is the load impact feature vector of the historical day;

For the similar days, historical days with a cosθ value above 0.8 are selected as similar days.

Further, the electric vehicle-related data in step 3 includes daily electric vehicle flow and daily electric vehicle charging times.

A distribution network load forecasting system, comprising a data extraction module, a data preprocessing module, a deep learning module, a data acquisition module, a load forecasting module and an output display module;

The data extraction module is used to obtain historical power data and electric vehicle-related data in the power database and the electric vehicle database, and based on this, extract the load impact data and load data of similar historical days and the electric vehicle daily load data;

The data preprocessing module is used to preprocess the load impact data and load data of the similar historical days and the daily load data of the electric vehicle;

The deep learning module is configured to use the preprocessed load impact data and load data of similar historical days and the daily load data of electric vehicles as a training set, and perform learning based on a deep learning network to generate a regional short-term load prediction model;

The data collection module is used to collect the relevant data of the forecast day in the area to be forecasted;

The load forecasting module is used for inputting the relevant data of the forecast day of the area to be forecasted, and predicting based on the short-term load forecasting model of the area, and generating a short-term load forecasting result of the area;

The output display module is used for outputting and displaying the short-term load forecasting result of the area, so as to provide an important reference for reasonably controlling the start and stop of the generator set.

Compared with the prior art, the beneficial effects of the present invention are:

The present application discloses a distribution network load forecasting method and system, which determine the corresponding load influencing factors according to the load influence degrees of different load influencing factors in different regions, and select the load on similar historical days similar to the forecast day through the similarity measurement model Influence data and load data, and at the same time determine the daily load data of electric vehicles through the daily driving electric vehicle flow and the daily driving electric vehicle charging times, and finally conduct training and prediction through the deep network model; this application fully considers other load impact related factors and electric vehicles. The impact of load access on the load of the distribution network is conducive to improving the accuracy of regional short-term power load forecasting, which in turn is conducive to providing an important reference for the rational control of the start and stop of generator sets.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention.

Fig. 1 is the overall flow chart of a distribution network load prediction method proposed by the present invention;

FIG. 2 is an overall schematic diagram of a distribution network load prediction system proposed by the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments.

In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inside", " The orientation or positional relationship indicated by "outside" is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the indicated device or element must have a specific orientation, so as to The specific orientation configuration and operation are therefore not to be construed as limitations of the present invention.

Referring to FIG. 1 , this embodiment discloses a method for predicting the load of a distribution network. The specific steps of the method are as follows:

Step 1: Obtain historical power data, and obtain historical power data of the area to be predicted; historical power data includes historical load impact data and historical load data;

Specifically, the historical load influence data includes specific influence factor data and random influence factor data; specific influence factor data includes electricity price, time and weather; random influence factor data needs to be determined according to the region to which it belongs, and the random influence factor includes but is limited to policy. factors, epidemic prevention and control factors, and heating time, etc., which are mainly determined according to the conditions of a specific area.

Step 2: Determine the load influencing factors and select similar days, determine the main influencing factors of the load in the area to be predicted according to the changes of historical load data, and extract the similar historical days of the predicted days in the area to be predicted according to the similarity measurement model;

Specifically, the specific process of determining the load influencing factor is as follows: select a certain historical day of a certain load influencing factor one by one, and extract the historical load data of the certain historical day; A similar historical day with a different load impact factor is used as a comparison day; the historical load data of the comparison day and the historical load data of a certain historical day are divided to obtain the operation result. If the operation result is between 0.5 and 1.5 If not, the certain load influence factor is determined as the main influence factor of the load in the area to be predicted, otherwise, the certain load influence factor is excluded.

式中：cosθ为相似度，区间[0,1]；A为待预测区域预测日的负荷影响特征向量；B为历史日的负荷影响特征向量；若cosθ值在0.8以上，则将该历史日作为相似日。Specifically, the selection of the similar day is determined according to the main influencing factors of the regional load to be predicted, which is calculated by the similarity measurement model, and its specific formula is as follows:

In the formula: cosθ is the similarity, the interval is [0, 1]; A is the load impact eigenvector of the forecast day in the area to be predicted; B is the load impact eigenvector of the historical day; if the cosθ value is above 0.8, the historical day as a similar day.

Step 3: Determine the load data of electric vehicles, obtain relevant data of electric vehicles, and determine the daily load data of electric vehicles in the area to be predicted according to them;

Specifically, the electric vehicle-related data includes the daily driving electric vehicle flow and the daily driving electric vehicle charging times; the electric vehicle load data is calculated based on the daily driving electric vehicle flow and the daily driving electric vehicle charging times, and based on the electric vehicle power, to obtain Electric vehicle load data.

Step 4: Extracting the training set, extracting the load impact data and load data of similar historical days and the daily load data of electric vehicles as training set, test set and verification set;

Step 5: Build a regional short-term load forecasting model, construct a deep network model, and use the training set as the input data of the deep network model for training to form a regional short-term load forecasting model, and verify it through the verification test set. If the regional short-term load forecasting model error If it is within the threshold, output the model; otherwise, skip to step 4 to re-extract the training set; specifically, the threshold is within ±5%;

Step 6: Load forecasting, collecting and inputting the relevant data of the forecast day in the area to be forecasted, and forecasting based on the regional short-term load forecasting model, obtaining the regional short-term load forecasting result, and outputting it.

Referring to FIG. 2, the present embodiment discloses a power distribution network load forecasting system, including a data extraction module, a data preprocessing module, a deep learning module, a data acquisition module, a load forecasting module and an output display module;

The data extraction module is used to obtain the historical power data and electric vehicle related data in the power database and the electric vehicle database, and based on this, extract the load impact data and load data of similar historical days and the electric vehicle daily load data;

The data preprocessing module is used to preprocess the load impact data and load data of similar historical days and the daily load data of electric vehicles;

Specifically, the preprocessing includes but is not limited to filling missing data, correcting noise data, data smoothing, data normalization and data format unification;

The deep learning module is used to use the preprocessed load impact data and load data of similar historical days and electric vehicle daily load data as a training set, and learn based on the deep learning network to generate a regional short-term load forecasting model;

The data collection module is used to collect the relevant data of the forecast day in the area to be forecasted;

Specifically, the relevant data includes each load impact data on the forecast day of the area to be forecasted;

The load forecasting module is used to input the relevant data of the forecast date of the area to be forecasted, and make predictions based on the regional short-term load forecasting model to generate the regional short-term load forecasting results;

The output display module is used to output the short-term load prediction results of the display area, which provides an important reference for the reasonable control of the start and stop of the generator set.

The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited to this. The equivalent replacement or change of the inventive concept thereof shall be included within the protection scope of the present invention.

Claims (6)

1. A power distribution network load prediction method is characterized by comprising the following specific steps:

the method comprises the following steps: acquiring historical power data and acquiring historical power data of an area to be predicted; the historical power data comprises historical load impact data and historical load data;

step two: determining load influence factors and selecting similar days, determining main influence factors of the load of the area to be predicted according to the change condition of the historical load data, and extracting the similar historical days of the prediction days of the area to be predicted through a similarity measurement model according to the main influence factors;

step three: determining electric vehicle load data, acquiring electric vehicle related data, and determining electric vehicle daily load data of an area to be predicted according to the electric vehicle related data;

step four: extracting a training set, namely extracting the load influence data and the load data of the similar historical days and the daily load data of the electric vehicle as the training set, the test set and the verification set;

step five: constructing a regional short-term load prediction model, constructing a deep network model, training the training set as input data of the deep network model to form a regional short-term load prediction model, verifying through a verification test set, outputting the model if the regional short-term load prediction model error is within a threshold value, otherwise, skipping to the fourth step to re-extract the training set;

step six: and load prediction, namely acquiring and inputting relevant data of a prediction day of a region to be predicted, predicting based on the region short-term load prediction model to obtain a region short-term load prediction result, and outputting the region short-term load prediction result.

2. The method according to claim 1, wherein the historical load influence data comprises specific influence factor data and random influence factor data; the specific influence factor data includes electricity price, time and weather; the random influence factor data is specifically determined according to the region to which the random influence factor data belongs.

3. The method for predicting the load of the power distribution network according to claim 1, wherein the specific process for determining the load influence factors in the second step is as follows:

s1: selecting a certain historical day of a certain load influence factor one by one, and extracting historical load data of the certain historical day;

s2: acquiring a history day which is the same as or similar to other influence factors of the history day and different from the certain load influence factor as a comparison day;

s3: and performing an addition operation on the historical load data of the comparison day and the historical load data of the certain historical day to obtain an operation result, if the operation result is 0.5-1.5, determining that the certain load influence factor is a main influence factor of the load of the area to be predicted, and otherwise, rejecting the certain load influence factor.

4. The method for predicting the load of the power distribution network according to claim 1, wherein the similar day in the second step is determined according to main influence factors of the load of the area to be predicted, and the main influence factors are calculated through a similarity measurement model, and the specific formula is as follows:

in the formula: cos θ is the similarity, interval [0,1 ]; a is a load influence characteristic vector of a prediction day of an area to be predicted; b is a load influence characteristic vector of a historical day;

the similar day is selected from historical days with cos theta values above 0.8.

5. The method according to claim 1, wherein the electric vehicle related data in step three includes a daily electric vehicle flow and a daily electric vehicle charging number.

6. A power distribution network load prediction system is characterized by comprising a data extraction module, a data preprocessing module, a deep learning module, a data acquisition module, a load prediction module and an output display module;

the data extraction module is used for acquiring historical electric power data and electric automobile related data in an electric power database and an electric automobile database, and extracting load influence data and load data of similar historical days and daily load data of the electric automobile based on the historical electric power data and the electric automobile related data;

the data preprocessing module is used for preprocessing the load influence data and the load data of the similar historical days and the daily load data of the electric automobile;

the deep learning module is used for taking the preprocessed load influence data and load data of similar historical days and daily load data of the electric vehicle as a training set, learning based on a deep learning network and generating a regional short-term load prediction model;

the data acquisition module is used for acquiring relevant data of a prediction day of an area to be predicted;

the load prediction module is used for inputting relevant data of the prediction day of the area to be predicted, predicting the data based on the area short-term load prediction model and generating an area short-term load prediction result;

and the output display module is used for outputting and displaying the regional short-term load prediction result so as to provide an important reference basis for reasonably controlling the start and stop of the generator set.

Images