A method and system for power load forecasting based on kinetic driving

Abstract

The present application relates to the field of power load fluctuation prediction, and provides a power load prediction method and system based on dynamics driving. The power load prediction method based on dynamics driving comprises the following steps: adopting a deterministic learning algorithm to perform dynamics modeling on a historical power load time training sequence obtained, and extracting inherent dynamics information of data; based on the dynamics information, adopting a multi-step forward iteration predictor based on a fourth-order Runge-Kutta model to obtain a power load prediction sequence composed of prediction values from the first step to the nth step; adopting a corrector to perform periodic correction on the power load prediction sequence composed of the prediction values from the first step to the nth step, and obtaining a corrected value of the power load prediction sequence of the first step; and based on the prediction value and the corrected value, obtaining a power load prediction result. The power load prediction accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of power load fluctuation prediction technology, and in particular to a power load prediction method and system based on dynamics. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Electricity load sequences, as a typical type of chaotic time series, are special time-series data generated by the complex nonlinear dynamic system of the power system. Their dynamic behavior exhibits complex, irregular, and seemingly random characteristics. Due to the extreme sensitivity of the power system to initial conditions—even small differences in the initial load state can amplify exponentially over time—the long-term evolution trajectory of electricity load is difficult to predict accurately. Therefore, accurate prediction of electricity load has always been a significant challenge in the field of power system operation and dispatch.

[0004] However, power load forecasting holds an irreplaceable and crucial position in the power industry. Its forecasting results directly impact the safe and stable operation of the power system, the rational formulation of power generation plans, the orderly trading of electricity market resources, and the efficient utilization of energy resources. For example, accurate short-term load forecasting can provide a basis for intraday unit combination optimization, while medium- and long-term load forecasting is a key reference for grid planning and power source construction. Effective power load forecasting can provide crucial decision support for practical scenarios such as economic dispatching of the power system, fault early warning, and renewable energy consumption, possessing significant theoretical research value and substantial practical application benefits.

[0005] Currently, power load forecasting methods are mainly divided into two categories: traditional statistical model methods and intelligent model methods. Traditional statistical model methods, represented by Autoregressive Moving Average (ARIMA) and Autoregressive Conditional Heteroskedasticity (ARCH), make predictions by establishing mathematical models under the assumptions of linearity or stationarity. However, these methods are difficult to adapt to the strong nonlinear and non-stationary characteristics of power load sequences due to the influence of user behavior, meteorological factors, social events, etc., and their prediction accuracy is limited, making it difficult to meet the refined dispatching needs of modern power systems.

[0006] With the development of machine learning technology, intelligent model methods have gradually become the mainstream approach for power load forecasting. Among these, artificial neural networks, with their powerful nonlinear approximation capabilities, are widely used, including feedforward neural networks such as Multilayer Perceptrons (MLP) and Radial Basis Functions (RBF), as well as recurrent neural networks such as Long Short-Term Memory (LSTM) and Echo State Networks (ESN). Furthermore, Transformer models based on attention mechanisms have also been introduced into the field of power load forecasting and have demonstrated excellent performance in some scenarios. Although intelligent model methods have improved the accuracy of power load forecasting to some extent, the following key issues still exist in practical applications:

[0007] 1. Existing intelligent models mostly adopt data-driven static mapping methods, which only focus on the superficial relationship between historical load data and future load data, but ignore the inherent dynamic evolution law contained in the power load sequence as a chaotic system. As a result, the model cannot fully explore the deep driving mechanism of load changes and has insufficient overall prediction potential.

[0008] 2. In multi-step iterative prediction scenarios, existing models generally lack targeted error correction mechanisms and cannot effectively suppress the error diffusion effect unique to chaotic systems, resulting in a sharp decline in the accuracy of medium- and long-term power load prediction, making it difficult to support medium- and long-term planning decisions for the power grid.

[0009] 3. Most mainstream neural network models are "black box" structures, and their prediction process relies on implicit fitting of a large number of parameters, which cannot clearly reveal the dynamic mechanism of power load sequence evolution, resulting in a lack of interpretability in the prediction results. This deficiency is particularly prominent in power systems with extremely high requirements for safety and reliability, severely restricting the application of models in critical power dispatching and control scenarios. Summary of the Invention

[0010] To address the technical problems existing in the background art, the present invention provides a power load forecasting method and system based on dynamics. The present invention can integrate chaotic dynamics characteristics, has error suppression capabilities and strong interpretability, thereby improving the accuracy and practicality of power load forecasting and providing stronger technical support for the safe, economical and efficient operation of power systems.

[0011] To achieve the above objectives, the present invention adopts the following technical solution:

[0012] The first aspect of the present invention provides a dynamically driven method for predicting power load.

[0013] A dynamically driven power load forecasting method includes:

[0014] A deterministic learning algorithm is used to perform dynamic modeling on the acquired historical power load time training sequence to extract the inherent dynamic information of the data;

[0015] Based on dynamic information, a multi-step forward iterative predictor based on a fourth-order Runge-Kutta model is used to obtain the prediction from the th... Step to the first A power load forecast sequence composed of step-by-step forecast values;

[0016] Using an error corrector to correct the error from the first Step to the first The power load forecast sequence composed of the predicted values ​​from step 1 is periodically corrected to obtain the 1st step. Corrected values ​​for the power load forecast sequence at step 1; based on the first step Step-by-step predicted value and the first The power load forecast sequence correction value is used to obtain the power load forecast result.

[0017] Furthermore, after obtaining the historical power load time training sequence, the process includes data cleaning of the historical power load time training sequence, and standardization or normalization of the historical power load time training sequence after cleaning.

[0018] Furthermore, the method for dynamically modeling the acquired historical power load time training sequence using a deterministic learning algorithm includes:

[0019] A dynamic identifier is constructed using an RBF neural network, and a weight update law based on Lyapunov stability is employed to accurately model the dynamics of the unknown system along the system's sampling trajectory.

[0020]

[0021] in, Representing the dynamics of an unknown system, To train the constant weights of the converged RBF neural network, For the regression vector of the RBF neural network, This represents the modeling error.

[0022] Furthermore, the multi-step forward iterative predictor based on the fourth-order Runge-Kutta model is expressed by the following formula:

[0023]

[0024] in, For the first The predicted value of the step, Sampling time, These are the constant weights embedded in the neural network, representing the learned dynamics. Indicates the first Instantaneous dynamics of a step Indicates based on The ( Predicting dynamics step by step, Indicates based on The ( Step-by-step improvement of dynamics, Indicates based on The ( Predicting dynamics step by step, to The calculations need to be performed sequentially to obtain the final result. Predicted value of step ;

[0025] when At time 1, the multi-step forward iterative predictor predicts the value at the first future time point based on the initial input; when At this time, the multi-step forward iterative predictor uses the predicted value of the previous step as the new input and repeats the iterative process to generate a power load prediction sequence containing multiple future time points.

[0026] Furthermore, the process of training the error corrector includes: slicing the historical power load training time series using a sliding window to generate a length of... Several subsequences Using the starting point of the subsequence as the initial value, a multi-step forward iterative predictor based on the fourth-order Runge-Kutta model is used to generate corresponding predicted subsequences. Construct data pairs of predicted and actual values. Training error corrector.

[0027] Furthermore, during the training of the error corrector, a confidence factor is added to the original loss function. The regularization term optimizes the error corrector; the improved loss function is:

[0028]

[0029]

[0030] in, For MLP network weights, for Regularization coefficient, As the confidence factor, This indicates normalization of the mean probability. This represents the original loss function.

[0031] Furthermore, after the error corrector is trained, the error corrector is used to obtain the first... Correction values ​​for step power load forecasting sequence Then based on the first Step-by-step predicted value and the first The power load forecast result is obtained by adjusting the step value; it is expressed by the following formula:

[0032]

[0033] in, Indicates the first Step-by-step power load forecast results These are weighting coefficients. Indicates by The power load forecast sequence is composed of step-by-step forecast values. For the first Predicted value of step This indicates an error corrector.

[0034] Furthermore, the error corrector employs an MLP neural network.

[0035] A second aspect of the present invention provides a dynamically driven power load prediction system.

[0036] A dynamically driven power load forecasting system includes:

[0037] The dynamic modeling module is configured to: use a deterministic learning algorithm to perform dynamic modeling on the acquired historical power load time training sequence and extract the inherent dynamic information of the data;

[0038] The multi-step prediction module is configured to: based on dynamic information, employ a multi-step forward iterative predictor based on a fourth-order Runge-Kutta model to obtain predictions from the first step... Step to the first A power load forecast sequence composed of step-by-step forecast values;

[0039] The error correction module is configured to use an error corrector to correct the error from the first... Step to the first The power load forecast sequence composed of the predicted values ​​from step 1 is periodically corrected to obtain the 1st step. Corrected values ​​for the power load forecast sequence at step 1; based on the first step Step-by-step predicted value and the first The power load forecast sequence correction value is used to obtain the power load forecast result.

[0040] A third aspect of the present invention provides a computer device comprising:

[0041] A processor, adapted to execute computer programs;

[0042] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the dynamically driven power load forecasting method described in the first aspect above.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] This invention uses a deterministic learning algorithm to accurately model the inherent nonlinear dynamics of the chaotic time series of power load. Unlike traditional machine learning methods that only perform a superficial static mapping of "historical load data - future load data", this invention can deeply capture the essential evolutionary characteristics of power load series under the coupling influence of multiple factors such as user behavior, meteorological conditions, and social activities. This significantly improves the model's prediction accuracy for short-term, medium-term, and long-term power load, providing a more reliable decision-making basis for the refined scheduling and economic operation of the power system.

[0045] To address the issue of error propagation in multi-step iterative power load forecasting, this invention designs a forecast-correction mechanism that considers the changing patterns of forecast errors. By introducing a confidence factor based on the trend of forecast error changes into the corrector network, the corrector neural network can more accurately identify and focus on reliable information in the input sequence, effectively reducing the accumulated error in the iterative forecasting process and significantly improving the performance of medium- and long-term power load forecasting. Its forecast results can better support medium- and long-term decision-making scenarios such as power grid planning and power source construction.

[0046] The prediction mechanism of this invention relies heavily on the inherent physical information contained in the time-series data of power load (such as the dynamic correlation between load and electricity consumption periods, seasonal cycles, and temperature changes), which can reveal the dynamic mechanism of power load sequence evolution to a certain extent and has good physical interpretability. This feature effectively solves the problem of limited application of traditional "black box" models in critical scenarios of power systems, and greatly enhances the application potential of the model in fields with extremely high safety requirements such as power system safety and stability control and power supply guarantee for important users, providing more solid technical support for the safe operation of power systems.

[0047] In summary, this invention effectively solves the problems of insufficient utilization of the inherent dynamic characteristics of data, large cumulative error in multi-step prediction, poor medium- and long-term prediction performance, and lack of physical interpretability in existing power load prediction methods by integrating the chaotic dynamic characteristics of power load sequences, optimizing error correction mechanisms, and enhancing the physical interpretability of prediction results. It significantly improves the accuracy, reliability, and practicality of power load prediction. Attached Figure Description

[0048] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0049] Figure 1 This is a flowchart illustrating a dynamically driven power load forecasting method according to an embodiment of the present invention;

[0050] Figure 2 This is a flowchart illustrating another embodiment of the dynamic-driven power load forecasting method of the present invention;

[0051] Figure 3 This is a structural diagram of a dynamically driven power load forecasting system shown in an embodiment of the present invention;

[0052] Figure 4 This is a structural diagram of a computer device shown in an embodiment of the present invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0054] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0055] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0056] To facilitate understanding of the technical solutions of this invention, some technical terms involved in this invention will be introduced below.

[0057] Deterministic learning algorithms are existing machine learning methods. Against the backdrop of continuous evolution in machine learning technology, deterministic learning algorithms, as an effective learning tool for nonlinear systems in dynamic environments, have emerged in numerous research and applications. This method includes the following core elements:

[0058] (1) The use of radial basis function (RBF) neural networks provides them with powerful function approximation capabilities. With its unique structure, the RBF neural network can efficiently fit complex nonlinear functions, which is particularly advantageous when dealing with data such as power loads that have complex nonlinear characteristics. In previous related machine learning application research, such as in the modeling and control scenarios of some complex industrial processes, the RBF neural network has been successfully applied to describe the dynamic characteristics of the system, demonstrating good adaptability and accuracy.

[0059] (2) The weight adjustment law based on Lyapunov stability theory ensures the stability of the algorithm during the learning process. Lyapunov stability theory provides a solid mathematical foundation for the stability analysis of dynamic systems. By introducing it into the weight adjustment process, it ensures that deterministic learning maintains overall stability while continuously approximating the real dynamic characteristics.

[0060] (3) Accurate modeling of the intrinsic dynamics of time-series data is achieved along the system trajectory, and this characteristic is stored in the form of constant weights in a neural network, enabling the deterministic learning algorithm to effectively capture the essential dynamic laws of the system. In the scenario of power load forecasting, traditional intelligent models often only focus on the superficial correlation between historical and future data, while the deterministic learning algorithm mines the intrinsic dynamic evolution laws contained in the power load sequence as a chaotic system, fixes these laws in the form of constant weights in a neural network, and thus achieves a deeper understanding and prediction of power load changes.

[0061] As described in the background section, existing power load forecasting suffers from technical problems such as insufficient utilization of the inherent dynamic characteristics of data, large cumulative errors in multi-step forecasting, poor medium- and long-term forecasting performance, and a lack of physical interpretability in the forecast results. This invention provides a dynamic-driven power load forecasting method and system. Several embodiments are described in detail below.

[0062] Figure 1 This is a flowchart illustrating a dynamically driven power load forecasting method according to an embodiment of the present invention; see also Figure 1 The method includes:

[0063] A deterministic learning algorithm is used to perform dynamic modeling on the acquired historical power load time training sequence to extract the inherent dynamic information of the data;

[0064] Based on dynamic information, a multi-step forward iterative predictor based on a fourth-order Runge-Kutta model is used to obtain the prediction from the th... Step to the first A power load forecast sequence composed of step-by-step forecast values;

[0065] Using an error corrector to correct the error from the first Step to the first The power load forecast sequence composed of the predicted values ​​from step 1 is periodically corrected to obtain the 1st step. Corrected values ​​for the power load forecast sequence at step 1; based on the first step Step-by-step predicted value and the first The power load forecast sequence correction value is used to obtain the power load forecast result.

[0066] This invention significantly improves the short-term forecast accuracy of power load time series and enhances the stability of medium- and long-term forecasts by exploring the inherent dynamic characteristics of power load time series and constructing an effective prediction-correction mechanism.

[0067] Figure 2 This is a flowchart of another embodiment of the dynamic-driven power load forecasting method shown in this invention; see also Figure 2 The dynamic-driven power load forecasting method includes the following steps:

[0068] Step (1): Obtain historical power load time series data and preprocess the data.

[0069] In this embodiment, the data preprocessing operation includes: first, cleaning the historical power load time series data; then, standardizing or normalizing the historical power load time series data; and finally, dividing the power load time training sequence and the power load time test sequence according to a 7:3 ratio.

[0070] Data cleaning includes removing outliers and filling in missing values.

[0071] Step (2): Use a deterministic learning algorithm to perform dynamic modeling on the preprocessed power load time training sequence, extract the inherent dynamic information of the data, and store it in the form of constant weights of the neural network.

[0072] Among them, the power load time training sequence can be regarded as a type of multivariate time series data generated by a nonlinear dynamic system, which can be represented by the following differential equation:

[0073] (1)

[0074] in, It is a system state variable. Represents the rate of change of the system state. It is a system dimension. It is the system's parameter vector. It is the unknown dynamics of the system. Electricity load time series data is analyzed through... dimensional state variables Obtained by sampling.

[0075] In some embodiments, a deterministic learning method is used to perform dynamic modeling of power load time series data. Specifically, this involves constructing a dynamic identifier using an RBF neural network and employing a weight update law based on Lyapunov stability design to model the unknown system dynamics along the system sampling trajectory. Accurate modeling:

[0076] (2)

[0077] in, To train the converged neural network with constant weights, i.e., to store the dynamic knowledge, For the regression vector of the RBF neural network, This represents the modeling error.

[0078] This invention uses a deterministic learning algorithm to accurately model the inherent nonlinear dynamics of the chaotic time series of power load. Unlike traditional machine learning methods that only perform a superficial static mapping of "historical load data - future load data", this invention can deeply capture the essential evolutionary characteristics of power load series under the coupling influence of multiple factors such as user behavior, meteorological conditions, and social activities. This significantly improves the model's prediction accuracy for short-term, medium-term, and long-term power load, providing a more reliable decision-making basis for the refined scheduling and economic operation of the power system.

[0079] Step (3): Using the obtained dynamic information, construct a multi-step forward iterative predictor based on the fourth-order Runge-Kutta model and realize multi-step iterative prediction.

[0080] The fourth-order Runge-Kutta model iterative predictor, which incorporates dynamical knowledge, is represented by the following formula:

[0081] (3)

[0082] in, For the first The predicted value of the step, Sampling time, These are the constant weights of the embedded neural network, representing the learned dynamics. Indicates the first Instantaneous dynamics of a step Indicates based on The ( Predicting dynamics step by step, Indicates based on The ( Step-by-step improvement of dynamics, Indicates based on The ( Predicting dynamics step by step, to The calculations need to be performed sequentially to obtain the final result. Predicted value of step .

[0083] In this embodiment, when When the predictor is based on the initial input, it predicts the value for the first future time point; when In this process, the predictor uses the previous prediction as new input and repeats the iterative process to generate a prediction sequence containing multiple future time points. In this architecture, the model does not predict all future steps at once, but rather makes predictions step by step.

[0084] Step (4): Construct a neural network-based error corrector to periodically correct the predictor output in real time.

[0085] The goal of the error corrector is to correct the erroneous predicted values ​​to the true values ​​as closely as possible. This corrector is built based on an MLP neural network, and the network's input is from the first... Step to the first A sequence of predicted values The network output is the first step correction value .

[0086] In this embodiment, the training strategy of the error corrector is as follows: first, a sliding window is used to slice the power load time training sequence data, generating a length of... Several subsequences Then, using the starting point of the subsequence as the initial value, the predictor generates the corresponding predicted subsequence for the subsequence. Finally, construct a "predicted value - actual value" data pair. It is used to train MLP neural networks.

[0087] It is important to note that, considering the evolutionary trend of iterative models where prediction errors are often small initially but gradually increase with each iteration, inference neural networks should focus more on the initial part of the input sequence, as the information in this part is relatively more reliable. Therefore, a confidence factor should be introduced into the loss function. The regularization term makes the neural network pay more attention to the prediction information with higher confidence in the input sequence. The modified loss function is as follows:

[0088] (4)

[0089] in, This represents the original loss function, which can be either the cross-entropy loss function or the mean squared error loss function. For MLP network weights, for Regularization coefficient, The confidence factor is obtained by normalizing the average value of the prediction error sequence to reflect the trend of error variation.

[0090] (5)

[0091] After the corrector training is complete, the corrected values ​​are output using the trained corrector and then weighted and fused with the predicted values ​​output by the predictor to obtain the final corrected prediction result.

[0092] (6)

[0093] in, These are weighting coefficients. Indicates by The sequence of predicted values. This indicates a trained corrector. This indicates the output value of the corrector.

[0094] In this embodiment, periodic correction, i.e., every time steps ( ) Correct once, of which The step size is set based on the degree of chaos in the time series data: for time series data with a lower degree of chaos, a larger correction step size can be used. This reduces the calibration frequency and avoids unnecessary fluctuations caused by excessively frequent calibrations.

[0095] To address the issue of error propagation in multi-step iterative power load forecasting, this invention designs a forecast-correction mechanism that considers the changing patterns of forecast errors. By introducing a confidence factor based on the trend of forecast error changes into the corrector network, the corrector neural network can more accurately identify and focus on reliable information in the input sequence, effectively reducing the accumulated error in the iterative forecasting process and significantly improving the performance of medium- and long-term power load forecasting. Its forecast results can better support medium- and long-term decision-making scenarios such as power grid planning and power source construction.

[0096] Step (5): Determine whether the iteration has ended. After the iteration ends, output the power load prediction result.

[0097] This invention effectively solves the problems of insufficient utilization of the inherent dynamic characteristics of data, large cumulative error in multi-step prediction, poor medium- and long-term prediction performance, and lack of physical interpretability in existing power load prediction methods by integrating the chaotic dynamic characteristics of power load sequences, optimizing error correction mechanisms, and enhancing the physical interpretability of prediction results. It significantly improves the accuracy, reliability, and practicality of power load prediction.

[0098] The above combination Figure 1 The dynamic-driven power load forecasting method provided in the embodiments of the present invention has been described in detail. Next, the dynamic-driven power load forecasting system provided in the embodiments of the present invention will be described in conjunction with the accompanying drawings.

[0099] Figure 3 This is a schematic diagram of the structure of a dynamically driven power load forecasting system according to an embodiment of the present invention, with reference to... Figure 3 The system described in this invention includes:

[0100] The data acquisition module is configured to collect power load time series data. It performs preprocessing on the collected power load time series data, including but not limited to data cleaning (removing outliers and filling missing values), data standardization or normalization, and data partitioning (dividing the processed data into power load time training sequences and power load time test sequences, with the power load time training sequences used for dynamic modeling and error corrector training). After preprocessing, the prepared power load time training sequences are transferred to the dynamic modeling module.

[0101] The dynamics modeling module is configured to receive the preprocessed power load time training sequence output from the data acquisition module, and model the intrinsic dynamics of the power load time training sequence based on a deterministic learning algorithm. Specifically, it constructs a radial basis function (RBF) neural network dynamic identifier, employs a weight update law based on Lyapunov stability design to approximate the nonlinear dynamic function inherent in the time series, and stores constant weights. The stored constant weights are then transmitted to the multi-step prediction module, providing dynamic knowledge for predictor construction.

[0102] The multi-step prediction module is configured to receive dynamic information in constant-weight form output from the dynamic modeling module, and construct a multi-step forward iterative predictor based on a 4th-order Runge-Kutta model. This predictor uses the first state of the power load time-test sequence as initial input, and iteratively predicts based on the 4th-order Runge-Kutta model embedded with dynamic knowledge, generating predicted values ​​for multiple future time steps. The generated prediction sequence is transmitted in real-time to the error correction module for correction processing.

[0103] The error correction module is configured to include an error corrector based on a neural network (such as a multilayer perceptron, MLP). This corrector needs to be pre-trained using historical power load training sequences. It extracts sub-sequences from the historical power load training sequences using a sliding window, generates predicted values ​​for the corresponding sub-sequences using the predictor of the multi-step prediction module, constructs "predicted value-actual value" data pairs, trains the corrector to learn the changing patterns of prediction errors, and introduces a confidence factor based on the prediction error trend into the training loss function, making the corrector focus more on reliable prediction information. During the prediction phase, it receives the prediction sequence output from the multi-step prediction module and corrects the predicted values ​​in real time according to a preset period. Based on the prediction sequence output from the multi-step prediction module and the corrected values ​​of the power load prediction sequence, the power load prediction result is obtained.

[0104] According to embodiments of the present invention, the dynamic-driven power load forecasting system can correspond to the execution of the methods described in the embodiments of the present invention, and the above and other operations and / or functions of each module of the dynamic-driven power load forecasting system are respectively for implementing Figure 1 For the sake of brevity, the corresponding processes of each method in the code will not be elaborated here.

[0105] See Figure 4 The diagram shows the structure of a computer device, which includes a processor, a communication interface, and a computer-readable storage medium. The processor, communication interface, and computer-readable storage medium are connected via a bus or other means. The communication interface is used to receive and send data. The computer-readable storage medium can be stored in the computer device's memory. The computer-readable storage medium stores computer programs, including program instructions, and the processor executes the program instructions stored in the computer-readable storage medium. The processor (or CPU, Central Processing Unit) is the computing and control core of the computer device, adapted to implement one or more instructions, specifically adapted to load and execute one or more instructions to implement the corresponding steps in the embodiment of the dynamic-driven power load forecasting method.

[0106] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0107] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0111] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamically driven power load forecasting method, characterized in that, include: A deterministic learning algorithm is used to perform dynamic modeling on the acquired historical power load time training sequence to extract the inherent dynamic information of the data; Based on dynamic information, a multi-step forward iterative predictor based on a fourth-order Runge-Kutta model is used to obtain the prediction from the th... Step to the first A power load forecast sequence composed of step-by-step forecast values; Using an error corrector to correct the error from the first Step to the first The power load forecast sequence composed of the predicted values ​​from step 1 is periodically corrected to obtain the 1st step. Corrected values ​​for the power load forecast sequence at step 1; based on the first step Step-by-step predicted value and the first The corrected values ​​of the power load forecast sequence are used to obtain the power load forecast results; The method for performing dynamic modeling on the acquired historical power load time training sequence using a deterministic learning algorithm includes: A dynamic identifier is constructed using an RBF neural network, and a weight update law based on Lyapunov stability is employed to accurately model the dynamics of the unknown system along the system's sampling trajectory. in, Representing the dynamics of an unknown system, To train the constant weights of the converged RBF neural network, For the regression vector of the RBF neural network, This represents the modeling error.

2. The power load forecasting method based on dynamics as described in claim 1, characterized in that, After obtaining the historical power load time training sequence, the process includes data cleaning of the historical power load time training sequence, and then standardization or normalization of the historical power load time training sequence after cleaning.

3. The power load forecasting method based on dynamics as described in claim 1, characterized in that, The multi-step forward iterative predictor based on the fourth-order Runge-Kutta model is expressed by the following formula: in, For the first The predicted value of the step, Sampling time, These are the constant weights embedded in the neural network, representing the learned dynamics. Indicates the first Instantaneous dynamics of a step Indicates based on The ( Predicting dynamics step by step, Indicates based on The ( Step-by-step improvement of dynamics, Indicates based on The ( Predicting dynamics step by step, to The calculations need to be performed sequentially to obtain the final result. Predicted value of step ; when At time 1, the multi-step forward iterative predictor predicts the value at the first future time point based on the initial input; when At this time, the multi-step forward iterative predictor uses the predicted value of the previous step as the new input and repeats the iterative process to generate a power load prediction sequence containing multiple future time points.

4. The power load forecasting method based on dynamics as described in claim 1, characterized in that, The process of training the error corrector includes: slicing the historical power load time series using a sliding window to generate a length of... Several subsequences Using the starting point of the subsequence as the initial value, a multi-step forward iterative predictor based on the fourth-order Runge-Kutta model is used to generate corresponding predicted subsequences. Construct data pairs of predicted and actual values. Training error corrector.

5. The power load forecasting method based on dynamics as described in claim 4, characterized in that, During the training of the error corrector, a confidence factor is added to the original loss function. The regularization term optimizes the error corrector; the improved loss function is: in, For MLP network weights, for Regularization coefficient, As the confidence factor, This indicates normalization of the mean probability. This represents the original loss function.

6. The power load forecasting method based on dynamics-driven methods according to claim 1, characterized in that, After the error corrector is trained, the error corrector is used to obtain the first... Correction values ​​for step power load forecasting sequence Then based on the first Step-by-step predicted value and the first Step correction values ​​are used to obtain the power load forecast results; The following formula is used to express this: in, Indicates the first Step-by-step power load forecast results These are weighting coefficients. Indicates by Electricity load forecasting sequence composed of step-by-step forecast values For the first The predicted value of the step, This indicates an error corrector.

7. The power load forecasting method based on dynamics as described in claim 1, characterized in that, The error corrector uses an MLP neural network.

8. A dynamically driven power load forecasting system, characterized in that, include: The dynamic modeling module is configured to: use a deterministic learning algorithm to perform dynamic modeling on the acquired historical power load time training sequence and extract the inherent dynamic information of the data; The multi-step prediction module is configured to: based on dynamic information, employ a multi-step forward iterative predictor based on a fourth-order Runge-Kutta model to obtain predictions from the first step... Step to the first A power load forecast sequence composed of step-by-step forecast values; The error correction module is configured to use an error corrector to correct the error from the first... Step to the first The power load forecast sequence composed of the predicted values ​​from step 1 is periodically corrected to obtain the 1st step. Corrected values ​​for the power load forecast sequence; Based on the Step-by-step predicted value and the first The corrected values ​​of the power load forecast sequence are used to obtain the power load forecast results; The method for performing dynamic modeling on the acquired historical power load time training sequence using a deterministic learning algorithm includes: A dynamic identifier is constructed using an RBF neural network, and a weight update law based on Lyapunov stability is employed to accurately model the dynamics of the unknown system along the system's sampling trajectory. in, Representing the dynamics of an unknown system, To train the constant weights of the converged RBF neural network, For the regression vector of the RBF neural network, This represents the modeling error.

9. A computer device, characterized in that, A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps of the dynamically driven power load forecasting method as described in any one of claims 1-7.

Images