Meteorological feature extraction method and system for bus load prediction, medium and processor

Abstract

The invention relates to the technical field of feature extraction, and discloses a meteorological feature extraction method and system for bus load prediction, a medium and a processor. The method comprises the steps of obtaining meteorological data of candidate meteorological stations, performing weighted calculation on an average absolute percentage error of meteorological total elements of the candidate meteorological stations according to the meteorological data, and screening a first target meteorological station with a preset proportion according to a result of the average absolute percentage error; obtaining load data of a selected bus, analyzing correlation between meteorological elements of each meteorological station in the first target meteorological stations and the load data of the selected bus, and screening a second target meteorological station in a preset proportion according to a correlation result; and taking the second target meteorological station as the input of a neural network, and generating the final meteorological characteristics of the selected bus load prediction. According to the invention, combination of meteorological data and bus load prediction is realized, the prediction precision of the bus load is improved, and safe and stable operation of a power system is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of feature extraction, and in particular to a meteorological feature extraction method, system, medium and processor for bus load prediction. Background Art

[0002] The function of the power system is to convert energy from a naturally existing form into electrical energy, thereby achieving efficient transmission of electrical energy, so that the user group can use efficient, economical, and low-risk electrical energy. However, electrical energy is not easy to store in large quantities, and in practice, the power consumption of users is constantly changing. In order to ensure the dynamic balance of power generation, transmission, and consumption in the power system, the power demand must be predicted in advance. Bus load forecasting, as an important part of power load forecasting, is of great significance to the stable operation and dispatching decision-making of the power system.

[0003] Factors such as temperature, humidity, rainfall, wind speed, and wind direction in the meteorological environment are important factors affecting bus load changes. The quality and accuracy of meteorological forecast data have a great impact on bus load forecasting. Analyzing and mining meteorological data to provide more applicable meteorological forecast information for bus load forecasting is an important method to improve the accuracy of bus load. Bus load forecasting is a key link in the operation of the power system. Strengthening the combination of meteorological data and bus load forecasting and improving forecast accuracy are of great significance to ensuring the safe and stable operation of the power system.

[0004] In view of this, a meteorological feature extraction method, system, medium and processor for bus load forecasting are needed. Summary of the invention

[0005] In view of the problem of how to combine meteorological data with bus load prediction in the prior art, the present invention provides a meteorological feature extraction method, system, medium and processor for bus load prediction, which can combine meteorological data with bus load prediction and improve the prediction accuracy of bus load. The specific technical solution is as follows:

[0006] A meteorological feature extraction method for bus load forecasting, comprising:

[0007] Acquire meteorological data of candidate meteorological sites, calculate the average absolute percentage error of all meteorological elements of candidate meteorological sites by weight according to the meteorological data, and select a preset proportion of first target meteorological sites according to the result of the average absolute percentage error;

[0008] Obtaining the load data of the selected bus, analyzing the correlation between the meteorological elements of each meteorological station in the first target meteorological station and the load data of the selected bus, and selecting a preset proportion of second target meteorological stations according to the correlation result;

[0009] The second target meteorological site is used as an input of the neural network to generate final meteorological characteristics for the selected bus load forecast.

[0010] Preferably, the step of acquiring meteorological data of the candidate meteorological site and weightedly calculating the mean absolute percentage error of all meteorological elements of the candidate meteorological site according to the meteorological data comprises:

[0011] Obtaining the actual data and forecast data of various meteorological elements of the candidate meteorological sites, and calculating the average absolute percentage errors of various meteorological elements within various preset periods based on the acquired data;

[0012] According to the forecast accuracy of meteorological elements in various preset periods and the preset period weight coefficient, the average absolute percentage error of various meteorological elements is weightedly calculated;

[0013] According to the average absolute percentage error of various meteorological elements and the preset element weight coefficient, the average absolute percentage error of all meteorological elements of the candidate meteorological station is weightedly calculated.

[0014] Preferably, the step of screening a preset proportion of first target meteorological stations according to the result of the mean absolute percentage error comprises:

[0015] The results of the mean absolute percentage errors are sorted from small to large, and a preset proportion of the first target meteorological stations in the top sorting are selected.

[0016] Preferably, the step of acquiring the load data of the selected bus, analyzing the correlation between the meteorological elements of each meteorological station in the first target meteorological station and the load data of the selected bus, and selecting a preset proportion of second target meteorological stations according to the result of the correlation comprises:

[0017] Obtain the location of the substation where the selected bus is located, and calculate the distance between each meteorological station in the first target meteorological station and the substation where the selected bus is located to obtain a distance result;

[0018] Selecting a preset number of third target meteorological stations according to the distance result;

[0019] Combining the meteorological element forecast data of the third target meteorological station into a meteorological characteristic matrix;

[0020] The meteorological characteristic matrix and the load data of the selected bus are used as the input and output of the random forest algorithm for training respectively, and the importance score results of each meteorological characteristic are obtained;

[0021] According to the importance score result, meteorological features of a preset proportion are screened out as the second target meteorological site of the selected bus correlation.

[0022] Preferably, the meteorological characteristic matrix and the load data of the selected bus are used as the input and output of the random forest algorithm for training respectively, and the importance score results of each meteorological characteristic are obtained, including:

[0023] Aligning the timestamps of the meteorological characteristic matrix and the load data of the selected bus;

[0024] Preprocessing the meteorological characteristic matrix;

[0025] The processed meteorological characteristic matrix is ​​used as input and the load data of the selected bus is used as output to train the random forest regression model;

[0026] The importance scores of each meteorological feature were extracted from the random forest regression model after training.

[0027] A meteorological feature extraction system for bus load forecasting, applied to the aforementioned meteorological feature extraction method for bus load forecasting, comprises:

[0028] A first target meteorological site acquisition unit is used to acquire meteorological data of candidate meteorological sites, calculate the average absolute percentage error of all meteorological elements of candidate meteorological sites based on the meteorological data, and select a preset proportion of first target meteorological sites based on the result of the average absolute percentage error;

[0029] A second target meteorological site acquisition unit, used to acquire load data of a selected bus, analyze the correlation between meteorological elements of each meteorological site in the first target meteorological site and the load data of the selected bus, and select a preset proportion of second target meteorological sites according to the correlation result;

[0030] The final meteorological feature selection unit is used to use the second target meteorological site as an input of the neural network to generate the final meteorological features of the selected bus load forecast.

[0031] A computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to use the aforementioned meteorological feature extraction method for bus load prediction.

[0032] A processor is used to run a program, wherein the program executes the aforementioned meteorological feature extraction method for bus load prediction when running.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] A meteorological feature extraction method for bus load prediction of the present invention obtains meteorological data of candidate meteorological sites, calculates the average absolute percentage error of all meteorological elements of candidate meteorological sites based on the meteorological data, and selects a preset proportion of first target meteorological sites based on the result of the average absolute percentage error; obtains the load data of the selected bus, and analyzes the correlation between the meteorological elements of each meteorological site in the first target meteorological site and the load data of the selected bus, and selects a preset proportion of second target meteorological sites based on the result of the correlation; and uses the second target meteorological site as the input of the neural network to generate the final meteorological features of the selected bus load prediction. The present invention realizes the combination of meteorological data and bus load prediction, and at the same time, based on the weighted calculation of multiple meteorological sites and all meteorological elements, selects meteorological features with large influence weights on bus load prediction, thereby improving the prediction accuracy of bus load and ensuring the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0036] Figure 1 The present invention is a flow chart of a meteorological feature extraction method for bus load prediction.

[0037] Figure 2 The present invention is a flow chart of an embodiment of a method for extracting meteorological features for bus load prediction according to the present invention.

[0038] Figure 3 The present invention is a flow chart of another embodiment of a method for extracting meteorological features for bus load prediction according to the present invention.

[0039] Figure 4 The present invention is a flow chart of another embodiment of a method for extracting meteorological features for bus load prediction according to the present invention.

[0040] Figure 5 This is a schematic diagram of a meteorological feature extraction system for bus load prediction according to the present invention.

[0041] Figure 6 This is a neural network structure diagram for extracting meteorological features of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0043] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0044] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0045] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0046] See the following examples Figures 1 to 6 .

[0047] The embodiment of the present application provides a meteorological feature extraction method for bus load prediction, including:

[0048] Step S01, obtaining meteorological data of candidate meteorological sites, weightedly calculating the average absolute percentage error of all meteorological elements of candidate meteorological sites according to the meteorological data, and selecting a preset proportion of first target meteorological sites according to the result of the average absolute percentage error;

[0049] In this embodiment, meteorological data of candidate meteorological sites are collected, including multiple key meteorological elements such as temperature, humidity, rainfall, wind speed, and wind direction. Subsequently, these data are used to calculate the mean absolute percentage error (MAPE) of all meteorological elements of each candidate meteorological site, which is a common indicator for measuring the error between the predicted value and the actual value. According to the results of MAPE, meteorological sites with a preset ratio of small errors and high data quality are selected as the first target meteorological sites. In this way, meteorological sites with poor data quality or large errors are eliminated to ensure that the basic data for subsequent analysis is more accurate.

[0050] Step S02, obtaining load data of the selected bus, analyzing the correlation between the meteorological elements of each meteorological site in the first target meteorological site and the load data of the selected bus, and selecting a preset proportion of second target meteorological sites according to the correlation result;

[0051] By analyzing the correlation between the meteorological elements in the first target meteorological station and the selected bus load data. Correlation analysis can use a variety of statistical methods, such as random forest method, Pearson phase correlation and Spearman rank correlation coefficient. According to the analysis results, the meteorological stations with a preset ratio of high correlation with the bus load are selected as the second target meteorological stations. By analyzing the correlation between the bus load data and the characteristics of meteorological elements, the range of meteorological characteristics is further narrowed, and those meteorological elements that have a significant impact on the bus load are retained, thereby improving the prediction accuracy of the subsequent neural network model.

[0052] Step S03: using the second target meteorological site as an input of the neural network to generate final meteorological characteristics for the selected bus load prediction.

[0053] The meteorological data of the selected second target meteorological station is used as the input of the neural network. After training, the neural network model can extract the features most relevant to bus load forecasting from these input data and generate the final meteorological feature set.

[0054] In this embodiment, the meteorological feature extraction network structure is as follows Figure 6 As shown in the figure, the meteorological feature generation neural network consists of three parts:

[0055] Convolutional neural network: According to the number of meteorological element features, the convolution kernel size is selected as 5 × 5. The convolution kernel moves along the time axis, performs convolution operations on the time series, and splices the features obtained by n convolution kernels together.

[0056] Recurrent Neural Network: For each sample, each vector in can get the corresponding n-dimensional hidden state, which is obtained by concatenating the hidden states together.

[0057] Eigenvalue residual connection: The original input feature matrix is ​​added to the output of the neural network, and then normalized to obtain the final meteorological feature matrix.

[0058] The neural network model has powerful nonlinear mapping and self-learning capabilities, and can mine potential patterns and rules from complex input data. Through the learning of the neural network, a set of refined and effective meteorological features can be obtained. These features have extremely high predictive value for bus load forecasting. After the above processing, the meteorological feature extraction for bus load forecasting is completed.

[0059] A meteorological feature extraction method for bus load prediction of the present invention obtains meteorological data of candidate meteorological sites, calculates the average absolute percentage error of all meteorological elements of candidate meteorological sites based on the meteorological data, and selects a preset proportion of first target meteorological sites based on the result of the average absolute percentage error; obtains the load data of the selected bus, and analyzes the correlation between the meteorological elements of each meteorological site in the first target meteorological site and the load data of the selected bus, and selects a preset proportion of second target meteorological sites based on the result of the correlation; and uses the second target meteorological site as the input of the neural network to generate the final meteorological features of the selected bus load prediction. The present invention realizes the combination of meteorological data and bus load prediction, and at the same time, based on the weighted calculation of multiple meteorological sites and all meteorological elements, selects meteorological features with large influence weights on bus load prediction, thereby improving the prediction accuracy of bus load and ensuring the safe and stable operation of the power system.

[0060] Specifically, in a preferred implementation manner of the present application, the obtaining of meteorological data of the candidate meteorological site and weighted calculation of the mean absolute percentage error of all meteorological elements of the candidate meteorological site according to the meteorological data include:

[0061] Step S11, obtaining real data and forecast data of various meteorological elements of the candidate meteorological site, and calculating the average absolute percentage error of various meteorological elements in various preset periods based on the obtained data;

[0062] Collect real data and forecast data of various meteorological elements (such as temperature, humidity, rainfall, wind speed, wind direction, etc.) at candidate meteorological sites. Include multiple preset periods (such as day, week, month, and year, etc.) to ensure the comprehensiveness and accuracy of the evaluation. Then, for each meteorological element in each preset period, calculate its mean absolute percentage error (MAPE). MAPE is a common indicator for measuring forecast accuracy, which represents the percentage average of the error between the predicted value and the actual value. By calculating MAPE, the forecast accuracy of each meteorological element in different time periods can be quantified, providing basic data for subsequent weighted calculations.

[0063] Step S12, weighted calculation of the average absolute percentage error of various meteorological elements according to the prediction accuracy of the meteorological elements in various preset periods and the preset period weight coefficients;

[0064] After obtaining the MAPE of each meteorological element in different preset periods, weighted calculation is performed according to the prediction accuracy of these elements and the preset period weight coefficient. The period weight coefficient can be set according to actual needs to reflect the importance of different time periods to the prediction accuracy. For example, for bus load forecasting, recent meteorological data may be more important, so it can be given a higher weight. Through weighted calculation, the prediction accuracy of each meteorological element in different time periods is more comprehensively evaluated, and the impact of different time periods on the prediction results is taken into account.

[0065] Step S13: weightedly calculate the average absolute percentage error of all meteorological elements of the candidate meteorological station according to the average absolute percentage error of various meteorological elements and the preset element weight coefficient.

[0066] After obtaining the weighted MAPE of each meteorological element, a weighted calculation is performed according to the preset element weight coefficients (these coefficients can reflect the importance of different meteorological elements to bus load forecasting) to obtain the average absolute percentage error of all meteorological elements of the candidate meteorological station. By comprehensively considering the impact of all meteorological elements on bus load forecasting, a comprehensive error evaluation index is obtained. This index can be used to screen meteorological stations with high data quality and good prediction accuracy, and provide support for subsequent meteorological feature extraction and bus load forecasting.

[0067] Specifically, in a preferred implementation manner of the present application, screening a preset proportion of first target meteorological stations according to the result of the mean absolute percentage error includes:

[0068] The results of the mean absolute percentage errors are sorted from small to large, and a preset proportion of the first target meteorological stations in the top sorting are selected.

[0069] By sorting the MAPE, we can intuitively see which meteorological stations have smaller prediction errors, that is, higher data quality and prediction accuracy. Selecting the top-ranked meteorological station as the first target meteorological station can ensure that the basic data for subsequent analysis is accurate and reliable, which helps to improve the accuracy and stability of bus load forecasting.

[0070] In this embodiment, the MAPE results are sorted from small to large and a preset proportion of the first target meteorological stations in the top sort are selected, which helps to improve data quality and prediction accuracy, optimize resource allocation and computing efficiency, reduce noise interference and misjudgment risks, and provide support for subsequent analysis.

[0071] Specifically, in a preferred embodiment of the present application, the step of acquiring the load data of the selected bus, analyzing the correlation between the meteorological elements of each meteorological station in the first target meteorological station and the load data of the selected bus, and selecting a preset proportion of second target meteorological stations according to the result of the correlation includes:

[0072] Step S21, obtaining the substation location where the selected bus is located, and respectively calculating the distance between each meteorological station in the first target meteorological station and the substation where the selected bus is located to obtain a distance result;

[0073] Geographic distance is an important factor affecting the impact of meteorological factors on bus load. Generally, the meteorological data provided by meteorological stations that are closer can more accurately reflect the meteorological conditions in the area where the selected bus is located. By selecting meteorological stations that are closer to the bus, more valuable meteorological stations for bus load prediction can be selected.

[0074] Step S22, screening a preset number of third target meteorological stations according to the distance result;

[0075] Step S23, forming a meteorological feature matrix with the meteorological element forecast data of the third target meteorological station;

[0076] The meteorological feature matrix will be used as input data for subsequent random forest algorithm training. The construction of the meteorological feature matrix makes it possible to systematically process and analyze the impact of multiple meteorological elements on bus load, facilitating subsequent correlation analysis and feature screening.

[0077] Step S24, using the meteorological characteristic matrix and the load data of the selected bus as the input and output of the random forest algorithm for training, and obtaining the importance score result of each meteorological characteristic;

[0078] The random forest algorithm can handle high-dimensional data and complex nonlinear relationships. By training the random forest algorithm, we can get the importance ranking of various meteorological characteristics for bus load forecasting, and thus identify which meteorological elements have the greatest impact on bus load.

[0079] Step S25: According to the importance score result, a preset proportion of meteorological features is screened out as a second target meteorological site of selected bus correlation.

[0080] The meteorological feature set is further refined to retain only the most valuable meteorological stations and meteorological elements for bus load forecasting. These stations and elements will serve as important inputs for subsequent bus load forecasting models.

[0081] The above steps in this embodiment together constitute a scientific and systematic meteorological feature screening method, which can more accurately identify meteorological stations and meteorological elements that are highly correlated with the selected bus load, and provide support for the construction of subsequent bus load prediction models.

[0082] Specifically, in a preferred implementation of the present application, the meteorological characteristic matrix and the load data of the selected bus are used as the input and output of the random forest algorithm for training, and the importance score results of each meteorological characteristic are obtained, including:

[0083] Step 241, aligning the timestamps of the meteorological characteristic matrix and the load data of the selected bus;

[0084] Before inputting the meteorological feature matrix and the load data of the selected bus into the random forest algorithm, it is first necessary to ensure that the timestamps of the two are aligned. The time point of each meteorological feature should match the corresponding bus load data time point. Aligning the timestamps of the two is the basis for ensuring data consistency and model training accuracy. Only when the meteorological features and bus load data match at the same time point can the correlation between the two be effectively analyzed.

[0085] Step 242, preprocessing the meteorological characteristic matrix;

[0086] Preprocessing of the meteorological feature matrix, including processing of missing values, outlier detection and processing, data standardization or normalization, etc., can improve data quality, reduce noise, and ensure the stability and accuracy of model training.

[0087] Step 243, using the processed meteorological characteristic matrix as input and the load data of the selected bus as output to train the random forest regression model;

[0088] The random forest regression model can handle high-dimensional data and nonlinear relationships, making it very suitable for analyzing the complex relationship between meteorological characteristics and bus load. By training such a model, we can gain a deep understanding of bus load forecasting and identify which meteorological characteristics have the greatest impact on the forecast results.

[0089] Step 244: extract the importance score of each meteorological feature from the trained random forest regression model.

[0090] After the random forest regression model is trained, the importance scores of each meteorological feature are extracted from the model. These scores reflect the contribution of each feature in the model prediction, that is, the impact of each feature on the bus load prediction results. Extracting the importance scores of meteorological features helps to identify which meteorological elements are most critical to bus load prediction.

[0091] In this embodiment, by analyzing the relationship between meteorological characteristics and selected bus load data and obtaining the importance score of each meteorological characteristic, support is provided for subsequent feature selection and bus load prediction model construction.

[0092] In order to better illustrate the scheme of the meteorological feature extraction method for bus load prediction of the present application, an embodiment of distance will be described in detail below, and the steps are as follows:

[0093] Step 1: Selection of candidate meteorological stations: Number the meteorological stations and record them as M1, M2, ..., Mn. Taking M1 as an example, the actual temperature data and forecast data on date t are M1 and M2 respectively. t 、M1 t′ , the actual rainfall data and forecast data on date t are M1 and M2 respectively. r 、M1 r′ , the actual humidity data and forecast data on date t are M1 and M2 respectively. h 、M1 h′ , the actual wind speed data and forecast data on date t are M1 and M2 respectively. ws 、M1 ws′ , the actual data and forecast data of wind direction on date t are M1 and M2 respectively. wd 、M1 wd′ MAPE is used to calculate the prediction accuracy of each element of the meteorological station. For example, the calculation formula of rainfall MAPE of meteorological station M1 on date t is:

[0094]

[0095] Calculate the MAPE of the last week, last month, last three months, last six months, and last twelve months respectively. For example, the rainfall accuracy of weather station M1 in the last week is:

[0096]

[0097] Calculate the rainfall MAPE of weather station M1 in the last week, last month, last three months, last six months, and last twelve months, and record it as: M1WeekMAPE r ,M1Month1MAPE r ,M1Month3MAPE r ,M1Month6MAPE r ,M1Month12MAPE r The above MAPEs are weighted and the accuracy of the rainfall forecast data of the weather station is obtained. The calculation process is:

[0098] M1MAPE r =0.2*M1Month1MAPE r+0.3*M1Month3MAPE r +0.3*M1Month6MAPE r +0.2*M1Month12MAPE r

[0099] In the same way, the MAPE of temperature, humidity, rainfall, wind speed, wind direction and other factors of weather station M1 are calculated respectively, which is recorded as M1MAPE t 、M1MAPE h 、M1MAPE r 、M1MAPE ws 、M1MAPE wd . Then weightedly calculate the total factor MAPE of the weather station, the formula is as follows:

[0100] M1MAPE=0.2*M1MAPE wd +0.3*M1MAPE t +0.3*M1MAPE r +0.1*M1MAPE h +0.1*M1MAPE wd

[0101] The total factor MAPE values ​​calculated by M1, M2, ..., Mn are sorted in ascending order, and the first 50% of the meteorological stations are selected as candidate meteorological stations, which are recorded as the meteorological station set MS.

[0102] Step 2: Meteorological feature selection based on bus load impact weight: record the bus load data as B1, B2, ..., B p , from M s Select busbar B p The most relevant factor of the load data is used as the input factor for the subsequent prediction model. p Taking as an example, the specific method for selecting meteorological elements is:

[0103] Get bus B p The coordinates of the substation where the M s The distance from the meteorological station to the substation is selected, and the three nearest meteorological stations are recorded as MS1, MS2, and MS3. The forecast data of temperature, humidity, rainfall, wind speed, wind direction, etc. of the three meteorological stations are spliced ​​to form the feature matrix F bp ,Right now:

[0104] F bp =[MS1 r MS1 t MS1 h ...MS3 r MS3 t MS3h ]

[0105] Using the random forest algorithm, the above matrix F bp As the input of the algorithm, B p As the output of the algorithm, a regression model is constructed. After the model is fitted, the importance values ​​of each feature of the random forest model are calculated, the feature importance is sorted in descending order, and the features with the top 70% of feature importance are selected as the mother line B. p The meteorological feature set of correlation is denoted as F bp′ .

[0106] Step 3: Generation method of weather forecast features based on convolutional neural network and recurrent neural network: bp′ As the input of the entire neural network, step S3, the method for generating weather forecast features based on convolutional neural network and recurrent neural network: F bp′ As the input of the entire neural network, the weather forecast feature generation neural network consists of three parts:

[0107] Convolutional neural network: According to the number of meteorological element features, the convolution kernel size is selected as 5 × 5. The convolution kernel moves along the time axis, performs convolution operations on the time series, and splices the features obtained by n convolution kernels together.

[0108] Recurrent Neural Network: For each sample, each vector in can get the corresponding n-dimensional hidden state, which is obtained by concatenating the hidden states together.

[0109] Eigenvalue residual connection: The original input feature matrix is ​​added to the output of the neural network, and then normalized to obtain the final meteorological feature matrix.

[0110] After the above processing, the meteorological feature extraction operation for bus load forecasting is completed.

[0111] The embodiment of the present application further provides a meteorological feature extraction system for bus load prediction, which is applied to the aforementioned meteorological feature extraction method for bus load prediction, comprising:

[0112] A first target meteorological site acquisition unit is used to acquire meteorological data of candidate meteorological sites, calculate the average absolute percentage error of all meteorological elements of candidate meteorological sites based on the meteorological data, and select a preset proportion of first target meteorological sites based on the result of the average absolute percentage error;

[0113] A second target meteorological site acquisition unit, used to acquire load data of a selected bus, analyze the correlation between meteorological elements of each meteorological site in the first target meteorological site and the load data of the selected bus, and select a preset proportion of second target meteorological sites according to the correlation result;

[0114] The final meteorological feature selection unit is used to use the second target meteorological site as an input of the neural network to generate the final meteorological features of the selected bus load forecast.

[0115] The functional explanation of each unit / module of a meteorological feature extraction system for bus load prediction in this embodiment is the same as that of an embodiment of a meteorological feature extraction method for bus load prediction, and the technical effects are the same, so they will not be repeated here.

[0116] An embodiment of the present application also provides a computer-readable storage medium, which includes a stored program, wherein when the program is executed, the aforementioned meteorological feature extraction method for bus load prediction is controlled in the device where the computer-readable storage medium is located.

[0117] The technical effect of this embodiment is the same as the technical effect of an embodiment of a meteorological feature extraction method for bus load prediction, and will not be repeated here.

[0118] The present invention can be used in many general or special computer system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like.

[0119] An embodiment of the present application further provides a processor, which is used to run a program, wherein the program executes the aforementioned meteorological feature extraction method for bus load prediction when running.

[0120] The technical effect of this embodiment is the same as the technical effect of an embodiment of a meteorological feature extraction method for bus load prediction, and will not be repeated here.

[0121] The processor in this embodiment may be a central processing unit (CPU), a controller, a microcontroller, or other data processing chips.

[0122] Those of ordinary skill in the art will appreciate that the units of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0123] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.

[0124] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0125] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nlyMemory), random access memory (RAM, RandomAccessMemory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A meteorological feature extraction method for bus load forecasting, characterized in that: include: Acquire meteorological data of candidate meteorological sites, calculate the average absolute percentage error of all meteorological elements of candidate meteorological sites by weight according to the meteorological data, and select a preset proportion of first target meteorological sites according to the result of the average absolute percentage error; Obtaining the load data of the selected bus, analyzing the correlation between the meteorological elements of each meteorological station in the first target meteorological station and the load data of the selected bus, and selecting a preset proportion of second target meteorological stations according to the correlation result; The second target meteorological site is used as an input of the neural network to generate final meteorological characteristics for the selected bus load forecast.

2. A meteorological feature extraction method for bus load prediction according to claim 1, characterized in that: The step of obtaining meteorological data of the candidate meteorological site and weightedly calculating the average absolute percentage error of all meteorological elements of the candidate meteorological site according to the meteorological data comprises: Obtaining the actual data and forecast data of various meteorological elements of the candidate meteorological sites, and calculating the average absolute percentage errors of various meteorological elements within various preset periods based on the acquired data; According to the forecast accuracy of meteorological elements in various preset periods and the preset period weight coefficient, the average absolute percentage error of various meteorological elements is weightedly calculated; According to the average absolute percentage error of various meteorological elements and the preset element weight coefficient, the average absolute percentage error of all meteorological elements of the candidate meteorological station is weightedly calculated.

3. A meteorological feature extraction method for bus load prediction according to claim 2, characterized in that: The step of selecting a preset proportion of first target meteorological stations according to the result of the mean absolute percentage error comprises: The results of the mean absolute percentage errors are sorted from small to large, and a preset proportion of the first target meteorological stations in the top sorting are selected.

4. The method for extracting meteorological features for bus load prediction according to claim 2, characterized in that: The step of acquiring the load data of the selected bus, analyzing the correlation between the meteorological elements of each meteorological station in the first target meteorological station and the load data of the selected bus, and selecting a preset proportion of second target meteorological stations according to the result of the correlation comprises: Obtain the location of the substation where the selected bus is located, and calculate the distance between each meteorological station in the first target meteorological station and the substation where the selected bus is located to obtain a distance result; Selecting a preset number of third target meteorological stations according to the distance result; Combining the meteorological element forecast data of the third target meteorological station into a meteorological characteristic matrix; The meteorological characteristic matrix and the load data of the selected bus are used as the input and output of the random forest algorithm for training respectively, and the importance score results of each meteorological characteristic are obtained; According to the importance score result, meteorological features of a preset proportion are screened out as the second target meteorological site of the selected bus correlation.

5. A meteorological feature extraction method for bus load prediction according to claim 4, characterized in that: The meteorological characteristic matrix and the load data of the selected bus are used as the input and output of the random forest algorithm for training respectively, and the importance score results of each meteorological characteristic are obtained, including: Aligning the timestamps of the meteorological characteristic matrix and the load data of the selected bus; Preprocessing the meteorological characteristic matrix; The processed meteorological feature matrix is ​​used as input, and the load data of the selected bus is used as output to train the random forest regression model; the importance score of each meteorological feature is extracted from the trained random forest regression model.

6. A meteorological feature extraction system for bus load forecasting, characterized in that: The meteorological feature extraction method for bus load prediction applied to any one of claims 1 to 5 comprises: A first target meteorological site acquisition unit is used to acquire meteorological data of candidate meteorological sites, calculate the average absolute percentage error of all meteorological elements of candidate meteorological sites based on the meteorological data, and select a preset proportion of first target meteorological sites based on the result of the average absolute percentage error; A second target meteorological site acquisition unit, used to acquire load data of a selected bus, analyze the correlation between meteorological elements of each meteorological site in the first target meteorological site and the load data of the selected bus, and select a preset proportion of second target meteorological sites according to the correlation result; The final meteorological feature selection unit is used to use the second target meteorological site as an input of the neural network to generate the final meteorological features of the selected bus load forecast.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the meteorological feature extraction method for bus load prediction according to any one of claims 1 to 5.

8. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the meteorological feature extraction method for bus load prediction according to any one of claims 1 to 5.

Images