CN113723717B - System day-ahead short-term load forecasting method, apparatus, device and readable storage medium - Google Patents

Abstract

The invention relates to a method, device, equipment and readable storage medium for short-term load forecasting in the day before a system. The method includes: collecting historical data, and preprocessing the historical data; using the preprocessed historical data to construct a training sample set; The training sample set is used to train the pre-established XGBoost multi-objective regression model, and the trained XGBoost multi-objective regression model is obtained; the predicted sample features are generated; the predicted sample features are input into the trained XGBoost multi-objective regression model to obtain the predicted short-term load . The technical solution provided by this application not only improves the efficiency of model training, deployment and prediction, but also improves the accuracy of short-term load prediction.

Description

System day-ahead short-term load forecasting method, apparatus, device and readable storage medium

technical field

The invention belongs to the technical field of energy monitoring equipment, and in particular relates to a method, device, equipment and readable storage medium for short-term load forecasting in the day-ahead system.

Background technique

The production, transmission, distribution and use of electric energy are carried out at the same time. Since electric energy cannot be stored in a large amount, the power supply and demand must be balanced in real time. Power dispatching needs to formulate the start and stop of units and the maintenance of power equipment in advance according to the future load demand. Therefore, accurate load forecasting is of great significance to the dispatching operation of the power grid, and the level of load forecasting directly affects the economic and social benefits of the power system.

The short-term load forecast of the system before the day generally refers to the forecast of the load demand at 96 times every 15 minutes in the next day in advance, which is a typical supervised regression problem in the field of machine learning. Traditional load forecasting usually disassembles the problem modeling into a regression problem of a single target by time. It is believed that the time to be predicted is related to the load value and meteorological factors at the same time in the historical day, and 96 regression models are trained using historical data. To realize the prediction of the load value of the next 96 moments. This method often ignores the continuous characteristics of the load curve in a long time range, which is easy to cause under-fitting of the model. To train 96 models at the same time, it is necessary to perform feature engineering and sample engineering for each model, which is often time-consuming and is not conducive to training complex models with a large amount of data.

SUMMARY OF THE INVENTION

In view of this, the object of the present invention is to overcome the deficiencies in the prior art, and provide a method, device, equipment and readable storage medium for short-term load forecasting in the prior art, to solve the underfitting of models in the prior art, and to train 96 The model is time-consuming, which is not conducive to the problem of training complex models with large amounts of data.

According to a first aspect of the embodiments of the present application, there is provided a method for predicting a system day-ahead short-term load, the method comprising:

collecting historical data, and preprocessing the historical data;

Use the preprocessed historical data to construct a training sample set;

Use the training sample set to train the pre-established XGBoost multi-objective regression model to obtain the trained XGBoost multi-objective regression model;

Generate predicted sample features;

The predicted sample features are input into the trained XGBoost multi-objective regression model to obtain the predicted short-term load.

Further, the historical data includes: historical short-term load value, historical meteorological value and historical load variation;

The meteorological indicators in the historical meteorological values include: temperature, humidity, rainfall, wind and cloudiness.

Further, the preprocessing of the historical data includes:

An interpolation method is used to fill in the missing values in the historical data, and the historical data after filling the missing values is normalized to obtain the preprocessed historical data.

Further, the use of the preprocessed historical data to construct a training sample set includes:

The time of the preprocessed historical data is divided into 96 time points, and a training sample set is constructed by using the preprocessed historical data of the 96 time points.

Further, the described training sample set is used to train the pre-established XGBoost multi-objective regression model, and the trained XGBoost multi-objective regression model is obtained, including:

dividing the training sample set into a training set and a validation set;

Use the training set to train the pre-established XGBoost multi-objective regression model, until when the pre-established XGBoost multi-objective regression model is verified using the verification set, the predicted historical load value obtained is different from the actual historical load value. When the error of the load value is less than the first threshold, the training ends, and the trained XGBoost multi-objective regression model is obtained.

Further, the predicted sample features include:

、待预测时刻前k~k-M个时刻点的气象值

、待预测时刻前k~k-M个时刻点的负荷变化量

、月份特征、日期特征、日期类型特征、待预测开始时刻所在小时特征和待预测开始时刻分钟特征；Load value at k~kM time points before the time to be predicted

, the meteorological value of k~kM time points before the forecast time

, the load variation at k~kM time points before the prediction time

, month feature, date feature, date type feature, hour feature of the start time to be predicted, and minute feature of the start time to be predicted;

，以此类推。Among them, k is the time point to be predicted; M is an adjustable hyperparameter, and M is a positive integer value; the load change at kM time points before the time to be predicted

, and so on.

Further, inputting the predicted sample features into the trained XGBoost multi-objective regression model to obtain the predicted short-term load, including:

The objective function of the trained XGBoost multi-objective regression model is determined as follows:

，n为预测样本特征的总数量，

，m为待预测的时刻点的总数量；

为第i个预测样本特征的特征向量，

为第i个预测样本特征的第k个待预测的时刻点的实际负荷值，

为第t-1轮迭代时对第i个预测样本特征第k个待预测的时刻点的预测负荷值，

为预测第k个待预测的时刻点的负荷值时第t轮迭代中增加的新模型，

为预测第k个待预测的时刻点的负荷值时第t轮迭代增加的新模型的复杂度，

为损失函数；In the above formula,

, n is the total number of predicted sample features,

, m is the total number of time points to be predicted;

is the feature vector of the i-th predicted sample feature,

is the actual load value of the kth to-be-predicted moment of the i-th prediction sample feature,

is the predicted load value of the k-th time point to be predicted for the i-th predicted sample feature during the t-1 round of iteration,

is the new model added in the t-th iteration when predicting the load value at the k-th time point to be predicted,

is the complexity of the new model added by the t-th iteration when predicting the load value at the k-th time point to be predicted,

is the loss function;

小于等于第二阈值或迭代轮数t等于第三阈值时，输出预测的所有时刻点的负荷，所述预测的所有时刻点的负荷为预测的短期负荷。The predicted sample feature is input to the objective function of the XGBoost multi-objective regression model after the training as an independent variable, when

When it is less than or equal to the second threshold or when the number of iteration rounds t is equal to the third threshold, the predicted loads at all time points are output, and the predicted loads at all time points are predicted short-term loads.

According to a second aspect of the embodiments of the present application, a system day-ahead short-term load forecasting device is provided, the device comprising:

a preprocessing module for collecting historical data and preprocessing the historical data;

The building module is used to construct a training sample set using the preprocessed historical data;

A training module for training the pre-established XGBoost multi-objective regression model using the training sample set to obtain the trained XGBoost multi-objective regression model;

A generation module is used to generate predicted sample features;

A prediction module, configured to input the predicted sample features into the trained XGBoost multi-objective regression model to obtain a predicted short-term load.

According to a third aspect of the embodiments of the present application, a system day-ahead short-term load forecasting device is provided, including:

a memory on which an executable program is stored;

The processor is configured to execute the executable program in the memory, so as to implement the steps of the above-mentioned method for short-term load forecasting for the day ahead of the system.

According to a fourth aspect of the embodiments of the present application, a readable storage medium is provided, on which an executable program is stored, and when the executable program is executed by a processor, the above-mentioned method for predicting a system day-ahead short-term load is implemented. step.

The present invention adopts the above technical solutions, and the beneficial effects that can be achieved include: by collecting historical data, preprocessing the historical data, using the preprocessed historical data to construct a training sample set, and using the training sample set to pre-establish a The XGBoost multi-objective regression model is trained, the XGBoost multi-objective regression model after training is obtained, the predicted sample features are generated, the predicted sample features are input into the XGBoost multi-objective regression model after the training, and the predicted short-term load is obtained, not only It improves the efficiency of model training, deployment and prediction, and also improves the accuracy of short-term load prediction.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that are used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

FIG. 1 is a flowchart of a method for predicting a system day-ahead short-term load according to an exemplary embodiment;

FIG. 2 is a structural block diagram of a system day-ahead short-term load forecasting apparatus according to an exemplary embodiment.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

Fig. 1 is a flow chart of a method for predicting a system day-ahead short-term load according to an exemplary embodiment. As shown in Fig. 1 , the method can be used in a terminal, but is not limited to, including the following steps:

Step 101: collect historical data, and preprocess the historical data;

Step 102: use the preprocessed historical data to construct a training sample set;

Step 103: using the training sample set to train the pre-established XGBoost multi-objective regression model to obtain a trained XGBoost multi-objective regression model;

Step 104: generating predicted sample features;

Step 105: Input the predicted sample features into the trained XGBoost multi-objective regression model to obtain the predicted short-term load.

The embodiment of the present invention provides a method for short-term load forecasting for the day ahead of the system. Different from the traditional method, the present application defines the short-term load forecasting problem for the day ahead as a multi-objective learning task. Prediction of load values. Specifically, by collecting historical data and preprocessing the historical data, using the preprocessed historical data to construct a training sample set, and using the training sample set to train the pre-established XGBoost multi-objective regression model, the trained XGBoost multi-objective regression model is obtained. The target regression model realizes the training and prediction of the multi-target regression model; by generating the predicted sample features, the predicted sample features are input to the XGBoost multi-target regression model after training, and the predicted short-term load is obtained, which not only improves the model training and deployment. and forecasting efficiency, and also improve the accuracy of short-term load forecasting.

Further, historical data, including: historical short-term load value, historical meteorological value and historical load variation;

The meteorological indicators in the historical meteorological values include: temperature, humidity, rainfall, wind and cloudiness.

Further, in step 101, the historical data is preprocessed, including:

The interpolation method is used to fill in the missing values in the historical data, and the historical data after filling the missing values is normalized to obtain the preprocessed historical data.

It should be noted that the methods of "using interpolation to fill missing values in historical data" and "normalizing historical data after filling missing values" involved in the embodiments of the present invention are well known to those skilled in the art. Therefore, its specific implementation will not be described too much.

Further, step 102 includes:

The time of the preprocessed historical data is divided into 96 time points, and a training sample set is constructed by using the preprocessed historical data of the 96 time points.

It is understandable that the time of day is divided into 96 time points, that is, each time point is spaced 15 minutes apart. Compared with the traditional method, such a sample construction method can obtain nearly 96 times the sample size, which can effectively support the sufficient update of model parameters in a large-scale feature space and avoid model underfitting.

It should be noted that for the multi-objective regression task, the same model and features are used for modeling each label in the sample, so compared with the traditional method, the model training complexity will be greatly reduced, and it is beneficial to use A large dataset is used for model training.

Further, step 103 includes:

Divide the training sample set into training set and validation set;

Use the training set to train the pre-established XGBoost multi-objective regression model, until when the validation set is used to verify the pre-established XGBoost multi-objective regression model, the error between the predicted historical load value and the actual historical load value is less than the first When a threshold is reached, the training ends, and the trained XGBoost multi-objective regression model is obtained.

Further, predict the sample characteristics, including:

、待预测时刻前k~k-M个时刻点的气象值

、待预测时刻前k~k-M个时刻点的负荷变化量

、月份特征、日期特征、日期类型特征、待预测开始时刻所在小时特征和待预测开始时刻分钟特征；Load value at k~kM time points before the time to be predicted

, the meteorological value of k~kM time points before the forecast time

, the load variation at k~kM time points before the prediction time

, month feature, date feature, date type feature, hour feature of the start time to be predicted, and minute feature of the start time to be predicted;

，以此类推。Among them, k is the time point to be predicted; M is an adjustable hyperparameter, and M is a positive integer value; the load change at kM time points before the time to be predicted

, and so on.

Specifically, the month feature refers to 12 months of a year, the date feature refers to each day in each month, and the date type feature refers to Monday to Sunday.

For example, according to the above content, one day is divided into 96 time points, each time point is 15 minutes apart, then the time points for predicting the sample features are also divided into 96 time points, and 0:00 in the morning is the first time point. . Suppose you need to predict the load value at 1 am tomorrow morning, then k is 5. Suppose M is 4, then k-M=5-4=1, k-M+1=5-4+1=2, k-M+2=5-4+2=3, k-M+3=5- 4+3=4.

It should be noted that the native XGBoost multi-objective regression model of dmlc cannot support the training and prediction of the multi-objective regression model. As shown below, the loss function defined by the XGBoost model of the single-objective regression task is:

，n为样本的总数量；

为第i个样本label，

第i个样本的特征向量，

为模型在第t-1轮迭代时对第i个样本的预测值，

为第t轮迭代要新增的模型，

为第t轮迭代要新增的模型的复杂度，

为损失函数。In the above formula,

, n is the total number of samples;

is the i-th sample label,

The eigenvector of the ith sample,

is the predicted value of the model for the ith sample at the t-1th iteration,

The model to be added for the t-th iteration,

is the complexity of the model to be added for the t-th iteration,

is the loss function.

The technical solution provided by the embodiment of the present invention modifies the loss function of the XGBoost multi-objective regression model so that it can support the multi-objective regression model. Further, step 105 includes:

Determine the objective function of the trained XGBoost multi-objective regression model as follows:

，n为预测样本特征的总数量，

，m为待预测的时刻点的总数量；

为第i个预测样本特征的特征向量，

为第i个预测样本特征的第k个待预测的时刻点的实际负荷值，

为第t-1轮迭代时对第i个预测样本特征第k个待预测的时刻点的预测负荷值，

为预测第k个待预测的时刻点的负荷值时第t轮迭代中增加的新模型，

为预测第k个待预测的时刻点的负荷值时第t轮迭代增加的新模型的复杂度，

为损失函数；In the above formula,

, n is the total number of predicted sample features,

, m is the total number of time points to be predicted;

is the feature vector of the i-th predicted sample feature,

is the actual load value of the kth to-be-predicted moment of the i-th prediction sample feature,

is the predicted load value of the k-th time point to be predicted for the i-th predicted sample feature during the t-1 round of iteration,

is the new model added in the t-th iteration when predicting the load value at the k-th time point to be predicted,

is the complexity of the new model added by the t-th iteration when predicting the load value at the k-th time point to be predicted,

is the loss function;

小于等于第二阈值或迭代轮数t等于第三阈值时，输出预测的所有时刻点的负荷，预测的所有时刻点的负荷为预测的短期负荷。Input the predicted sample features as independent variables to the objective function of the trained XGBoost multi-objective regression model, when

When it is less than or equal to the second threshold or when the number of iteration rounds t is equal to the third threshold, the predicted loads at all time points are output, and the predicted loads at all time points are predicted short-term loads.

It can be understood that the method provided by the embodiment of the present invention can reflect the power load as a continuous time series, and the load value at the moment to be predicted is not only related to the historical value at the same time, but also is related to the load value of the nearest point that can be obtained. Sex is also great.

Experiments have shown that the method for short-term load forecasting in the day-ahead system provided by the embodiment of the present invention can achieve a statistical average RMPSE accuracy of 97.29% after a period of stable operation, which is significantly improved from the 94.43% accuracy of the traditional method deployed in the early stage. . In terms of operating efficiency, the new method architecture disassembles the model training process and the prediction process. The training process is a scheduled task, and the prediction process is a calling task. Compared with the traditional method, the training process and the prediction process are coupled together. , the prediction efficiency has been greatly improved, from the previous time-consuming nearly 30s to 400ms.

The embodiment of the present invention provides a method for short-term load forecasting for the day ahead of the system. Different from the traditional method, the present application defines the short-term load forecasting problem for the day ahead as a multi-objective learning task. Prediction of load values. Specifically, by collecting historical data and preprocessing the historical data, using the preprocessed historical data to construct a training sample set, and using the training sample set to train the pre-established XGBoost multi-objective regression model, the trained XGBoost multi-objective regression model is obtained. The target regression model realizes the training and prediction of the multi-target regression model; by generating the predicted sample features, the predicted sample features are input to the XGBoost multi-target regression model after training, and the predicted short-term load is obtained, which not only improves the model training and deployment. and forecasting efficiency, and also improve the accuracy of short-term load forecasting, compared with traditional methods.

An embodiment of the present invention also provides a system day-ahead short-term load forecasting device, as shown in FIG. 2 , the device includes:

The preprocessing module is used to collect historical data and preprocess the historical data;

The building module is used to construct a training sample set using the preprocessed historical data;

The training module is used to train the pre-established XGBoost multi-objective regression model by using the training sample set to obtain the trained XGBoost multi-objective regression model;

A generation module is used to generate predicted sample features;

The prediction module is used to input the predicted sample features into the trained XGBoost multi-objective regression model to obtain the predicted short-term load.

Further, historical data, including: historical short-term load value, historical meteorological value and historical load variation;

The meteorological indicators in the historical meteorological values include: temperature, humidity, rainfall, wind and cloudiness.

Further, the preprocessing module is specifically used for:

The interpolation method is used to fill in the missing values in the historical data, and the historical data after filling the missing values is normalized to obtain the preprocessed historical data.

Further, building modules, specifically for:

The time point length of the preprocessed historical data is set to 96, and a training sample set is constructed by using the preprocessed historical data of the 96 time points.

Further, the training module is specifically used for:

Divide the training sample set into training set and validation set;

Use the training set to train the pre-established XGBoost multi-objective regression model, until when the validation set is used to verify the pre-established XGBoost multi-objective regression model, the error between the predicted historical load value and the actual historical load value is less than the first When a threshold is reached, the training ends, and the trained XGBoost multi-objective regression model is obtained.

Further, predict the sample characteristics, including:

、待预测时刻前k~k-M个时刻点的气象值

、待预测时刻前k~k-M个时刻点的负荷变化量

、月份特征、日期特征、日期类型特征、待预测开始时刻所在小时特征和待预测开始时刻分钟特征；Load value at k~kM time points before the time to be predicted

, the meteorological value of k~kM time points before the forecast time

, the load variation at k~kM time points before the prediction time

, month feature, date feature, date type feature, hour feature of the start time to be predicted, and minute feature of the start time to be predicted;

，以此类推。Among them, k is the time point to be predicted; M is an adjustable hyperparameter, and M is a positive integer value; the load change at kM time points before the time to be predicted

, and so on.

Further, the prediction module is specifically used for:

Determine the objective function of the trained XGBoost multi-objective regression model as follows:

，n为预测样本特征的总数量，

，m为待预测的时刻点的总数量；

为第i个预测样本特征的特征向量，

为第i个预测样本特征的第k个待预测的时刻点的实际负荷值，

为第t-1轮迭代时对第i个预测样本特征第k个待预测的时刻点的预测负荷值，

为预测第k个待预测的时刻点的负荷值时第t轮迭代中增加的新模型，

为预测第k个待预测的时刻点的负荷值时第t轮迭代增加的新模型的复杂度，

为损失函数；In the above formula,

, n is the total number of predicted sample features,

, m is the total number of time points to be predicted;

is the feature vector of the i-th predicted sample feature,

is the actual load value of the kth to-be-predicted moment of the i-th prediction sample feature,

is the predicted load value of the k-th time point to be predicted for the i-th predicted sample feature during the t-1 round of iteration,

is the new model added in the t-th iteration when predicting the load value at the k-th time point to be predicted,

is the complexity of the new model added by the t-th iteration when predicting the load value at the k-th time point to be predicted,

is the loss function;

小于等于第二阈值或迭代轮数t等于第三阈值时，输出预测的所有时刻点的负荷，预测的所有时刻点的负荷为预测的短期负荷。Input the predicted sample features as independent variables to the objective function of the trained XGBoost multi-objective regression model, when

When it is less than or equal to the second threshold or when the number of iteration rounds t is equal to the third threshold, the predicted loads at all time points are output, and the predicted loads at all time points are predicted short-term loads.

The embodiment of the present invention provides a method for short-term load forecasting for the day ahead of the system. Different from the traditional method, the present application defines the short-term load forecasting problem for the day ahead as a multi-objective learning task. Prediction of load values. Specifically, the historical data is collected through the preprocessing module, and the historical data is preprocessed. The building module uses the preprocessed historical data to construct a training sample set, and the training module uses the training sample set to train the pre-established XGBoost multi-objective regression model. , the trained XGBoost multi-objective regression model is obtained, which realizes the training and prediction of the multi-objective regression model; the predicted sample features are generated by the generation module, and the prediction module inputs the predicted sample features into the trained XGBoost multi-objective regression model to obtain the prediction. It not only improves the efficiency of model training, deployment and prediction, but also improves the accuracy of short-term load prediction, compared with traditional methods.

It can be understood that the above-mentioned apparatus embodiments correspond to the above-mentioned method embodiments, and the corresponding specific contents can be referred to each other, which will not be repeated here.

The embodiment of the present invention also provides a system day-ahead short-term load forecasting device, including:

a memory on which an executable program is stored;

The processor is configured to execute the executable program in the memory, so as to implement the steps of the method for predicting the short-term load ahead of the system day provided by the foregoing embodiment.

Embodiments of the present invention further provide a readable storage medium on which an executable program is stored, and when the executable program is executed by a processor, implements the steps of the method for predicting system day-ahead short-term load provided by the foregoing embodiments.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media having computer-usable program code embodied therein, including but not limited to disk storage, optical storage, and the like.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in one or more of the flowcharts and/or one or more blocks of the block diagrams.

These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer readable memory result in an article of manufacture comprising the method of the instructions, the instructions A method implements the functionality specified in a flow or flows of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in one or more of the flowcharts and/or one or more blocks of the block diagrams.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention. should be included within the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims (9)

1. A method for predicting a short-term load of a power system in the day ahead, the method comprising:

acquiring historical data and preprocessing the historical data;

constructing a training sample set by utilizing the preprocessed historical data;

training a pre-established XGboost multi-target regression model by using the training sample set to obtain a trained XGboost multi-target regression model;

generating predicted sample features;

inputting the predicted sample characteristics into the trained XGboost multi-target regression model to obtain a predicted short-term load;

inputting the predicted sample characteristics into the trained XGboost multi-objective regression model to obtain a predicted short-term load, wherein the predicted short-term load comprises the following steps:

determining an objective function of the trained XGboost multi-objective regression model according to the following formula:

in the above formula, the first and second carbon atoms are,

n is the total number of predicted sample features,

m is the total number of time points to be predicted;

for the feature vector of the ith prediction sample feature,

for the actual load value of the power grid at the kth time point to be predicted of the ith prediction sample characteristic,

predicting a load value of the power grid at a kth time point to be predicted of the ith prediction sample characteristic during the t-1 th iteration,

to predict the grid load value at the kth point of time to be predicted, a new model is added in the t-th iteration,

the complexity of a new model which is added in the t-th iteration when the grid load value of the kth time point to be predicted is predicted,

is a loss function;

inputting the predicted sample characteristics as independent variables into the objective function of the trained XGboost multi-objective regression model

Less than or equal to a second threshold or number of iterations t, etcAnd outputting the predicted loads of all the time points at the third threshold, wherein the predicted loads of all the time points are predicted short-term loads.

2. The method of claim 1, wherein the historical data comprises: historical short-term power grid load values, historical meteorological values and historical load variation;

the meteorological indexes in the historical meteorological values comprise: temperature, humidity, rainfall, wind, and cloud cover.

3. The method of claim 1, wherein the pre-processing the historical data comprises:

and utilizing an interpolation method to fill missing values in the historical data, and carrying out normalization processing on the historical data after the missing values are filled to obtain the preprocessed historical data.

4. The method of claim 1, wherein constructing a training sample set using the preprocessed historical data comprises:

and dividing the time of the preprocessed historical data into 96 time points, and constructing a training sample set by using the preprocessed historical data of the 96 time points.

5. The method as claimed in claim 1, wherein the training of the pre-established XGBoost multi-objective regression model by using the training sample set to obtain the trained XGBoost multi-objective regression model comprises:

dividing the training sample set into a training set and a verification set;

and training the pre-established XGboost multi-target regression model by using the training set until the error between the obtained predicted historical power grid load value and the actual historical power grid load value is smaller than a first threshold value when the pre-established XGboost multi-target regression model is verified by using the verification set, finishing training and obtaining the trained XGboost multi-target regression model.

6. The method of claim 1, wherein predicting the sample features comprises:

the power grid load value of k-M time points before the time to be predicted

And the meteorological values of k-M time points before the time to be predicted

Load variation of k-M time points before the time to be predicted

A month feature, a date type feature, an hour feature where the start time to be predicted is located, and a minute feature of the start time to be predicted;

wherein k is a time point to be predicted; m is a tunable hyperparameter, and M is a positive integer value; load variation of k-M time points before time to be predicted

And so on.

7. An apparatus for predicting a short-term load before a day in a power system, the apparatus comprising:

the preprocessing module is used for acquiring historical data and preprocessing the historical data;

the construction module is used for constructing a training sample set by utilizing the preprocessed historical data;

the training module is used for training a pre-established XGboost multi-target regression model by using the training sample set to obtain a trained XGboost multi-target regression model;

a generation module for generating predicted sample features;

the prediction module is used for inputting the characteristics of the prediction samples into the trained XGboost multi-target regression model to obtain predicted short-term load;

the prediction module is specifically configured to:

determining an objective function of the trained XGboost multi-objective regression model according to the following formula:

in the above formula, the first and second carbon atoms are,

n is the total number of predicted sample features,

m is the total number of time points to be predicted;

for the feature vector of the ith prediction sample feature,

for the actual load value of the power grid at the kth time point to be predicted of the ith prediction sample characteristic,

predicting a load value of the power grid at a kth time point to be predicted of the ith prediction sample characteristic during the t-1 th iteration,

to predict the grid load value at the kth point of time to be predicted, a new model is added in the t-th iteration,

for predicting the grid load of the kth time point to be predictedThe complexity of the new model added in the t-th iteration at load,

is a loss function;

inputting the predicted sample characteristics as independent variables into the target function of the trained XGboost multi-target regression model

And when the load is less than or equal to the second threshold or the iteration round number t is equal to the third threshold, outputting the predicted load of all the time points, wherein the predicted load of all the time points is the predicted short-term load.

8. A power system day-ahead short-term load prediction apparatus, comprising:

a memory having an executable program stored thereon;

a processor for executing the executable program in the memory to implement the steps of the method of any one of claims 1-6.

9. A readable storage medium having stored thereon an executable program, wherein the executable program, when executed by a processor, performs the steps of the method of any one of claims 1-6.

Images