User side power load prediction method and device

Abstract

The invention discloses a user side power load prediction method and device, and relates to the technical field of power. According to the main technical scheme of the invention, the method comprises the steps: building the correlation between data of two dimensions through the correlation between historical meteorological record data and historical power load data in time features, and obtaining time sequence data sorted according to dates; the time sequence data comprises a sequence order among a plurality of pieces of combined data, and each piece of combined data comprises data in two dimensions of weather and power load; and then, a preset machine learning architecture is adopted to perform training processing on the time sequence data, so that supervised learning training is performed on the time sequence to obtain a target prediction model with strong prediction capability, and the target prediction model is applied to an application scene of user-side power prediction.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a method and device for predicting user-side electric power load. Background Art

[0002] In the field of power technology, user-side power load, also known as electricity load, refers to the total power consumed by a certain range of power-consuming equipment in the power system at a certain moment. It is a constantly changing quantity that varies with time and season.

[0003] At present, with the construction of new power systems and the development of intelligent technologies, user-side power load management is developing in a more intelligent and refined direction. Among them, if the user-side power load can be effectively predicted, it will bring great benefits to optimizing power resource allocation, improving power system stability, reducing operating costs, etc. Therefore, how to propose an effective user-side power load prediction solution is a technical problem that needs to be solved urgently. Summary of the Invention

[0004] In view of this, the present application provides a user-side power load forecasting method and device, the main purpose of which is to construct time series data based on the two dimensions of meteorology and power load, and then use machine learning for training, so as to integrate the three factors of time, meteorology and historical power load to construct a more comprehensive user-side power forecasting model, thereby providing an effective user-side power load forecasting solution.

[0005] In order to achieve the above objectives, this application mainly provides the following technical solutions:

[0006] A first aspect of the present application provides a method for predicting user-side power load, the method comprising:

[0007] Obtaining historical meteorological record data, wherein the historical meteorological record data is daily meteorological record data within a preset time range, the daily meteorological record data includes meteorological data at each preset unit time interval, and the meteorological data includes at least temperature, humidity, light radiation, and wind speed; wherein the daily meteorological record data corresponds to an associated date and season;

[0008] Acquire historical power load data, wherein the historical power load data is daily power load data within the preset time range, and the daily power load data includes power load data at each preset unit time interval;

[0009] Associate the historical meteorological record data and the historical power load data according to the same time characteristics to obtain time series data sorted by date. The time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load;

[0010] Use a preset machine learning architecture to perform training processing on the time series data to construct a target prediction model for predicting the power load on the user side;

[0011] Use the target prediction model to process the obtained current meteorological record data and current power load data, and output the prediction result of the power load on the user side in the future time range.

[0012] The second aspect of this application provides a power load prediction device on the user side. The device includes:

[0013] The first acquisition unit is used to acquire historical meteorological record data. The historical meteorological record data is the daily meteorological record data within a preset time range. The daily meteorological record data is the meteorological data at every preset unit time interval, and the meteorological data at least includes temperature, humidity, light irradiance, and wind speed; wherein, the daily meteorological record data is correspondingly associated with dates and seasons;

[0014] The second acquisition unit is used to acquire historical power load data. The historical power load data is the daily power load data within the preset time range. The daily power load data is the power load data at every preset unit time interval;

[0015] The third acquisition unit is used to associate the historical meteorological record data and the historical power load data according to the same time characteristics to obtain time series data sorted by date. The time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load;

[0016] The processing unit is used to use a preset machine learning architecture to perform training processing on the time series data to construct a target prediction model for predicting the power load on the user side;

[0017] The prediction unit is used to use the target prediction model to process the obtained current meteorological record data and current power load data, and output the prediction result of the power load on the user side in the future time range.

[0018] The third aspect of this application provides a storage medium. The storage medium includes a stored program. When the program runs, it controls the device where the storage medium is located to execute the power load prediction method on the user side as described above.

[0019] The fourth aspect of the present application provides an electronic device, which includes at least one processor, at least one memory connected to the processor, and a bus;

[0020] Wherein, the processor and the memory complete communication with each other through the bus;

[0021] The processor is used to call program instructions in the memory to execute the user-side power load prediction method as described above.

[0022] By means of the above technical solutions, the technical solutions provided by the present application have at least the following advantages:

[0023] The present application provides a user-side power load prediction method and device. The present application utilizes the correlation between historical meteorological record data and historical power load data in terms of time characteristics to establish the association between these two-dimensional data, thereby obtaining time series data sorted by date, and the time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load; then the present application further uses a pre-set machine learning architecture to train and process the time series data, so as to perform supervised learning training on the time series to obtain a target prediction model with strong prediction ability for application to the application scenario of user-side power prediction.

[0024] Compared with the existing demand for predicting the user-side power load, the present application constructs time series data from two dimensions of meteorology and power load, and then uses machine learning for training, thereby comprehensively considering three factors of time, meteorology and historical power load to construct a more comprehensive user-side power prediction model, thus providing an effective user-side power load prediction solution.

[0025] The above description is only an overview of the technical solutions of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically illustrates the specific implementation manners of the present application. Description of the Drawings

[0026] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present application. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0027] Figure 1 It is a flowchart of a user-side power load prediction method provided by an embodiment of the present application;

[0028] Figure 2 Another flowchart of the user-side power load prediction method provided by the embodiment of the present application;

[0029] Figure 3 The ROC curve obtained by taking whether it is a rainy day as an example provided by the embodiment of the present application;

[0030] Figure 4 Determine the number of clustering clusters for all data provided by the embodiment of the present application;

[0031] Figure 5a Determine the number of clustering clusters for the data with light provided by the embodiment of the present application;

[0032] Figure 5b Determine the number of clustering clusters for the data without light provided by the embodiment of the present application;

[0033] Figure 6 Predict the photovoltaic output result by the sub-modal prediction model provided by the embodiment of the present application;

[0034] Figure 7 The block diagram of the composition of a user-side power load prediction device provided by the embodiment of the present application;

[0035] Figure 8 The block diagram of the composition of another user-side power load prediction device provided by the embodiment of the present application. Detailed implementation manners

[0036] The exemplary embodiments of the present application will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be completely conveyed to those skilled in the art.

[0037] The embodiment of the present application provides a user-side power load prediction method, as Figure 1 shown, and the following specific steps are provided for the embodiment of the present invention:

[0038] 101. Obtain historical meteorological record data, where the historical meteorological record data is the daily meteorological record data within a preset time range, and the daily meteorological record data includes meteorological data at every preset unit time interval, and the meteorological data includes at least temperature, humidity, light irradiance, and wind speed; wherein, the daily meteorological record data is associated with the corresponding date and season.

[0039] Meteorological changes will have a certain impact on the user-side power load. Exemplary examples include:

[0040] (1) Influence of temperature. Summer cooling load: High temperature weather will cause a sharp increase in the electricity consumption of cooling equipment such as air conditioners and fans, thus increasing the power load on the user side; with the improvement of living standards, the popularity of cooling equipment such as air conditioners in families has been continuously increasing, further exacerbating the growth of summer cooling load.

[0041] Winter heating load: Low temperature weather will cause an increase in the electricity consumption of heating equipment such as electric heaters and floor heating, which will also increase the power load on the user side; in some areas, the winter heating load may even exceed the summer cooling load and become the main factor affecting the power load on the user side.

[0042] (2) Influence of humidity: Especially in high temperature and high humidity or low temperature and high humidity environments, humidity changes will affect human comfort and equipment working efficiency, thus indirectly affecting the power load;

[0043] High temperature and high humidity environment: In this environment, people feel more stuffy, so more cooling equipment is needed to maintain comfort, resulting in an increase in power load; Low temperature and high humidity environment: Although the direct influence of humidity on the power load in a low temperature environment is small, high humidity may cause a decrease in the insulation performance of equipment and an increase in equipment failure rate, thus indirectly affecting the power load.

[0044] (3) Influence of other meteorological factors

[0045] In addition to temperature and humidity, other meteorological factors such as air pressure, wind speed, cloud cover or solar radiation intensity may also affect the power load on the user side. For example:

[0046] Air pressure change: The change in air pressure may affect air circulation and human comfort, thus indirectly affecting the power load; Wind speed change: The change in wind speed may affect the output of wind power generation equipment, and then affect the power supply capacity of the power grid and the power load on the user side; Cloud cover or solar radiation intensity: The change in cloud cover and solar radiation intensity will affect the output of solar power generation equipment, and may also affect the power supply capacity of the power grid and the power load on the user side.

[0047] Therefore, considering the influence brought by different meteorological characteristics, the historical meteorological record data obtained in the embodiments of the present application is required to include at least these meteorological characteristics. And in order to improve the accuracy of the subsequent training target prediction model, and at the same time in order to manage these large amounts of complex and diverse meteorological data, the embodiments of the present application may but are not limited to record in units of days, and the daily record includes meteorological data at every preset unit time interval (such as 1 hour as the time interval), and at least requires that these meteorological data contain meteorological characteristics such as temperature, humidity, light radiation intensity, and wind speed.

[0048] 102. Obtain historical power load data, where the historical power load data is the daily power load data within a preset time range, and the daily power load data includes the power load data at every preset unit time interval.

[0049] In order to establish an association with the historical meteorological record data obtained in 101, the embodiments of the present application can, but are not limited to, obtain the daily power load data within the same preset time range and with the same time unit (such as 1 day), and further perform fine-grained division on a daily basis at every preset unit time interval (such as 1 hour as the time interval) to obtain the change of the historical power consumption load at every preset unit time interval.

[0050] 103. According to the same time characteristics, establish an association between the historical meteorological record data and the historical power load data to obtain time series data sorted by date. The time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load.

[0051] As in 101 and 102, the embodiments of the present application establish an association between these two different-dimensional historical data based on the similarities in time characteristics of the historical meteorological record data and the historical power load data. For example, combined data composed of the historical meteorological record data and the historical power load data on a daily basis is obtained, and thus time series data with 1 day as the unit time is obtained by sorting by date.

[0052] 104. Use a preset machine learning architecture to perform training processing on the time series data to construct a target prediction model applied to predicting the power load on the user side.

[0053] 105. Use the target prediction model to process the obtained current meteorological record data and current power load data, and output the prediction result of the power load on the user side in a future time range.

[0054] The embodiments of the present application use the time series data obtained in 103 as samples for model training, and at the same time adopt a preset machine learning architecture (such as the XGBoost framework) to convert the training on the time series into a model training of supervised learning, so as to increase the ability of the model to predict the power load to a certain extent considering comprehensive meteorological factors.

[0055] It should be noted that the time series data obtained in 103 is sorted by date, and the time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load. However, the two are actually associated based on time characteristics. Therefore, in fact, each combined data contains three dimensions (such as meteorology, time, and historical power load).

[0056] Exemplarily, during the model training process, the key steps include extracting useful features from the original data, which refers to the time series data obtained at 103. The extracted features include two dimensions, such as meteorology and time.

[0057] First, for each combined data, since the data in the meteorology dimension can include some meteorological description information, in the embodiments of the present application during feature extraction, each combined data needs to be converted into numerical data, and then the representation of meteorological features is extracted. For example: the numerical representations of key indicators such as the highest temperature, the lowest temperature, the average temperature, the light radiation intensity, and the wind speed are extracted from each combined data. These numerical representations (corresponding to meteorological features) directly reflect the impact of meteorological conditions on the power load.

[0058] Exemplarily, the examples can include but are not limited to the following:

[0059] 1. According to the sunset and sunrise time strings, after extraction, the time difference is obtained to get the feature of the lighting time, and the unit is minutes (min).

[0060] 2. According to the temperature range string, text extraction is performed to obtain two columns of features of the daily minimum and maximum temperatures, and the unit is degrees Celsius (°C).

[0061] 3. According to the wind direction and wind level strings, text extraction is performed to obtain two columns of features of the wind direction and the wind level

[0062] 4. Extract the wind direction description information to construct discrete event features.

[0063] 5. Extract the wind level description information to construct two columns of features of the minimum and maximum wind levels

[0064] 6. Extract the meteorological condition description information to construct discrete event features.

[0065] Secondly, for each combined data, in addition to extracting meteorological features as above, time features also need to be extracted, such as extracting time period features including date, day of the week, hour, etc. These features help the model capture the impact of seasonal changes and time trends on the power load.

[0066] By performing feature extraction operations on the time series data, the feature data in two dimensions of meteorology and time included in each combined data is obtained, and these two dimensions (such as meteorology, time) are related to the historical power data included in the combined data.

[0067] After the feature extraction operation, for the processed time series data, the embodiments of the present application convert the time series data into supervised learning, which can utilize the powerful prediction ability of the supervised learning algorithm to analyze and predict the trends and patterns in the time series. Thus, by integrating the three factors of time, meteorology, and historical power load, a more comprehensive user-side power prediction model is constructed, thereby providing an effective user-side power load prediction solution.

[0068] To make a more detailed description of the above embodiments, the embodiments of the present application also provide another user-side power load prediction method, as Figure 2 shown. For this, the embodiments of the present application provide the following specific steps:

[0069] 201. Obtain historical meteorological record data, where the historical meteorological record data is the daily meteorological record data within a preset time range. The daily meteorological record data includes meteorological data at every preset unit time interval, and the meteorological data includes at least temperature, light irradiance, and wind speed; wherein, the daily meteorological record data is associated with the corresponding date and season.

[0070] 202. Obtain historical power load data, where the historical power load data is the daily power load data within a preset time range. The daily power load data includes power load data at every preset unit time interval.

[0071] 203. Establish an association between the historical meteorological record data and the historical power load data according to the same time feature to obtain time series data sorted by date. The time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load.

[0072] In the embodiments of the present application, for the explanations of 201-203 above, refer to 101-103, which will not be elaborated here. The following uses 204-208 to implement the training of a target prediction model for predicting the user-side power load.

[0073] 204. Perform feature processing on the time series data to obtain the feature data corresponding to each combined data in the time series data. The feature data includes at least: converting the description information in the meteorological data into a numerical representation.

[0074] It should be noted that the time series data obtained in 203 is sorted by date, and the time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load. However, the two are actually associated based on the time feature. Therefore, in fact, each combined data contains three dimensions (such as meteorology, time, and historical power load).

[0075] For each combined data, since the data in the meteorological dimension can include some meteorological description information, in the embodiments of the present application during feature extraction, it is necessary to convert each combined data into numerical data, and then extract the representation of meteorological features. For example: extract the numerical representations of key indicators such as the highest temperature, the lowest temperature, the average temperature, the light radiation intensity, and the wind speed from each combined data. These numerical representations (corresponding to meteorological features) will directly reflect the impact of meteorological conditions on the power load.

[0076] Secondly, for each combined data, in addition to extracting meteorological features as above, time features also need to be extracted, such as extracting time period features including: date, day of the week, hour, etc. These features help the model capture the impact of seasonal changes and time trends on the power load.

[0077] By performing feature extraction operations on the time series data, the feature data in two dimensions of meteorology and time included in each combined data are obtained, and these two dimensions (such as meteorology, time) are associated with the historical power data included in the combined data.

[0078] In some alternative embodiments, in the final stage of feature extraction, feature selection and dimensionality reduction operations can also be performed to remove redundant features and reduce computational complexity, retaining the features that contribute most to power load prediction, thereby improving the generalization ability and prediction accuracy of the model obtained in subsequent training.

[0079] 205. According to the sorting corresponding to different combined data in the time series data, add sorting serial numbers to the feature data corresponding to different combined data to obtain a unique sorting serial number for each feature data.

[0080] After undergoing the feature extraction operation, for the processed time series data, the embodiments of the present application convert the time series data into supervised learning, which can utilize the powerful prediction ability of supervised learning algorithms to analyze and predict the trends and patterns in the time series.

[0081] 206. Determine the data sequence composed of feature data according to the sorting sequence of the sorting serial numbers.

[0082] 207. By continuously sliding a preset sliding window on the data sequence at a preset interval, multiple groups of input features and output labels are determined. The input features are the feature data currently selected by the preset sliding window, and the output label is the next adjacent feature data of the feature data currently selected by the preset sliding window.

[0083] 208. Use a preset XGBoost framework to perform training processing on multiple groups of input features and output labels to construct a prediction model applied to predicting the power load on the user side.

[0084] 209. Process the acquired current meteorological record data and current power load data using the target prediction model, and output the prediction result of the power load on the user side in the future time range.

[0085] As an explanation of 205 - 209. In the embodiments of the present application, it is to convert the time series into supervised learning participation, such as using the sliding window technique to achieve.

[0086] For example, by setting a window of a fixed size and sliding it on the time series data, multiple time window samples are generated. Each window sample contains a series of consecutive time step data (i.e., input features), and the data point after the window is used as the target value (i.e., input label). Thus, based on the movement of the sliding window, multiple sets of data (including input features and output labels) are obtained, making it suitable for training a supervised learning model.

[0087] For example, in a time series containing daily meteorological data, the time window sample uses the data of the past 7 days as input features, and the target value of the 8th day after this window is used as the prediction result, that is, the output label. By adjusting the window size and the sliding window, patterns and trends at different time scales can be flexibly captured, that is, multiple sets of data (including input features and output labels). Apply these multiple sets of data to the model of supervised learning for training. Thus, by integrating the three factors of time, meteorology, and historical power load, a more comprehensive power prediction model for the user side is constructed, thereby providing an effective power load prediction solution for the user side.

[0088] As a supplementary optimization of the solutions of 201 - 208 in the embodiments of the present application, after acquiring the historical meteorological record data, the embodiments of the present application can also train a meteorological prediction model, so as to predict future meteorological data based on the historical meteorological record data and apply it to the target prediction model trained above for predicting power load to improve the accuracy of the target prediction model. Next, for training a meteorological prediction model, it can include but is not limited to the following steps:

[0089] A1. Through feature analysis and processing of the historical meteorological record data, extract the first meteorological record data corresponding to the meteorological adverse features.

[0090] A2. In the historical meteorological record data, determine the other data except the first meteorological record data as the second meteorological record data.

[0091] A3. Adopt the vector autoregressive framework, use the first meteorological record data as the positive sample and the second meteorological record data as the negative sample to train the meteorological prediction model.

[0092] If the difference in the amount of data between the first meteorological record data and the second meteorological record data is greater than a preset threshold; during the process of training the meteorological prediction model, adjust the recognition threshold of the classifier for the samples with a smaller number in the first meteorological record data and the second meteorological record data

[0093] A4 will apply the meteorological prediction data output by the meteorological prediction model to the target prediction model to predict the future user-side power load.

[0094] As in A1 - A4, the embodiments of the present application may, but are not limited to, adopt a vector autoregressive (VAR) model. VAR is an advanced multivariate time series analysis method for predicting the short-term trends of a series of key meteorological variables. The core advantage of the VAR model lies in its ability to simultaneously process multiple interrelated meteorological indicators, including temperature, dew point, relative humidity, atmospheric pressure, visibility, and wind speed, etc. By constructing a comprehensive prediction framework, the model created by the embodiments of the present application can capture the complex dynamics and interactions among these variables, thereby providing more accurate and comprehensive meteorological predictions.

[0095] The purpose of the embodiments of the present application is to use the VAR model to predict meteorological conditions for the next year. The training of the model is based on historical meteorological data, including multi-dimensional time series such as continuously recorded temperature, dew point, humidity, pressure, visibility, and wind speed. Through the VAR model, the joint probability distribution of these meteorological variables can be simulated, and their evolution in the future time period can be predicted.

[0096] During the model establishment process, first preprocess the data to ensure the quality and consistency of the data, such as obtaining the numerical data structure of the meteorological data after data normalization processing. Subsequently, identify and set the optimal lag order of the model, which is achieved by analyzing the autocorrelation function and partial autocorrelation function of each variable according to the Akaike information criterion (AIC criterion) or Bayesian information criterion (BIC criterion).

[0097] Furthermore, the embodiments of the present application use the VAR model to estimate the data, thereby obtaining the coefficient matrix of each variable. These coefficients capture the immediate and lagged effects between variables, and compare the predicted future data with the real future data.

[0098] In some alternative embodiments, the meteorological prediction model constructed by the embodiments of the present application may be a meteorological multi-modal classification model. Exemplary explanations include the following:

[0099] Adjusting the classification threshold is a key step in tackling classification tasks with an imbalance of positive and negative samples. Since severe weather events are typically rare, the model may tend to predict negative cases (i.e., normal weather) during training, resulting in uneven prediction performance and reduced effectiveness in real-world applications. Therefore, adjusting the threshold can help balance the classification results between positive and negative samples, optimizing the model's performance.

[0100] Specifically, the purpose of adjusting the threshold is to improve the classifier's ability to identify the minority class (i.e., severe weather) by changing the decision boundary. In standard binary classification models, a default threshold of 0.5 is typically used, meaning that predicted probabilities greater than 0.5 are classified as positive, and those less than 0.5 are classified as negative. However, in cases where there is a severe imbalance between positive and negative samples, the default threshold may cause the prediction of severe weather to be biased towards the negative class. By adjusting the threshold, the probability of predicting the positive class can be increased, thereby improving the model's recognition rate for severe weather.

[0101] Methods for adjusting the threshold include optimization based on performance metrics, such as adjusting the threshold to maximize the F1 score, precision, or recall. Alternatively, the receiver operating characteristic (ROC) curve or precision-recall (PR) curve can be used to select the optimal threshold. These methods can ensure that the model can better capture severe weather events when faced with imbalanced data, improving the accuracy and reliability of forecasts.

[0102] In the examples of this application, the ROC-AUC curve is used to find the balance between maximizing the true positive rate (TPR) and minimizing the false positive rate (FPR). This is usually achieved by evaluating the difference between the TPR and the FPR, which is sometimes called the "Youden's J statistic." It measures the classifier's ability to reduce false negatives and false positives. A larger value of tpr-fpr indicates better classification performance.

[0103] The decision threshold determination formula is as follows:

[0104] optimal_idx=argmax i (TPR i -FPR i ) formula (1);

[0105] Among them, TPR i is the true rate (recall rate) at the i-th threshold, FPR i is the false positive rate at the i-th threshold, argmax i It means finding the index i that maximizes the given expression.

[0106] Taking whether it is a rainy day as an example, a random forest classifier is used to construct a model, and the ROC curve is obtained as Figure 3 .

[0107] According to the curve in the above figure, the decision boundary is calculated to be 0.284, that is:

[0108]

[0109] The classification results of the model on the test set are as follows, including the confusion matrix and recall rate of the model on the test set.

[0110]

[0111] The classification effect of the model on the future numerical meteorological data predicted by the VAR model is as follows, including the confusion matrix and recall rate of the model on the future prediction set.

[0112]

[0113] Above, by reasonably adjusting the threshold, the model can more effectively identify minority class events when dealing with imbalanced classification tasks, and improve the prediction performance in practical applications.

[0114] As a supplementary optimization of the solutions of Embodiments 201-208 of this application, after obtaining historical power load data, this application embodiment can also train a new energy output prediction model to utilize the predicted future new energy output data and apply it to the target prediction model trained above for predicting power load to improve the accuracy of the target prediction model. Next, for training a new energy output prediction model, it may but is not limited to including the following steps:

[0115] B1 Analyze and process the historical power load data to extract the target historical data corresponding to the influence of new energy output.

[0116] New energy output refers to the electric power generated by new energy generating units in a power system at a certain moment or time period. These new energy generating units usually use renewable energy (such as solar energy, wind energy, etc.) for power generation, and the pollution generated during the power generation process is less, belonging to clean energy. New energy output includes at least wind energy output and photovoltaic output.

[0117] B2 Use the pre-set LightGBM framework to train and process the target historical data to construct a new energy output prediction model.

[0118] The LightGBM framework is based on the gradient boosting algorithm and improves the model's accuracy by constructing multiple weak learners (usually decision trees). Each new tree is fitted to the residuals (i.e., the difference between the predicted value and the actual value) based on the previous tree. Its goal is to iteratively train weak learners and combine them into a strong learner to achieve accurate prediction of the data.

[0119] B3 will utilize the output prediction data output by the new energy output prediction model and apply it to the target prediction model to predict the future user-side power load.

[0120] As described above for B1 - B3, the embodiments of this application are explained as follows:

[0121] By analyzing and processing historical power load data, the corresponding target historical data affected by the new energy output is extracted from it, thereby obtaining the historical new energy output data. And the obtained new energy output prediction model is a new energy output multi-modal classification model. Taking the photovoltaic output as an example, an explanation is given.

[0122] To further improve the prediction accuracy, the embodiments of this application introduce a sub-modal classification model and add it as a new feature to the training process of the LightGBM model. Specifically, first, a sub-modal classification model is trained to divide the meteorological data and the photovoltaic output power into several modes. These mode information reflects the response characteristics of the photovoltaic system under different meteorological conditions. Subsequently, the classification results are used as new features and combined with the original meteorological data to retrain the LightGBM model. In this way, the mode information extracted by the classification model can be used to enhance the adaptability of the prediction model to the photovoltaic output power.

[0123] In the embodiments of this application, the potential laws of the data are mainly discovered through the clustering algorithm and divided into different modes. Usually, the clustering algorithm needs to determine the number of clustering clusters in advance. In this study, the number of modes that can be classified is determined according to the elbow method and the silhouette coefficient.

[0124] The elbow method is a heuristic method for selecting the optimal number of clusters in K-means clustering. It is based on the concept of the within-cluster sum of squares (WCSS). WCSS is the sum of the squares of the distances from all points within the cluster to the cluster center, and the calculation formula (3) is:

[0125]

[0126] where k is the number of clusters, C i is the set of points in the i-th cluster, x is a point in C i , and μ i is the cluster Ci The center point (centroid), ||x - μ i || is the Euclidean distance from point x to the centroid μ i of the Euclidean distance.

[0127] The elbow method is implemented by plotting the WCSS against the number of clusters k. The optimal number of clusters k is the point where the rate of decrease of the WCSS suddenly slows down, i.e., the "elbow".

[0128] The silhouette coefficient is a value between -1 and 1, used to evaluate the rationality of clustering for each data point. The silhouette coefficient s i For each point x i is calculated using the following formula (4):

[0129]

[0130] where a i is the average distance from point x i to the nearest point in the same cluster, and b i is the average distance from point x i to the nearest point in the nearest cluster (not its own cluster).

[0131] The average of the silhouette coefficients for all points can be used to evaluate the quality of the clustering. A silhouette coefficient close to 1 indicates good clustering, close to -1 indicates poor clustering, and close to 0 indicates that the point may be on the boundary of two clusters. When determining the optimal number of clusters, the average of the silhouette coefficients for all points is usually calculated for different values of k, and the k value that maximizes the average silhouette coefficient is selected.

[0132] Both of these methods are used to assist in determining the number of clusters in cluster analysis, but they focus on different aspects: the elbow method focuses on the change in within-cluster cohesion, while the silhouette coefficient takes into account both within-cluster cohesion and between-cluster separation. Therefore, in the embodiments of this application, both methods are considered comprehensively to determine the number of clusters in clustering, as Figure 4 shown to determine the number of clusters for all data.

[0133] Considering that photovoltaic output can be divided into two categories with maximized differences according to whether there is sunlight, in this study, the data is divided into two parts based on whether the 'total irradiance (W / ㎡)' is equal to 0, and the above methods are used to determine the number of classification clusters for these two parts respectively, as Figure 5a shown to determine the number of clusters for data with light, as Figure 5b shown to determine the number of clusters for data without light.

[0134] According to the above Figure 4 、 Figure 5a 、 Figure 5b , taking the number of classification clusters for all data as the main basis and the data with light and without light as corroboration, finally the number of classification clusters in this study is set to 5.

[0135] Using the KMeans clustering algorithm, set the parameter 'n_clusters' to 5. After clustering the prediction data, use the clustering results as new features and add them to the prediction set. Then call the LightGBM algorithm in the above research without adjusting the parameters for prediction. The prediction results are as follows: Figure 6 shown, the multimodal prediction model predicts the photovoltaic output results.

[0136] The experimental results show that after adding the multimodal classification model as new features, the prediction accuracy of the LightGBM model is significantly improved. This indicates that the additional information provided by the modal classification can effectively supplement the original features and improve the prediction performance of the model for photovoltaic output under complex meteorological conditions. This improvement strategy demonstrates the potential of combining multimodal classification with the main prediction model, providing a better solution for the photovoltaic output prediction task.

[0137] Above, in combination with the supplementary technical solutions, the meteorological prediction model and the new energy output prediction model trained in the embodiments of the present application, compared with the target prediction model for predicting power load trained in 201 - 208, the former two models can play a corroborative role and can provide input and auxiliary information in the training of the latter model. For the ultimate purpose, all are to improve the accuracy of the target prediction model for predicting power load.

[0138] Further, as an implementation of the above Figure 1 , Figure 2 shown method, the embodiments of the present application provide a user-side power load prediction device. The device embodiments correspond to the foregoing method embodiments. For ease of reading, the device embodiments will not repeat the details of the foregoing method embodiments one by one. However, it should be clear that the devices in this embodiment can correspondingly implement all the contents of the foregoing method embodiments. The device is applied to construct a more comprehensive user-side power prediction model by integrating three factors: time, meteorology, and historical power load, as specifically shown in Figure 7 shown, the device includes:

[0139] The first acquisition unit 31 is used to acquire historical meteorological record data, where the historical meteorological record data is daily meteorological record data within a preset time range. The daily meteorological record data is meteorological data at every preset unit time interval, and the meteorological data includes at least temperature, light irradiance, and wind speed; among them, the daily meteorological record data is correspondingly associated with the date and season;

[0140] The second acquisition unit 32 is used to acquire historical power load data, where the historical power load data is daily power load data within the preset time range. The daily power load data is power load data at every preset unit time interval;

[0141] A third acquisition unit 33, configured to establish an association between the historical meteorological record data and the historical power load data according to the same time feature, so as to obtain time series data sorted by date, where the time series data includes a sequence sorting between multiple combined data, and each of the combined data includes data in two dimensions of meteorology and power load;

[0142] A processing unit 34, configured to perform training processing on the time series data by using a preset machine learning architecture, so as to construct a target prediction model applied to predicting the power load on the user side;

[0143] A prediction unit 35, configured to process the acquired current meteorological record data and current power load data by using the target prediction model, and output a prediction result of the power load on the user side in a future time range.

[0144] Further, as Figure 8 shown, the preset machine learning architecture is an XGBoost framework, and the processing unit 34 includes:

[0145] A processing module 341, configured to perform feature processing on the time series data to obtain feature data corresponding to each of the combined data in the time series data, where the feature data at least includes: converting the description information in the meteorological data into a numerical representation;

[0146] An adding module 342, configured to add a sorting serial number to the feature data corresponding to different combined data according to the sorting corresponding to different combined data in the time series data, so as to obtain a unique sorting serial number corresponding to each of the feature data;

[0147] A first determination module 343, configured to determine a data sequence composed of the feature data according to the sorting sequence of the sorting serial numbers;

[0148] A second determination module 344, configured to continuously slide a preset sliding window on the data sequence at a preset interval, so as to determine multiple groups of input features and output labels, where the input features are the feature data currently selected by the preset sliding window, and the output labels are the next adjacent feature data of the feature data currently selected by the preset sliding window;

[0149] A construction module 345, configured to perform training processing on the multiple groups of input features and output labels by using a preset XGBoost framework, so as to construct a prediction model applied to predicting the power load on the user side.

[0150] Further, as Figure 8 shown, after acquiring the historical meteorological record data, the device further includes:

[0151] A dividing unit 36, configured to extract first meteorological record data corresponding to severe meteorological features by performing feature analysis and processing on the historical meteorological record data;

[0152] The dividing unit 36 is further configured to determine, in the historical meteorological record data, other data except the first meteorological record data as second meteorological record data;

[0153] A first training unit 37, configured to use a vector autoregressive framework to train a meteorological prediction model with the first meteorological record data as positive samples and the second meteorological record data as negative samples;

[0154] A first execution unit 38, configured to apply meteorological prediction data output by the meteorological prediction model to the target prediction model to predict future user-side power loads.

[0155] Further, as Figure 8 shown, during the process of training the meteorological prediction model, the first training unit 37 is further specifically configured to:

[0156] If the difference in the amount of data between the first meteorological record data and the second meteorological record data is greater than a preset threshold;

[0157] During the process of training the meteorological prediction model, adjust the recognition threshold of the classifier for the samples with a smaller number in the first meteorological record data and the second meteorological record data.

[0158] Further, as Figure 8 shown, after obtaining the historical power load data, the device further includes:

[0159] An extraction unit 39, further configured to extract target historical data corresponding to the influence of new energy output by analyzing and processing the historical power load data, where the new energy output at least includes wind energy output and photovoltaic output;

[0160] A second training unit 310, configured to perform training processing on the target historical data using a preset LightGBM framework to construct a new energy output prediction model;

[0161] A second execution unit 311, configured to apply the output power prediction data of the new energy output prediction model to the target prediction model to predict future user-side power loads.

[0162] The user-side power load prediction device includes a processor and a memory. The above first acquisition unit, second acquisition unit, third acquisition unit, processing unit, prediction unit, etc. are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions.

[0163] The processor contains a kernel, which retrieves the corresponding program units from the memory. One or more kernels can be set. By adjusting the kernel parameters, time series data is constructed in two dimensions of meteorology and power load, and then machine learning is used for training. Thus, considering the three factors of time, meteorology, and historical power load, a more comprehensive user-side power prediction model is constructed, providing an effective user-side power load prediction solution.

[0164] An embodiment of the present application provides a storage medium with a program stored thereon. When the program is executed by a processor, the user-side power load prediction method is implemented.

[0165] An embodiment of the present application provides a processor for running a program. When the program runs, the user-side power load prediction method is executed.

[0166] The present application also provides a computer program product, which is suitable for executing a program initialized with the steps of the user-side power load prediction method when executed on a data processing device.

[0167] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0168] In a typical configuration, the device includes one or more processors (CPUs), a memory, and a bus. The device may also include an input / output interface, a network interface, etc.

[0169] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory includes at least one storage chip. The memory is an example of computer-readable media.

[0170] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0171] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0172] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0173] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for predicting user-side power load, characterized in that, The method includes: Obtaining historical meteorological record data, where the historical meteorological record data is daily meteorological record data within a preset time range, and the daily meteorological record data includes meteorological data at preset unit time intervals, and the meteorological data includes at least temperature, humidity, light irradiance, and wind speed; wherein, the daily meteorological record data is correspondingly associated with a date and a season; Obtaining historical power load data, where the historical power load data is daily power load data within the preset time range, and the daily power load data includes power load data at preset unit time intervals; Establishing an association between the historical meteorological record data and the historical power load data according to the same time characteristics to obtain time series data sorted by date, and the time series data includes a sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load; Performing training processing on the time series data by using a preset machine learning architecture to construct a target prediction model applied to predicting the power load on the user side; Processing the obtained current meteorological record data and current power load data by using the target prediction model, and outputting a prediction result of the power load on the user side in a future time range.

2. The method according to claim 1, characterized in that The preset machine learning architecture is the XGBoost framework, and performing training processing on the time series data by using the preset machine learning architecture to construct a target prediction model applied to predicting the power load on the user side includes: Performing feature processing on the time series data to obtain feature data corresponding to each combined data in the time series data, and the feature data at least includes: converting the description information in the meteorological data into a numerical representation; Adding a sorting serial number to the feature data corresponding to different combined data according to the sorting corresponding to different combined data in the time series data, so as to obtain a unique sorting serial number corresponding to each feature data; Determining a data sequence composed of the feature data according to the sorting sequence of the sorting serial numbers; By continuously sliding a preset sliding window on the data sequence at a preset interval, determining multiple groups of input features and output labels, where the input feature is the feature data currently selected by the preset sliding window, and the output label is the next adjacent feature data of the feature data currently selected by the preset sliding window; Performing training processing on the multiple groups of input features and output labels by using the preset XGBoost framework to construct a prediction model applied to predicting the power load on the user side.

3. The method according to claim 1, wherein After obtaining the historical meteorological record data, the method further includes: Performing feature analysis and processing on the historical meteorological record data to extract the first meteorological record data corresponding to the severe meteorological features; In the historical meteorological record data, determining other data except the first meteorological record data as the second meteorological record data; Using a vector autoregressive framework to train a meteorological prediction model with the first meteorological record data as positive samples and the second meteorological record data as negative samples. Apply the meteorological prediction data output by the meteorological prediction model to the target prediction model to predict the future user-side power load.

4. The method according to claim 3, characterized in that, During the process of training the meteorological prediction model, the method further includes: If the data volume difference between the first meteorological record data and the second meteorological record data is greater than a preset threshold; During the process of training the meteorological prediction model, adjust the recognition threshold of the classifier for the samples with a smaller number in the first meteorological record data and the second meteorological record data.

5. The method according to claim 1, characterized in that, After obtaining the historical power load data, the method further includes: Through analyzing and processing the historical power load data, extract the target historical data corresponding to the influence of new energy output, where the new energy output at least includes wind energy output and photovoltaic output; Use the preset LightGBM framework to train and process the target historical data to build a new energy output prediction model; Apply the output power prediction data of the new energy output prediction model to the target prediction model to predict the future user-side power load.

6. A user-side power load forecasting device, characterized in that, The device includes: A first acquisition unit, configured to acquire historical meteorological record data, where the historical meteorological record data is daily meteorological record data within a preset time range, and the daily meteorological record data is meteorological data at every preset unit time interval, and the meteorological data at least includes temperature, humidity, light irradiance, and wind speed; wherein, the daily meteorological record data is associated with the corresponding date and season; A second acquisition unit, configured to acquire historical power load data, where the historical power load data is daily power load data within the preset time range, and the daily power load data is power load data at every preset unit time interval; A third acquisition unit, configured to establish an association between the historical meteorological record data and the historical power load data according to the same time feature to obtain time series data sorted by date, where the time series data includes the sequence sorting between multiple combined data, and each combined data includes data in two dimensions of meteorology and power load; A processing unit, configured to use a preset machine learning architecture to train and process the time series data to build a target prediction model for predicting the user-side power load; A prediction unit, configured to use the target prediction model to process the acquired current meteorological record data and current power load data, and output a prediction result of the user-side power load in a future time range.

7. The device according to claim 6, wherein The preset machine learning architecture is the XGBoost framework, and the processing unit includes: A processing module, configured to perform feature processing on the time series data to obtain the feature data corresponding to each combined data in the time series data, where the feature data at least includes: converting the description information in the meteorological data into a numerical representation; An adding module, configured to add a sorting serial number to the feature data corresponding to different combined data according to the sorting corresponding to different combined data in the time series data, so as to obtain a unique sorting serial number corresponding to each feature data; A first determination module, configured to determine a data sequence composed of the feature data according to the sorting order of the sorting serial numbers; A second determination module, configured to continuously slide a preset sliding window on the data sequence at a preset interval to determine multiple groups of input features and output labels, where the input features are the feature data currently selected by the preset sliding window, and the output label is the next adjacent feature data of the feature data currently selected by the preset sliding window; A construction module, configured to perform training processing on the multiple groups of input features and output labels by using a preset XGBoost framework to construct a prediction model applied to predicting the user-side power load.

8. The device according to claim 6, characterized in that, After obtaining the historical meteorological record data, the device further includes: A division unit, configured to extract first meteorological record data corresponding to meteorological severe features by performing feature analysis processing on the historical meteorological record data; The division unit is further configured to determine, in the historical meteorological record data, other data except the first meteorological record data as second meteorological record data; A first training unit, configured to use a vector autoregressive framework to train a meteorological prediction model with the first meteorological record data as positive samples and the second meteorological record data as negative samples; A first execution unit, configured to apply the meteorological prediction data output by using the meteorological prediction model to the target prediction model to predict the future user-side power load.

9. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for predicting the user-side power load according to any one of claims 1-5 is implemented.

10. An electronic device, characterized in that, The device includes at least one processor, and at least one memory and a bus connected to the processor; Wherein, the processor and the memory complete communication with each other through the bus; The processor is configured to call program instructions in the memory to execute the method for predicting the user-side power load according to any one of claims 1-5.

Images