Self-learning prediction method and device for power system load and storage medium

Abstract

The invention discloses a power system load self-learning prediction method and device and a storage medium. The method relates to the field of artificial intelligence, and comprises the following steps: receiving influence factor data of a power system to be subjected to load prediction; a target load prediction model is obtained, and the target load prediction model is a model obtained by adjusting a design matrix in the initial load prediction model according to the influence factor data; and inputting the influence factor data into the target load prediction model, and predicting a load prediction result of the power system. According to the invention, the technical problem of low accuracy of power system load prediction in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular, to a self-learning prediction method, device, and storage medium for power system load. Background Art

[0002] Power system load forecasting is one of the key tasks in the operation and management of power systems. Accurate load forecasting can not only help power companies optimize the allocation of power resources, ensure the stability and reliability of power supply, but also effectively reduce operating costs and improve energy utilization efficiency. With the continuous expansion of the scale of power systems and the increasing complexity of power demand, the accuracy and timeliness of load forecasting become particularly important.

[0003] In the prior art, power system load forecasting methods are mainly divided into two categories: statistical methods and machine learning methods. Statistical methods mainly include time series analysis (such as autoregressive models, moving average models, autoregressive moving average models and their extended models such as autoregressive integrated moving average models, etc.) and exponential smoothing methods (such as first-order, second-order, and third-order exponential smoothing methods, etc.). These methods predict through the trends, seasonality, and periodic changes of historical load data. Statistical methods have the advantages of solid theoretical foundation and high computational efficiency, but in dealing with complex non-linear relationships and multi-variable influencing factors, the prediction accuracy is often limited. Machine learning methods mainly include methods such as support vector machines, decision trees, random forests, and artificial neural networks. Machine learning methods have significant advantages in dealing with large-scale data and complex pattern recognition, but usually require a large amount of training data and computational resources, and the interpretability of the models is poor.

[0004] Although the prior art has improved the prediction accuracy to a certain extent, there are still the following deficiencies: First, the adaptability is insufficient. The power system load is affected by various factors such as weather, economic activities, and population changes. The dynamic changes of these factors make the load pattern have strong time-variability. The prior art shows insufficient adaptability in dealing with this dynamic change and is difficult to update model parameters in real time. Second, the ability to handle non-linear relationships is limited. There are complex non-linear relationships between power load and influencing factors. Although machine learning methods such as neural networks can capture these non-linear relationships to a certain extent, their training process is complex and affects the prediction performance. Third, the model complexity and computational cost are high. High-precision prediction usually requires more complex models and higher computational costs. In application scenarios with high real-time requirements, the prior art is difficult to balance prediction accuracy and computational efficiency. Fourth, the interpretability is poor: Although many machine learning methods have high prediction accuracy, the black-box nature of the models makes their results lack interpretability, which may lead to a decrease in trust in practical applications and limit the promotion and application of the models.

[0005] In summary, when the existing power system load forecasting methods are used for real-time load forecasting, it is difficult to respond to new data inputs in a timely manner, lack the ability to adaptively adjust model parameters based on new data, and cannot effectively cope with the rapid changes in load patterns and comprehensively reflect long-term dynamic trends. As a result, when faced with complex and dynamically changing data, the existing technologies have insufficient adaptability and thus lower accuracy.

[0006] For the above problems, no effective solutions have been proposed yet. Summary of the Invention

[0007] Embodiments of the present invention provide a self-learning prediction method, device, and storage medium for power system load, so as to at least solve the technical problem of low accuracy in power system load forecasting in the prior art.

[0008] To achieve the above object, according to one aspect of the present application, there is provided a self-learning prediction method for power system load, the method including: receiving influence factor data for power system load forecasting to be performed; obtaining a target load forecasting model, where the target load forecasting model is a model obtained by adjusting a design matrix in an initial load forecasting model according to the influence factor data; and inputting the influence factor data into the target load forecasting model to predict a load forecasting result of the power system.

[0009] Further, obtaining the target load forecasting model includes: obtaining a load forecasting generalized additive model, where the load forecasting generalized additive model is used to predict load data based on the influence factor data; converting the load forecasting generalized additive model into a state space model to determine the initial load forecasting model; and adjusting the design matrix in the initial load forecasting model according to the influence factor data to obtain the target load forecasting model.

[0010] Further, obtaining the load forecasting generalized additive model includes: obtaining historical influence factor data and historical load data; determining a plurality of basis functions according to the feature types of the historical influence factor data; determining a smoothing function and a basis function matrix according to the plurality of basis functions, where the smoothing function includes the basis function and the coefficients of the basis function; splicing the basis function matrix to determine the design matrix; using the historical influence factor data as a feature variable and the historical load data as a target variable, and determining the load forecasting generalized additive model according to the smoothing function.

[0011] Further, converting the load forecasting generalized additive model into a state space model to determine the initial load forecasting model includes: determining a state vector and a state transition equation of the state space model according to the coefficients of the basis function in the load forecasting generalized additive model; determining an observation equation of the state space model according to the design matrix, the state vector, and the state transition equation; and determining the initial load forecasting model according to the state vector, the state transition equation, and the observation equation.

[0012] Further, after inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system, it further includes: receiving the real load data of the power system, and updating the target load prediction model according to the real load data, the load prediction result, and the influencing factor data.

[0013] Further, updating the target load prediction model according to the real load data, the load prediction result, and the influencing factor data includes: obtaining the historical influencing factor data corresponding to the preset recursive model and the target load prediction model; updating the historical influencing factor data according to the influencing factor data; updating the model parameters of the target load prediction model according to the updated historical influencing factor data, the real load data, the load prediction result, and the preset recursive model.

[0014] Further, before updating the historical influencing factor data according to the influencing factor data, it further includes: obtaining the feature type of the historical influencing factor data as the historical feature type; determining whether the feature types of the influencing factor data all belong to the historical feature type. If there are feature types in the feature types of the influencing factor data that do not belong to the historical feature type, determining the newly emerged feature types according to the influencing factor data and the historical feature type; determining the basis function corresponding to the newly emerged feature types as the new basis function, and updating the design matrix of the target load prediction model according to the new basis function.

[0015] Further, updating the historical influencing factor data according to the influencing factor data includes: if the feature types of the influencing factor data all belong to the historical feature type, writing the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data; if there are feature types in the feature types of the influencing factor data that do not belong to the historical feature type, adding the newly emerged feature types to the historical influencing factor data, and writing the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data.

[0016] To achieve the above object, according to another aspect of the present application, a self-learning prediction method for the load of a power system is provided, including: receiving a prediction instruction triggered by a client, where the prediction instruction is used to indicate to perform a load prediction on the power system; in a cloud server, in response to the prediction instruction, receiving the influencing factor data of the power system to be load predicted; obtaining a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system; and returning the load prediction result to the client.

[0017] To achieve the above object, according to another aspect of the present application, there is provided a self-learning prediction device for the load of a power system, the device comprising: a data acquisition unit for receiving data of influencing factors for load prediction of the power system; a model acquisition unit for acquiring a target load prediction model, wherein the target load prediction model is a model obtained by adjusting a design matrix in an initial load prediction model according to the data of influencing factors; and a load prediction unit for inputting the data of influencing factors into the target load prediction model to predict the load prediction result of the power system.

[0018] According to another aspect of the present application, there is provided a computer-readable storage medium, the computer-readable storage medium comprising a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute any one of the self-learning prediction methods for the load of a power system.

[0019] According to another aspect of the present application, there is provided an electronic device, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include those for executing any one of the self-learning prediction methods for the load of a power system.

[0020] According to another aspect of the present application, there is provided a computer program product, comprising computer instructions, which when executed by a processor implement the steps of the self-learning prediction method for the load of a power system as described in any one of the above.

[0021] In the embodiments of the present invention, by receiving data of influencing factors for load prediction of the power system; acquiring a target load prediction model, wherein the target load prediction model is a model obtained by adjusting a design matrix in an initial load prediction model according to the data of influencing factors; and inputting the data of influencing factors into the target load prediction model to predict the load prediction result of the power system, the technical problem of low accuracy in load prediction of the power system in the prior art is solved.

[0022] By adjusting the design matrix in the initial load prediction model according to the data of influencing factors to obtain the target load prediction model, using the design matrix as the carrier for updating and adjusting according to the data of influencing factors, the target load prediction model can be made to adaptively learn based on the data of influencing factors by adjusting the design matrix, so that the target load prediction model can self-learn the long-term change trend of the data of influencing factors and make corresponding dynamic adjustments, enhancing the adaptability of the model to the dynamic changes of influencing factors, thereby improving the accuracy of load prediction. Description of the Drawings

[0023] The accompanying drawings described herein are used to provide a further understanding of the present invention and form a part of this application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0024] Figure 1 shows a hardware structure block diagram of a computer terminal for implementing a self-learning prediction method for power system loads;

[0025] Figure 2 is a flowchart of a self-learning prediction method for power system loads provided according to an embodiment of the present application;

[0026] Figure 3 is a flowchart of obtaining a target load prediction model in an optional self-learning prediction method for power system loads provided according to an embodiment of the present application;

[0027] Figure 4 is a schematic diagram of a cyclic iterative prediction process of a self-learning prediction method for power system loads provided according to an embodiment of the present application;

[0028] Figure 5 is a schematic diagram of a self-learning prediction device for power system loads provided according to an embodiment of the present application;

[0029] Figure 6 is a flowchart of another self-learning prediction method for power system loads provided according to an embodiment of the present application;

[0030] Figure 7 is a structure block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0031] In order to enable those skilled in the art of the present technology to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] It should be noted that the terms "first", "second", etc. in the specification, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily limit to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0033] It should be noted that the relevant information and data involved in this application (including but not limited to data for training, data for analysis, etc.) are all information and data authorized by users or fully authorized by all parties. For example, an interface is set between this system and relevant users or institutions. Before obtaining relevant information, a request for obtaining needs to be sent to the aforementioned users or institutions through the interface, and after receiving the consent information feedback from the aforementioned users or institutions, the relevant information is obtained.

[0034] Embodiment 1

[0035] According to an embodiment of the present invention, an embodiment of a self-learning prediction method for the load of a power system is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0036] The method embodiment provided in the first embodiment of this application can be executed on a mobile terminal, a computer terminal or a similar computing device. Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a self-learning prediction method for the load of a power system is shown. As Figure 1 shown, the computer terminal 10 (or mobile device 10) may include one or more processors 102 (shown as 102a, 102b,..., 102n in the figure) (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. Those of ordinary skill in the art can understand, Figure 1The structure shown is only illustrative and does not limit the structure of the above electronic device. For example, computer terminal 10 may also include more or fewer components than those shown in Figure 1 or have a different configuration from that shown in Figure 1 .

[0037] It should be noted that one or more of the above processors 102 and / or other data processing circuits can generally be referred to as "data processing circuits" herein. The data processing circuit can be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit can be a single independent processing module, or be incorporated in whole or in part into any one of the other elements in computer terminal 10 (or mobile device). As involved in the embodiments of the present application, the data processing circuit is a kind of processor control (such as the selection of a variable resistor terminal path connected to an interface).

[0038] Memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the self-learning prediction method for the power system load in the embodiments of the present invention. Processor 102 executes various functional applications and data processing by running the software programs and modules stored in memory 104, that is, implements the above self-learning prediction method for the power system load. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, memory 104 may further include a memory remotely disposed relative to processor 102, and these remote memories can be connected to computer terminal 10 through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0039] Transmission device 106 is used to receive or send data via a network. Specific examples of the above network may include the wireless network provided by the communication provider of computer terminal 10. In one instance, transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0040] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables a user to interact with the user interface of computer terminal 10 (or mobile device).

[0041] Under the above operating environment, the present application provides as Figure 2Self-learning prediction method for power system load as shown Figure 2 It is a flowchart of the self-learning prediction method for power system load according to Embodiment 1 of the present invention

[0042] Step S201, receive the influencing factor data of the power system to be load predicted

[0043] Optionally, the influencing factor data can be the input data used by the power system to predict the load. The power system load is affected by multiple factors, and the characteristic types of the influencing factor data can include time characteristics, weather characteristics, economic characteristics, etc. Time characteristics can include date, season, etc. Weather characteristics can include temperature, humidity, rainfall, etc., and economic characteristics can include whether it is a working day, etc

[0044] For example, the influencing factor data of the power system to be load predicted is: March 5, 2025, temperature 15 degrees Celsius, sunny weather, working day

[0045] Optionally, after obtaining the influencing factor data, the method of the embodiment of the present application can also preprocess the influencing factor data. The above preprocessing can be data cleaning, missing value processing, outlier detection and processing, data normalization, feature engineering. Preprocessing the influencing factor data can ensure that the data input into the target load prediction model is complete and of a unified scale, thereby helping to improve the accuracy of the prediction

[0046] Step S202, obtain the target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data

[0047] Optionally, the initial load prediction model can be the model in the state at the previous moment of the time where the target load prediction model is located. For example, the time where the influencing factor data obtained in Step S201 is located is March 5, 2025, that is, the target load prediction model is going to predict the load on March 5, 2025. In the case of a one-day time interval, the corresponding initial load prediction model of this target load prediction model is a model constructed based on the historical data before March 4, 2025 (historical data includes historical influencing factor data and historical load data)

[0048] Optionally, the initial load prediction model includes a design matrix, which is a two-dimensional data structure constructed based on historical influencing factor data and is used to characterize the changes in the influencing factors corresponding to the initial load prediction model. For example, if the initial load prediction model is constructed based on the historical data from January 1, 2025 to March 4, 2025, and each piece of influencing factor data includes 3 features, then the design matrix in the initial load prediction model can be a matrix with 63 rows and 3 columns, where each row represents a historical data sample and each column corresponds to a feature.

[0049] Optionally, adjusting the design matrix in the initial load prediction model according to the influencing factor data can be adding the influencing factor data obtained in step S201 to the design matrix to make it a matrix with 64 rows and 3 columns, so as to reflect the latest information on March 5, 2025. In other words, the design matrix is a bridge connecting the influencing factor data and the load prediction model (target / initial load prediction model). By means of the design matrix, the long-term dynamic change law of the influencing factor data can be effectively reflected, the adaptability of the load prediction is improved, enabling it to be adjusted in real time with new data, and thus the accuracy of the load prediction is improved.

[0050] Step S203: Input the influencing factor data into the target load prediction model to predict the load prediction result of the power system.

[0051] Optionally, the initial load prediction model can also include a parameter mechanism and a prediction mechanism. The parameter mechanism is used to learn the change laws of the historical influencing factor data and the historical load data, and the prediction mechanism is used to continue predicting the load corresponding to the influencing factor data (i.e., the influencing factor data obtained in step S201) based on the parameter mechanism. After adjusting the design matrix in the initial load prediction model by using the influencing factor data, a target load prediction model including the updated design matrix, the parameter mechanism, and the prediction mechanism is obtained. For example, input the historical data (the historical data includes historical influencing factor data and historical load data) into the Kalman filter algorithm to use this algorithm as the basis for the parameter mechanism and the prediction mechanism, and use the design matrix as the coefficient of the state vector in the observation equation of this algorithm to integrate the design matrix into the parameter mechanism and the prediction mechanism to obtain the initial load prediction model. In addition, the recursive least squares method or the particle filter algorithm can also be selected as the basis for the parameter mechanism and the prediction mechanism. It should be noted that the improvement of this application lies in setting the design matrix in the model to self-learn the long-term dynamic changes of the influencing factor data. Those skilled in the art can use the recursive least squares method or the particle filter algorithm to establish a prediction model according to their needs to implement the parameter mechanism and the prediction mechanism on the premise of the above disclosure of this application, and this application will not elaborate on this.

[0052] In summary, in the present application, by adjusting the design matrix in the initial load prediction model according to the influencing factor data to obtain the target load prediction model, and using the design matrix as the carrier for updating and adjusting the influencing factor data, the target load prediction model can be made to adaptively learn based on the influencing factor data by adjusting the design matrix. As a result, the target load prediction model can self-learn the long-term change trend of the influencing factor data and make corresponding dynamic adjustments, enhancing the adaptability of the model to the dynamic changes of the influencing factors, thereby improving the accuracy of load prediction.

[0053] To improve the accuracy of power system load prediction, Figure 3 is a flowchart for obtaining the target load prediction model in the optional self-learning prediction method for power system load provided by an embodiment of the present application. Refer to Figure 3 As shown, optionally, in the self-learning prediction method for power system load provided by an embodiment of the present application, obtaining the target load prediction model includes:

[0054] S301, obtain a load prediction generalized additive model, where the load prediction generalized additive model is used to predict load data based on influencing factor data.

[0055] Optionally, obtain the historical load data and historical influencing factor data of the power system, and establish a load prediction generalized additive model according to the historical load data and historical influencing factor data to achieve predicting load data based on influencing factor data. The Generalized Additive Model (GAM) is a flexible regression model that can capture the non-linear relationship between each independent variable (i.e., influencing factor data) and the response variable (i.e., load data) through a smoothing function, enabling the prediction of load data based on influencing factor data to be extended to non-linear rules. At the same time, the smoothing function mechanism of the generalized additive model makes it possible to increase the characteristic types of historical influencing factor data.

[0056] For example, if there are 3 types of characteristic types of historical influencing factor data, specifically date, temperature, and humidity, construct corresponding smoothing functions for each characteristic type and establish a load prediction generalized additive model to learn the non-linear relationship between influencing factor data and load data. If the characteristic types of influencing factor data are date, temperature, humidity, and rainfall, only need to construct a new smoothing function for the rainfall characteristic alone to achieve an increase in characteristic types.

[0057] S302, convert the load prediction generalized additive model into a state space model to determine the initial load prediction model.

[0058] Optionally, the initial load prediction model includes a state vector, a state transition equation, and an observation equation. The observation equation includes a design matrix. The design matrix is a two-dimensional data structure constructed based on historical influencing factor data. For example, if there are 63 pieces of historical influencing factor data, and each piece of influencing factor data includes 3 features, then the design matrix can be a matrix with 63 rows and 3 columns. Each row represents a historical data sample, and each column corresponds to a feature. The coefficients of the load prediction generalized additive model are used as the state vector to transform the load prediction generalized additive model into a state space model, and the state transition equation and the observation equation are defined. The state transition equation is used to characterize the change of the state variable over time, and the observation equation is used to characterize the relationship between the design matrix, the state variable, and the load data.

[0059] For example, the observation equation is that the load data is equal to the product of the state variable and the design matrix. Optionally, an observation noise is also set in the observation equation, that is, first calculate the product of the state variable and the design matrix, and then superimpose the observation noise on this product to obtain the load data.

[0060] S303. Adjust the design matrix in the initial load prediction model according to the influencing factor data to obtain the target load prediction model.

[0061] Optionally, adjusting the design matrix in the initial load prediction model according to the influencing factor data can be adding the influencing factor data to the design matrix. For example, the design matrix in the initial load prediction model is a matrix with 63 rows and 3 columns before adjustment, and after adjustment, it becomes a matrix with 64 rows and 3 columns to add the latest information of the influencing factor data. If a new feature type appears in the influencing factor data, the design matrix is adjusted to a matrix with 64 rows and 4 columns. For example, add the new feature of the influencing factor data to the 64th row and 4th column, and fill the 4th column of the first 63 rows with 0 values.

[0062] In summary, by constructing the load prediction generalized additive model, the subsequent established initial load prediction model can learn the non-linear law between the influencing factor data and the load data, expands the type of laws that the initial load prediction model can adapt to, enhances the interpretability of the model. Even if the relationship between the influencing factor data and the load data is non-linear, this application can still make effective predictions, avoiding the common prediction biases in linear models. At the same time, through the smoothing function and the design matrix, when new feature types appear in the influencing factor data over time, the initial load prediction model can also be adaptively adjusted, improving the adaptability to the dynamic environment and the prediction accuracy. In addition, the mode of this application based on the load prediction generalized additive model to learn the non-linear law between the influencing factors and the load data significantly reduces the computational complexity compared with the existing machine learning or neural network methods, and improves the real-time performance of load prediction.

[0063] To improve the accuracy of power system load forecasting, optionally, obtaining a generalized additive model for load forecasting includes: obtaining historical influencing factor data and historical load data; determining multiple basis functions according to the characteristic types of the historical influencing factor data; determining a smoothing function and a basis function matrix according to the multiple basis functions, where the smoothing function includes the basis function and the coefficients of the basis function; splicing the basis function matrix to determine a design matrix; using the historical influencing factor data as a feature variable and the historical load data as a target variable, and determining the generalized additive model for load forecasting according to the smoothing function.

[0064] For example, obtaining the influencing factor data within a past period as the historical influencing factor data, obtaining the real load data as the historical load data, determining the type of basis function (B-spline function or polynomial spline function), taking each characteristic type of the historical influencing factor data as a feature variable, assuming there are m feature variables X 1，t 、X 2，t 、X 3，t 、……、X m，t , constructing multiple basis functions for each feature variable to form a basis function matrix and a smoothing function, so as to construct a generalized additive model for load forecasting. Horizontally splicing the basis function matrices of all feature variables to construct a design matrix. Optionally, an intercept term is also set in the design matrix, that is, first obtaining a basic design matrix, where the first column of the basic design matrix is all constants 1 to represent the intercept, and then splicing the splicing result of the basis function matrices of all feature variables with the basic design matrix to form a final design matrix.

[0065] The generalized additive model for load forecasting is shown in the following formula:

[0066] Y t =β0+f1(X 1，t )+f2(X 2，t )+…+f M (X M，t )+∈ t

[0067] Wherein, Y t represents the load data at time t, β0 represents the intercept term, f M (X M，t ) represents the smoothing function, and ∈ t represents the random error at time t.

[0068] The smoothing function is:

[0069]

[0070] Wherein, f m (X m，t ) represents the m-th smoothing function at time t, B m，k (Xm，t ) represents the basis function matrix of the k-th basis function, and θ m，k，t represents the coefficient of the k-th basis function, and K m represents the total number of basis functions. Thus, the generalized additive model for load forecasting can also be expressed by the following formula:

[0071]

[0072] The design matrix can be as shown in the following formula:

[0073]

[0074] where B(X t ) represents the design matrix, represents the K M -th basis function matrix of the basis function.

[0075] In summary, by determining the basis function according to the feature type, further constructing the smoothing function and the design matrix, and finally determining the generalized additive model for load forecasting, the relationship between the influencing factor data and the load data is mapped to the non-linear law, enhancing the interpretability of the load forecasting and facilitating the improvement of the accuracy of the load forecasting.

[0076] Optionally, to improve the conversion effect from the generalized additive model to the state space model, the generalized additive model for load forecasting is converted into the state space model to determine the initial load forecasting model, including: determining the state vector and the state transition equation of the state space model according to the coefficients of the basis functions in the generalized additive model for load forecasting; determining the observation equation of the state space model according to the design matrix, the state vector and the state transition equation; and determining the initial load forecasting model according to the state vector, the state transition equation and the observation equation.

[0077] The state vector is the coefficients of the basis functions of all feature variables, specifically as follows:

[0078]

[0079] where X t represents the state vector, represents the coefficient of the K M -th basis function in the M-th smoothing function. K is equal to Therefore, X t is a K×1 vector.

[0080] Assuming that all coefficients follow independent random walks, the following state transition equation can be obtained:

[0081] x t = x t-1 + w t

[0082] Among them, w t represents the state noise, N represents the normal distribution, and Q is a K×K state noise covariance matrix, which can be set as a diagonal matrix to assume the independence between coefficients.

[0083] To determine the observation equation of the state space model, the following formula can be referred to:

[0084]

[0085] Among them, Y t represents the load data at time t, β0 represents the intercept term, B(X t ) represents the design matrix, υ t represents the observation noise, and R represents the variance of the observation noise.

[0086] The above state space model is a representation in principle. Correspondingly, when making predictions, that is, when inputting the influencing factor data into the target load prediction model, the load prediction result of the power system can be predicted by referring to the following formula:

[0087]

[0088] Among them, B(X t+1 ) represents the adjusted design matrix, β0 represents the intercept term, that is, the influencing factor data is the data at time t + 1, represents the load prediction result corresponding to the influencing factor data. represents the parameters of the initial load prediction model (that is, the latest parameters of the model after being updated using the historical data at time t).

[0089] In summary, based on the load prediction generalized additive model, the state vector, state transition equation, and observation equation are defined, and it is transformed into a state space model. That is, it is equivalent to introducing a self-learning mechanism on the basis of the generalized additive model, enabling the model parameters X t to continuously learn and adjust according to new data, so as to predict the load data. It can realize the real-time update and dynamic adjustment of the model parameters while maintaining the strong modeling ability of the generalized additive model, thereby ensuring the transformation effect from the generalized additive model to the state space model, and further improving the accuracy of load prediction.

[0090] To improve the adaptability of load prediction, optionally, after inputting the influencing factor data into the target load prediction model and predicting the load prediction result of the power system, it further includes: receiving the real load data of the power system, and updating the target load prediction model based on the real load data, load prediction result, and influencing factor data.

[0091] For example, the influencing factor data is the data at time t+1. After predicting the load prediction result at time t+1 using the target load prediction model, as time changes, when the power system obtains the real load data at time t+1, it receives the real load data at time t+1, and updates the target load prediction model based on the real load data at time t+1, the load prediction result, and the influencing factor data, so that the model can learn the latest information at time t+1 for subsequent iterative prediction.

[0092] Optionally, after receiving the real load data of the power system and updating the target load prediction model based on the real load data, the load prediction result, and the influencing factor data, the method of the embodiment of the present application further includes: receiving the influencing factor data at the next moment (i.e., the influencing factor data at time t+2) as the first loop influencing factor data; obtaining the first loop load prediction model, where the first loop prediction model is a model obtained by adjusting the design matrix in the updated target load prediction model based on the first loop influencing factor data; inputting the first loop influencing factor data into the first loop load prediction model to predict the first loop load prediction result of the power system (i.e., predicting the load data at time t+2).

[0093] Figure 4 It is a schematic diagram of the cyclic iterative prediction process of the self-learning prediction method for the load of the power system provided by the embodiment of the present application. Refer to Figure 4 As shown, an initial load prediction model is constructed based on the historical data from time 0 to time t. On this basis, through the iterative process of prediction, update, and prediction, as time evolves, the model can be continuously updated according to new data to predict the load data at times t+3, t+4, and even more moments.

[0094] In summary, by updating the target load prediction model after obtaining the load prediction result, the model can be updated accordingly based on the continuously updated real load data to learn the latest information, ensuring the growth of the model, being applicable to predicting the load prediction results at different times, and thus improving the prediction adaptability.

[0095] To ensure the effect of model iterative update and the accuracy of iterative prediction, optionally, updating the target load prediction model based on the real load data, the load prediction result, and the influencing factor data includes: obtaining a preset recursive model and the historical influencing factor data corresponding to the target load prediction model; updating the historical influencing factor data based on the influencing factor data; updating the model parameters of the target load prediction model based on the updated historical influencing factor data, the real load data, the load prediction result, and the preset recursive model.

[0096] For example, the preset recursive model is a Kalman filter. During the construction of the initial load prediction model, for the parameters at the corresponding first moment, obtaining the preset recursive model at this time includes: initializing the initial state mean of the Kalman filter Initial state covariance P 0|0 ; determining the state transition matrix, observation matrix, state noise covariance Q, and observation noise covariance R of the Kalman filter according to the target load prediction model to obtain the preset recursive model. Correspondingly, when updating the target load prediction model, as time goes by, for the parameters at the t-th moment, obtaining the preset recursive model at this time includes: initializing the initial state mean of the Kalman filter Initial state covariance P t|t ; determining the state transition matrix, observation matrix, state noise covariance Q, and observation noise covariance R of the Kalman filter according to the target load prediction model to obtain the preset recursive model. Updating the model parameters of the target load prediction model according to the updated historical influencing factor data, real load data, load prediction results, and preset recursive model can refer to the following formula:

[0097]

[0098] P t+1|t = P t|t + Q

[0099]

[0100] P t+1|t+1 = (I - K t+1 B(X t+1 ) T )P t+1|t

[0101] where K t+1 represents the Kalman gain matrix, the weight matrix used to update the state vector at the t + 1 moment; R represents the covariance matrix of the observation noise; I represents the identity matrix, that is, a square matrix with all elements on the diagonal being 1; Q represents the state noise covariance matrix, represents the predicted value of the state vector at the t + 1 moment based on all information before the t moment, represents the estimated value of the state vector at the t moment based on all observations before the t moment, P t+1|t represents the covariance matrix of the prediction of the state vector at the t + 1 moment based on all information before the t moment, P t|t represents the covariance matrix of the estimation of the state vector at the t moment based on all observations before the t moment, represents the estimated value of the state vector at the t + 1 moment based on all observations before the t + 1 moment, Y t+1denotes the actual observed load data at time t + 1 (i.e., the true load data), β0 denotes the intercept term, B(X t+1 ) denotes the design matrix in the target load prediction model, P t+1|t+1 denotes the covariance matrix of the state vector estimate obtained based on all the observed data before time t + 1 at time t + 1.

[0102] After the update, it can be used to predict the load data at time t + 2 (i.e., the first-cycle load prediction result mentioned above). The prediction process can refer to the following formula:

[0103]

[0104] where, denotes the first-cycle load prediction result, B(X t+2 ) denotes the design matrix obtained by adjusting the design matrix in the target load prediction model according to the first-cycle influencing factor data, denotes the estimated value of the state vector obtained based on all the observations before time t + 1 at time t + 1.

[0105] In summary, by updating the model parameters in the target load prediction model, the model can continuously learn and adapt to new data, thereby improving the accuracy of power system load prediction. In addition, the present application adopts the method of combining a generalized additive model with a preset recursive model (such as a Kalman filter), captures the non-linear relationship between variables through a smoothing function, and then realizes the recursive update of the target load prediction model based on the preset recursive model. Therefore, it can efficiently respond to new data input and improve the real-time performance of load prediction.

[0106] To enhance the accuracy of load data prediction, optionally, before updating the historical influencing factor data according to the influencing factor data, it further includes: obtaining the feature type of the historical influencing factor data as the historical feature type; determining whether the feature types of the influencing factor data all belong to the historical feature type. If there are feature types in the feature types of the influencing factor data that do not belong to the historical feature type, determining the newly emerged feature type according to the influencing factor data and the historical feature type; determining the basis function corresponding to the newly emerged feature type as the new basis function, and updating the design matrix of the target load prediction model according to the new basis function.

[0107] For example, the feature types of the historical influencing factor data are date, temperature, and humidity. If the feature types of the influencing factor data are date, temperature, humidity, and rainfall, then rainfall is the newly emerged feature type. Set a new basis function for this feature, determine the corresponding new basis function matrix according to the new basis function, and add a new column in the design matrix to add the new basis function matrix corresponding to rainfall. If there is no newly emerged feature type, there is no need to add a new basis function.

[0108] In summary, by updating the design matrix according to the newly emerged feature types, it can be ensured that the model can not only handle the known influencing factors, but also adapt to the newly emerged feature types that may appear in the load forecasting of power systems, thereby enhancing the prediction ability and adaptability of the model, being suitable for dealing with the evolving load forecasting problems of power systems, and the model can learn and adapt to these changes in a timely manner to improve the prediction accuracy.

[0109] To ensure the update effect of the historical influencing factor data, optionally, updating the historical influencing factor data based on the influencing factor data includes: if the feature types of the influencing factor data all belong to the historical feature types, writing the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data; if there are feature types in the feature types of the influencing factor data that do not belong to the historical feature types, adding the newly emerged feature types to the historical influencing factor data and writing the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data.

[0110] For example, the feature types of the historical influencing factor data are date, temperature, and humidity. If the feature types of the influencing factor data are date, temperature, and humidity, then directly write the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data; if the feature types of the influencing factor data are date, temperature, humidity, and rainfall, and rainfall is the newly emerged feature type, then a column of rainfall data can be added to the historical influencing factor data, and the influencing factor data can be written into the historical influencing factor data to obtain the updated historical influencing factor data.

[0111] In summary, by distinguishing whether the feature types in the new data are already included in the historical influencing factor data and adopting different update strategies, it can be ensured that the load forecasting model can not only handle the changes in the known feature types, but also adapt to the introduction of the new feature types in a timely manner. This method improves the generalization ability of the historical influencing factor data and ensures the update effect of the historical influencing data.

[0112] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0113] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0114] Embodiment 2

[0115] According to an embodiment of the present invention, there is also provided a self-learning prediction device for the power system load for implementing the above self-learning prediction method for the power system load, as Figure 5 shown. The device includes: a data acquisition unit 501, a model acquisition unit 502, and a load prediction unit 503.

[0116] Specifically, the data acquisition unit 501 is configured to receive the influencing factor data of the power system to be load predicted;

[0117] The model acquisition unit 502 is configured to acquire a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data;

[0118] The load prediction unit 503 is configured to input the influencing factor data into the target load prediction model and predict the load prediction result of the power system.

[0119] The self-learning prediction device for the power system load provided by the embodiments of the present application receives the influencing factor data of the power system to be load predicted through the data acquisition unit 501; acquires a target load prediction model through the model acquisition unit 502, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; inputs the influencing factor data into the target load prediction model through the load prediction unit 503 and predicts the load prediction result of the power system, solving the technical problem of low accuracy of power system load prediction in the prior art, and thus achieving the technical effect of improving the accuracy of power system load prediction.

[0120] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, the model acquisition unit 502 includes: a generalized additive model acquisition module, configured to acquire a load prediction generalized additive model, where the load prediction generalized additive model is used to predict load data based on influencing factor data; a transformation module, configured to transform the load prediction generalized additive model into a state space model to determine an initial load prediction model; and an adjustment module, configured to adjust the design matrix in the initial load prediction model according to the influencing factor data to obtain a target load prediction model.

[0121] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, the generalized additive model acquisition module includes: a historical data acquisition sub-module, configured to acquire historical influencing factor data and historical load data; a basis function sub-module, configured to determine a plurality of basis functions according to the feature type of the historical influencing factor data; a smoothing function sub-module, configured to determine a smoothing function and a basis function matrix according to the plurality of basis functions, where the smoothing function includes the basis functions and the coefficients of the basis functions; a design matrix sub-module, configured to splice the basis function matrix to determine a design matrix; and a generalized additive model construction sub-module, configured to use the historical influencing factor data as feature variables, use the historical load data as target variables, and determine a load prediction generalized additive model according to the smoothing function.

[0122] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, the transformation module includes: a first transformation sub-module, configured to determine the state vector and the state transition equation of the state space model according to the coefficients of the basis functions in the load prediction generalized additive model; a second transformation sub-module, configured to determine the observation equation of the state space model according to the design matrix, the state vector, and the state transition equation; and a third transformation sub-module, configured to determine the initial load prediction model according to the state vector, the state transition equation, and the observation equation.

[0123] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, it further includes: an update unit, configured to, after inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system, receive the real load data of the power system, and update the target load prediction model according to the real load data, the load prediction result, and the influencing factor data.

[0124] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, the update unit includes: a recursive acquisition module, configured to acquire a preset recursive model and the historical influencing factor data corresponding to the target load prediction model; a factor update module, configured to update the historical influencing factor data according to the influencing factor data; and a parameter update module, configured to update the model parameters of the target load prediction model according to the updated historical influencing factor data, the real load data, the load prediction result, and the preset recursive model.

[0125] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, the updating unit further includes: a feature type determination module, configured to obtain the feature type of the historical influencing factor data as the historical feature type before updating the historical influencing factor data according to the influencing factor data; a new type module, configured to determine whether the feature types of the influencing factor data all belong to the historical feature types, and if there are feature types in the feature types of the influencing factor data that do not belong to the historical feature types, determine the newly emerged feature types according to the influencing factor data and the historical feature types; a new basis function module, configured to determine the basis function corresponding to the newly emerged feature types as the new basis function, and update the design matrix of the target load prediction model according to the new basis function.

[0126] Optionally, in the self-learning prediction device for power system load provided in the embodiments of the present application, the factor updating module includes: a first writing sub-module, configured to write the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data if the feature types of the influencing factor data all belong to the historical feature types; a second writing sub-module, configured to add the newly emerged feature types to the historical influencing factor data and write the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data if there are feature types in the feature types of the influencing factor data that do not belong to the historical feature types.

[0127] It should be noted here that the above data acquisition unit 501, model acquisition unit 502, and load prediction unit 503 correspond to steps S201 to S203 in Embodiment 1. The instances and application scenarios implemented by the three units and the corresponding steps are the same, but are not limited to the content disclosed in the above Embodiment 1. It should be noted that the above modules, as part of the device, can run in the computer terminal 10 provided in Embodiment 1.

[0128] Embodiment 3

[0129] According to the embodiments of the present application, another self-learning prediction method for power system load is further provided, as Figure 6 shown. The method includes:

[0130] S601, receiving a prediction instruction triggered by a client, where the prediction instruction is used to indicate performing a load prediction on the power system.

[0131] S602, in the cloud server, in response to the prediction instruction, receiving the influencing factor data of the power system to be load predicted; obtaining a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system.

[0132] S603, return the load forecasting result to the client.

[0133] In the cloud server, the self-learning prediction method for the power system load is the same as the method in Embodiment 1, and will not be elaborated here.

[0134] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0135] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0136] Embodiment 4

[0137] The embodiment of the present invention can provide a computer terminal, and this computer terminal can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the above computer terminal can also be replaced with a terminal device such as a mobile terminal.

[0138] Optionally, in this embodiment, the above computer terminal can be located in at least one network device among multiple network devices of a computer network.

[0139] In this embodiment, the above computer terminal can execute the program code of the following steps in the self-learning prediction method for the power system load: receive the influencing factor data of the power system to be load predicted; obtain a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; input the influencing factor data into the target load prediction model, and predict the load prediction result of the power system.

[0140] Optionally, the above computer terminal may execute the program code for the following steps in the self-learning prediction method for the power system load: Obtain a load prediction generalized additive model, where the load prediction generalized additive model is used to predict load data based on influencing factor data; convert the load prediction generalized additive model into a state space model to determine an initial load prediction model; adjust the design matrix in the initial load prediction model according to the influencing factor data to obtain a target load prediction model.

[0141] Optionally, the above computer terminal may execute the program code for the following steps in the self-learning prediction method for the power system load: Determine multiple basis functions according to the characteristic types of historical influencing factor data; determine a smoothing function and a basis function matrix according to the multiple basis functions, where the smoothing function includes the basis functions and the coefficients of the basis functions; splice the basis function matrix to determine a design matrix; use the historical influencing factor data as feature variables and the historical load data as target variables, and determine a load prediction generalized additive model according to the smoothing function.

[0142] Optionally, the above computer terminal may execute the program code for the following steps in the self-learning prediction method for the power system load: Determine the state vector and the state transition equation of the state space model according to the coefficients of the basis functions in the load prediction generalized additive model; determine the observation equation of the state space model according to the design matrix, the state vector, and the state transition equation; determine the initial load prediction model according to the state vector, the state transition equation, and the observation equation.

[0143] Optionally, the above computer terminal may execute the program code for the following steps in the self-learning prediction method for the power system load: Receive the real load data of the power system, and update the target load prediction model according to the real load data, the load prediction result, and the influencing factor data.

[0144] Optionally, the above computer terminal may execute the program code for the following steps in the self-learning prediction method for the power system load: Obtain the preset recursive model and the historical influencing factor data corresponding to the target load prediction model; update the historical influencing factor data according to the influencing factor data; update the model parameters of the target load prediction model according to the updated historical influencing factor data, the real load data, the load prediction result, and the preset recursive model.

[0145] Optionally, the above computer terminal may execute the program code of the following steps in the self-learning prediction method for power system load: obtaining the feature types of historical influencing factor data as historical feature types; determining whether the feature types of the influencing factor data all belong to the historical feature types, and if there are feature types among the feature types of the influencing factor data that do not belong to the historical feature types, determining the newly emerged feature types based on the influencing factor data and the historical feature types; determining the basis functions corresponding to the newly emerged feature types as new basis functions, and updating the design matrix of the target load prediction model based on the new basis functions.

[0146] Optionally, the above computer terminal may execute the program code of the following steps in the self-learning prediction method for power system load: if the feature types of the influencing factor data all belong to the historical feature types, writing the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data; if there are feature types among the feature types of the influencing factor data that do not belong to the historical feature types, adding the newly emerged feature types to the historical influencing factor data and writing the influencing factor data into the historical influencing factor data to obtain the updated historical influencing factor data.

[0147] Optionally, Figure 7 is a structural block diagram of an electronic device according to an embodiment of the present application. As Figure 7 shown, the electronic device may include: one or more ( Figure 7 only one is shown in the figure) processors 602, a memory 604, a storage controller, and a peripheral interface, wherein the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0148] Among them, the memory may be used to store software programs and modules, such as the program instructions / modules corresponding to the self-learning prediction method and device for power system load in the embodiments of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the above self-learning prediction method for power system load. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory may further include a memory remotely disposed relative to the processor, and these remote memories may be connected to the terminal through a network. Examples of the above network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0149] The processor may call the information and application programs stored in the memory through a transmission device to execute the above steps.

[0150] An embodiment of the present invention provides a self-learning prediction solution for the load of a power system. By receiving the influencing factor data of the power system to be load predicted; obtaining a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system, the technical problem of low accuracy in load prediction of the power system in the prior art is solved, and thus the technical effect of improving the accuracy of load prediction of the power system is achieved.

[0151] Those of ordinary skill in the art can understand that Figure 7 the structure shown is only schematic, and the computer terminal can also be a smart phone, a tablet computer, a palm computer, and terminal devices such as Mobile Internet Devices (MID), PAD, etc. Figure 7 It does not limit the structure of the above electronic device. For example, the computer terminal may further include more or fewer components (such as a network interface, a display device, etc.) than those shown in Figure 7 and may have a different configuration from that shown in Figure 7 the figure.

[0152] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware of the terminal device through a program, and the program can be stored in a computer-readable storage medium. The storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0153] Embodiment 5

[0154] An embodiment of the present invention further provides a computer-readable storage medium. Optionally, in this embodiment, the above computer-readable storage medium can be used to store the program code executed by the self-learning prediction method for the load of the power system provided in the first embodiment above.

[0155] Optionally, in this embodiment, the above computer-readable storage medium can be located in any one of the computer terminals in the computer terminal group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

[0156] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for performing the following steps: The above computer terminal may execute the program code for the following steps in the self-learning prediction method for the power system load: receiving the influencing factor data of the power system to be load-predicted; obtaining a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system.

[0157] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: obtaining a load prediction generalized additive model, where the load prediction generalized additive model is used to predict load data based on the influencing factor data; converting the load prediction generalized additive model into a state space model to determine an initial load prediction model; adjusting the design matrix in the initial load prediction model according to the influencing factor data to obtain a target load prediction model.

[0158] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: determining a plurality of basis functions according to the feature types of the historical influencing factor data; determining a smoothing function and a basis function matrix according to the plurality of basis functions, where the smoothing function includes the basis function and the coefficients of the basis function; splicing the basis function matrix to determine a design matrix; using the historical influencing factor data as feature variables and the historical load data as target variables, and determining a load prediction generalized additive model according to the smoothing function.

[0159] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: determining the state vector and the state transition equation of the state space model according to the coefficients of the basis function in the load prediction generalized additive model; determining the observation equation of the state space model according to the design matrix, the state vector, and the state transition equation; determining the initial load prediction model according to the state vector, the state transition equation, and the observation equation.

[0160] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: receiving the real load data of the power system, and updating the target load prediction model according to the real load data, the load prediction result, and the influencing factor data.

[0161] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: obtaining historical influencing factor data corresponding to a preset recursive model and a target load prediction model; updating the historical influencing factor data according to the influencing factor data; and updating model parameters of the target load prediction model according to the updated historical influencing factor data, real load data, load prediction results, and the preset recursive model.

[0162] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: obtaining the feature types of the historical influencing factor data as historical feature types; determining whether the feature types of the influencing factor data all belong to the historical feature types, and if there are feature types among the feature types of the influencing factor data that do not belong to the historical feature types, determining newly emerged feature types according to the influencing factor data and the historical feature types; determining basis functions corresponding to the newly emerged feature types as new basis functions, and updating the design matrix of the target load prediction model according to the new basis functions.

[0163] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: if the feature types of the influencing factor data all belong to the historical feature types, writing the influencing factor data into the historical influencing factor data to obtain updated historical influencing factor data; if there are feature types among the feature types of the influencing factor data that do not belong to the historical feature types, adding the newly emerged feature types to the historical influencing factor data and writing the influencing factor data into the historical influencing factor data to obtain updated historical influencing factor data.

[0164] The present application also provides a computer program product, which is suitable for executing a program of the self-learning prediction method steps of the power system load when executed on a data processing device.

[0165] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0166] In the above embodiments of the present invention, the descriptions of the various embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0167] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0168] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0169] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0170] If the above-mentioned 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 such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs and other various media that can store program codes.

[0171] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A self-learning prediction method for power system load, characterized in that, Including: Receiving the influencing factor data of the power system to be load predicted; Obtaining a target load prediction model, wherein the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influencing factor data; Inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system.

2. The method according to claim 1, characterized in that, Obtaining the target load prediction model includes: Obtaining a generalized additive model for load prediction, wherein the generalized additive model for load prediction is used to predict load data based on influencing factor data; Converting the generalized additive model for load prediction into a state space model to determine an initial load prediction model; Adjusting the design matrix in the initial load prediction model according to the influencing factor data to obtain a target load prediction model.

3. The method according to claim 2, characterized in that Obtaining the generalized additive model for load prediction includes: Obtaining historical influencing factor data and historical load data; Determining a plurality of basis functions according to the characteristic types of the historical influencing factor data; Determining a smoothing function and a basis function matrix according to the plurality of basis functions, wherein the smoothing function includes the basis functions and the coefficients of the basis functions; Concatenating the basis function matrix to determine a design matrix; Taking the historical influencing factor data as a characteristic variable and the historical load data as a target variable, and determining a generalized additive model for load prediction according to the smoothing function.

4. The method according to claim 3, wherein Converting the generalized additive model for load prediction into a state space model to determine an initial load prediction model includes: Determining the state vector and the state transition equation of the state space model according to the coefficients of the basis functions in the generalized additive model for load prediction; Determining the observation equation of the state space model according to the design matrix, the state vector and the state transition equation; Determining an initial load prediction model according to the state vector, the state transition equation and the observation equation.

5. The method according to claim 4, wherein After inputting the influencing factor data into the target load prediction model to predict the load prediction result of the power system, the method further includes: Receiving the real load data of the power system, and updating the target load prediction model according to the real load data, the load prediction result and the influencing factor data.

6. The method according to claim 5, characterized in that Updating the target load prediction model according to the real load data, the load prediction result and the influencing factor data includes: Obtaining a preset recursive model and the historical influencing factor data corresponding to the target load prediction model; Updating the historical influencing factor data according to the influencing factor data; Updating the model parameters of the target load prediction model according to the updated historical influencing factor data, the real load data, the load prediction result and the preset recursive model.

7. The method according to claim 6, wherein Before updating the historical influencing factor data according to the influencing factor data, it further includes: Obtaining the characteristic type of the historical influencing factor data as the historical characteristic type; Determine whether the feature types of the influence factor data all belong to the historical feature types. If there are feature types among the feature types of the influence factor data that do not belong to the historical feature types, determine the newly emerged feature types based on the influence factor data and the historical feature types; Determine the basis function corresponding to the newly emerged feature type as the new basis function, and update the design matrix of the target load prediction model according to the new basis function.

8. The method according to claim 7, characterized in that Updating the historical influence factor data according to the influence factor data includes: If the feature types of the influence factor data all belong to the historical feature types, write the influence factor data into the historical influence factor data to obtain the updated historical influence factor data; If there are feature types among the feature types of the influence factor data that do not belong to the historical feature types, add the newly emerged feature types to the historical influence factor data, and write the influence factor data into the historical influence factor data to obtain the updated historical influence factor data.

9. A self-learning prediction method for power system load, characterized in that, Including: Receive a prediction instruction triggered by the client, where the prediction instruction is used to indicate load prediction for the power system; In the cloud server, in response to the prediction instruction, receive the influence factor data of the power system to be load predicted; obtain a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influence factor data; input the influence factor data into the target load prediction model, and predict the load prediction result of the power system; Return the load prediction result to the client.

10. A self-learning prediction device for a power system load, characterized in that, Including: A data acquisition unit, configured to receive the influence factor data of the power system to be load predicted; A model acquisition unit, configured to obtain a target load prediction model, where the target load prediction model is a model obtained by adjusting the design matrix in the initial load prediction model according to the influence factor data; A load prediction unit, configured to input the influence factor data into the target load prediction model and predict the load prediction result of the power system.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, where when the executable program runs, it controls the device where the computer-readable storage medium is located to execute the self-learning prediction method for the load of the power system according to any one of claims 1 to 8.

12. An electronic device, characterized in that, Including: A memory storing an executable program; A processor for running the program, where when the program runs, it executes the self-learning prediction method for the load of the power system according to any one of claims 1 to 8.

13. A computer program product, comprising computer instructions, characterized in that, When the computer instruction is executed by the processor, it implements the self-learning prediction method for the load of the power system according to any one of claims 1 to 8.

Images