Power load prediction method and device based on quadratic mode decomposition and double-model parallelism, and medium

Abstract

The invention discloses a power load prediction method and device based on quadratic mode decomposition and double-model parallelism, and a medium, and the method comprises the steps: carrying out the fine decomposition of an original load sequence through a quadratic mode decomposition method, carrying out the denoising and reconstruction through combining with a wavelet threshold method, and finally obtaining a series of stable mode components; in a prediction stage, a parallel prediction architecture of the Informer and the BiLSTM is constructed, all modal components are synchronously input, global long-term dependence is captured by using a multi-head probability sparse self-attention mechanism of the Informer, and local short-term dynamic is captured by using the BiLSTM; and carrying out splicing and nonlinear fusion on the heterogeneous features extracted by the two to obtain a final prediction value. Compared with the prior art, cooperative capture and accurate prediction of the multi-scale features of the non-stationary power load sequence are realized through secondary decomposition from coarse to fine and a targeted double-model parallel architecture.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, in particular to a power load prediction method, device and medium based on quadratic modal decomposition and double model parallelism. BACKGROUND

[0002] With the proposal of the "double carbon" target and the promotion of global energy transformation, the power system is undergoing profound changes. The high proportion of new energy access and the diversity of user behavior make the power load present high volatility, strong randomness and nonlinear characteristics. Accurate power load prediction is a key link to ensure the safe and stable operation of the power grid, optimize dispatching decisions and reduce operating costs.

[0003] Existing power load prediction methods are mainly divided into traditional prediction methods, machine learning methods and deep learning methods. Traditional prediction methods, such as regression analysis and time series method, are mostly linear models, which are difficult to capture the complex nonlinear characteristics of modern power system load, resulting in insufficient prediction accuracy; machine learning methods, such as support vector machines and random forests, have improved in handling nonlinear data, but have limited flexibility in model construction and feature extraction capabilities.

[0004] Deep learning methods, especially recurrent neural networks RNN and long short-term memory networks LSTM, are widely used due to their strong time series feature learning capabilities. However, a single deep learning model often struggles to fully capture the complex characteristics of multi-scale and multi-frequency in power load sequences with strong randomness and volatility. For example, long sequence prediction models such as Informer are good at capturing long-term dependencies, but may not be sensitive enough to short-term dramatic fluctuations; while BiLSTM and other models can better capture local time series features, but may face information forgetting and computational efficiency issues when dealing with ultra-long sequences. In addition, the original power load sequence often contains a large amount of high-frequency noise, and directly using the original data for prediction will interfere with the model's feature learning process and affect the prediction accuracy.

[0005] After searching, Chinese patent publication CN117592593A discloses a short-term power load prediction method based on improved quadratic modal decomposition and WOA optimized BiLSTM-attention. This method performs CEEMDAN and VMD quadratic decomposition on the load sequence, and uses the whale optimization algorithm WOA to optimize the hyperparameters of VMD and BiLSTM, and finally predicts each subsequence and superimposes the results to achieve load prediction.

[0006] However, this scheme is after modal decomposition, respectively for each sub-sequence component prediction, and then the prediction results of each component are simply superimposed to obtain the final result, which breaks the inherent internal correlation between different frequency modal components, and uses a single BiLSTM-Attention model to process all components in the final prediction stage, which may cause the model to be difficult to learn and utilize the complex nonlinear interaction between different scale features, thereby limiting the further improvement of prediction accuracy, and may affect the stability of the final result due to independent accumulation of errors.

[0007] Therefore, how to effectively decompose the multi-scale features of the power load, while synergistically utilizing the advantages of different deep learning models, to achieve accurate prediction of non-stationary and noisy power load, is a technical problem to be solved. SUMMARY

[0008] The purpose of the present application is to overcome the defects of the prior art and provide a power load prediction method based on secondary modal decomposition and double model parallelism.

[0009] The purpose of the present application can be achieved by the following technical solutions: According to a first aspect of the present application, a power load prediction method based on secondary modal decomposition and double model parallelism is provided, the method comprising: performing first modal decomposition and denoising on the original power load sequence to obtain a denoised reconstructed signal; inputting the denoised reconstructed signal into a variational modal decomposition algorithm for secondary modal decomposition to obtain a group of modal components with different center frequencies; synchronously and in parallel inputting all modal components obtained by secondary decomposition into a hybrid prediction model composed of an Informer model and a bidirectional long short-term memory network; wherein the Informer model uses a multi-head probabilistic sparse self-attention mechanism to extract global long-term dependency features from all modal components, and the bidirectional long short-term memory network extracts local short-term dynamic features from all modal components; concatenating the extracted global long-term dependency features and local short-term dynamic features in the feature dimension to form a fusion feature vector; mapping the fusion feature vector through a fully connected layer to obtain the final power load prediction value.

[0010] As a preferred technical solution, the first modal decomposition and denoising specifically comprises: using an adaptive noise complete empirical mode decomposition algorithm to decompose the original power load sequence into a group of intrinsic mode function (IMF) components and a residual term; identifying high-frequency IMF components containing noise, and using a wavelet threshold method to denoise the high-frequency IMF components; The de-noised high-frequency IMF component, the untreated low-frequency IMF component and the residual are accumulated to reconstruct a de-noised reconstructed signal.

[0011] As a preferred technical solution, when the wavelet threshold method is used for de-noising processing, the threshold λ used is adaptively determined according to the noise level, and the calculation method is: , Where σ is the noise standard deviation, n is the signal length, and ln is the natural logarithm function.

[0012] As a preferred technical solution, the number K of modal components in the variational modal decomposition algorithm is adaptively determined by the center frequency method, specifically including: through iterative search, taking the separation degree of the center frequency of the modal component as an evaluation index, when the minimum difference between the center frequency of the newly added modal component and the center frequency of the existing modal component is lower than a preset threshold, the current K value is determined as the optimal decomposition number.

[0013] As a preferred technical solution, extracting the global long-term dependency feature based on the Informer model specifically includes: Integrating all modal components into a multi-channel time series tensor to represent the composite fluctuations of power load at different time scales; Using its multi-head probabilistic sparse self-attention mechanism to filter the key time points and modal components that contribute most to long-term load prediction from the multi-channel time series tensor information; Based on the filtered results, output a fixed-dimensional global long-term dependency feature vector for representing long-term trends and periodic patterns in the power load sequence.

[0014] As a preferred technical solution, the bidirectional long short-term memory network is a BiLSTM model, and extracting the local short-term dynamic feature based on the BiLSTM model specifically includes: The total modal components are input into the forward and backward long short-term memory network layers in parallel; For each modal component sequence, through its gating mechanism, the past and future context information is cyclically fused to capture the transient fluctuation pattern of each time point in its local context; The final time step bidirectional hidden states of all modal components are collected to output a fixed-dimensional local short-term dynamic feature vector representing short-term fluctuations and local mutation details in the power load sequence.

[0015] As a preferred technical solution, the method further comprises a multi-feature screening step before the first modal decomposition and denoising of the original power load sequence: the linear correlation strength between each relevant feature and the power load sequence is evaluated by using Pearson correlation analysis, and the features with a linear correlation strength greater than a preset threshold are screened out for subsequent model training.

[0016] As a preferred technical solution, the method further comprises a dynamic iteration process, which adopts a multi-step rolling prediction strategy, comprising: iterative prediction is performed through a sliding time window to generate a future load prediction sequence; In the iterative prediction process, the hybrid prediction model is retrained or parameter fine-tuned using newly acquired actual load data to dynamically correct the prediction error of the model.

[0017] According to a second aspect of the present application, an electronic device is provided, comprising a memory and a processor, the memory having a computer program stored thereon, and the processor implementing the method when executing the program.

[0018] According to a third aspect of the present application, a computer readable storage medium is provided, having a computer program stored thereon, the program being executed by a processor to implement the method.

[0019] Compared with the prior art, the present application has the following advantages: 1. The present application converts a complex load sequence into a set of stationary modal components through secondary modal decomposition, providing high-quality input for subsequent models. Through the parallel design of Informer and BiLSTM, long-term global patterns and short-term local dynamics hidden in the modal components can be simultaneously and fully captured, realizing deep mining and collaborative utilization of multi-scale time series features of power load, and effectively improving the prediction accuracy.

[0020] 2. The secondary modal decomposition strategy proposed in the present application first coarsely decomposes and then finely decomposes, using CEEMDAN combined with wavelet thresholding for initial decomposition and coarse-grained denoising, effectively removing random noise interference and providing a purer input signal for subsequent decomposition, enhancing the stability and reliability of the features.

[0021] 3. The present application synchronously inputs all modal components into the parallel architecture of Informer and BiLSTM, and extracts long-term and short-term features in parallel. These heterogeneous features are then concatenated and nonlinearly fused at the deep network level, enabling the model to autonomously learn the complex interaction between different scale features, avoiding cognitive bias and error accumulation problems of single models, and effectively improving the accuracy and generalization performance of the prediction model. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is a general technical flowchart of the present application; Figure 2 is a structural schematic diagram of the Informer algorithm of the present application; Figure 3 is a structural schematic diagram of the BiLSTM neural network of the present application; Figure 4 is a general structural diagram of the Informer-BiLSTM prediction model based on secondary modal decomposition of the present application; Figure 5 is a raw load data curve diagram of the first group of data sets used in the embodiment of the present application; Figure 6 is a raw load data curve diagram of the second group of data sets used in the embodiment of the present application; Figure 7 is a result diagram after the first group of data sets is decomposed once by CEEMDAN; Figure 8 is a result diagram after the second group of data sets is decomposed once by CEEMDAN; Figure 9 is a result diagram after the first group of data sets is decomposed by secondary modal decomposition; Figure 10 is a result diagram after the second group of data sets is decomposed by secondary modal decomposition; Figure 11 is a comparison diagram of the prediction results of the first group of data sets under different modal decomposition strategies; Figure 12 is a comparison diagram of the prediction results of the second group of data sets under different modal decomposition strategies; Figure 13 is a comparison diagram of the prediction results of the model of the present application and other comparison models under the first group of data sets; Figure 14 is a comparison diagram of the prediction results of the model of the present application and other comparison models under the second group of data sets. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0024] It is apparent that the drawings in the following description merely show some examples or embodiments of the present application, and the present application can be applied to other similar situations without creative labor by those skilled in the art based on these drawings. In addition, it can be understood that, although the efforts made in the development process can be complex and lengthy, some design, manufacture or production changes made on the basis of the technical content disclosed in the present application by those skilled in the art related to the content disclosed in the present application are only routine technical means and should not be understood as insufficient disclosure of the content disclosed in the present application.

[0025] The details of one or more embodiments of the present application are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the present application will be apparent from the description of the embodiments and from the drawings, and various modifications can be made on the basis of the logical equivalents of the features of the application and of the teachings of the application presented in the description without departing from the scope of the application.

[0026] Reference to "an embodiment" or "one embodiment" in this application means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" or "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily all directed to the same embodiment, or to a common or a single embodiment to which all the other embodiments will necessarily be related.

[0027] Unless otherwise defined, technical terms or scientific terms used in the present application shall have the ordinary meanings commonly used by those skilled in the art to which the present application belongs. The terms "one", "a", "an", "the" and similar terms in the present application do not mean quantity limitation, but can mean single or multiple. The terms "include", "contain", "have" and any variations thereof in the present application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device including a series of steps or modules (units) is not limited to the listed steps or units, but can also include steps or units not listed, or can also include other steps or units inherent to the process, method, product or device. The terms "connect", "connect", "couple" and similar terms in the present application are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The term "multiple" in the present application means two or more. The association relationship of the associated objects is described by the term "and / or", which means that there can be three relationships, for example, "A and / or B" can mean that A exists alone, A and B exist together, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects. The terms "first", "second", "third" and the like in the present application are merely to distinguish similar objects, and do not represent a specific order for the objects.

[0028] Embodiment 1: As Figure 1 shown, the application provides a power load prediction method based on secondary modal decomposition and double model parallelism, the overall technical process mainly includes four parts of data processing, secondary modal decomposition, parallel prediction and result fusion, specifically including: Step S1: data acquisition and preprocessing Step S101: collect the historical power load data and related multi-dimensional feature data of the research area, such as maximum temperature, minimum temperature, relative humidity, date characteristics, etc., and preprocess the data, including processing missing values (such as using Lagrange interpolation method) and outliers; Step S102: use Pearson correlation analysis method for feature selection, calculate the linear correlation strength between each feature data and power load, set the threshold value to 0.3, retain the features with correlation greater than the threshold value as the input of the model, and eliminate redundant features, in order to reduce the model calculation complexity and improve the prediction accuracy, for example, temperature features usually have the strongest correlation with load, which should be used as the core prediction feature.

[0029] Step S2: secondary modal decomposition based on CEEMDAN-VMD This step aims to denoise and extract multi-scale features from the original power load sequence; Step S201: first decomposition using adaptive noise complete empirical mode decomposition algorithm CEEMDAN and wavelet threshold denoising; Step S2011: use CEEMDAN to decompose the original power load signal to obtain N intrinsic mode functions IMF and a residual component, high-frequency IMF components such as IMF1-IMF4 mainly contain noise and short-term fluctuations, while low-frequency IMF and residual components contain long-term trends and main signal features; Step S2012: to remove high-frequency noise, the identified high-frequency IMF components are jointly processed using wavelet threshold technology, the wavelet coefficients of each high-frequency IMF are obtained by wavelet transform, the wavelet coefficients are denoised by applying a threshold function, and then the denoised IMF is reconstructed by inverse transform to retain significant signal components: , Where λ is the threshold value, which is adaptively determined according to the noise level, and is expressed as: , Where, is the noise standard deviation, is the signal length; The original form of the low-frequency IMF and the residual term is retained, and the denoised high-frequency components are accumulated with the untreated low-frequency components to obtain the first decomposition signal: , in, This is a decomposition signal. The high-frequency IMF components after noise reduction. These are unprocessed low-frequency IMF components. For the residual term; This operation can remove high-frequency noise while retaining low-frequency effective signals, providing a cleaner input for subsequent VMD decomposition; Step S202: Using the Variational Mode Decomposition (VMD) algorithm, the clean signal reconstructed in S201 is input into the VMD model for secondary decomposition. VMD is a non-recursive signal processing method that can adaptively decompose a signal into multiple stationary components with definite frequency characteristics and bandwidth limitations. The specific steps for secondary feature extraction using VMD are as follows: Step S021: Construct a variational model, and... Decomposed into K modal components Each mode Corresponding to a center frequency And bandwidth, which is determined by the range supported by the frequency domain; Each mode It can be regarded as an amplitude-frequency modulation signal. Its frequency characteristics are analyzed by performing the Hilbert-Huang transform (HHT). To ensure the accuracy of the frequency information, exponential modulation is used to convert it to the vicinity of the fundamental frequency. The mathematical expression for constructing the unconstrained variational problem is shown below. First, the sum of the bandwidths of each mode is minimized to achieve frequency domain localization. At the same time, the signal reconstruction constraint needs to be satisfied to ensure that the sum of the decomposed modal components is equal to the original signal. , Where K is the preset total number of modal decompositions; k is the index of the modal component. ; This is the k-th modal component; Unit impact function; Together they constitute the Hilbert converter; For exponential modulation terms; Step S2022: To transform the constrained optimization problem into an unconstrained optimization problem and obtain the modified augmented Lagrange function, Lagrange multipliers need to be introduced. and secondary penalty factor ; When the signal contains noise or decomposition error, the second-order penalty factor It can balance bandwidth constraints and reconstruction errors, avoiding overfitting or underfitting; Step S2023: Iteratively optimize each modal component using the Alternating Direction Multiplier Method (ADMM). , center frequency and Lagrange multipliers ; Step S2024: After the iteration meets the convergence condition, K modal components are obtained , which have the characteristics of non-overlapping frequency domain components and strong stationarity, and different modes can correspond to different time scale components in the original signal, which is beneficial for subsequent feature analysis; Step S2025: By setting the decomposition layer number K (the optimal value can be determined by the center frequency method), VMD decomposes the signal into K intrinsic mode functions, which correspond to different time scale characteristic components in the load, such as daily cycle, weekly cycle, seasonal trend, and random fluctuations, have good frequency spectrum characteristics and stationarity, and are suitable as input for subsequent deep learning models.

[0030] Step S3: As shown in Figure 4 , an Informer-BiLSTM parallel prediction model is constructed: Step S301: The low-frequency modal component (representing long-term trend and seasonal change) obtained by VMD decomposition in step S202 is input into the Informer model. Informer is an improved model based on the Transformer architecture, as shown in Figure 2 , its core is the multi-head probability sparse self-attention mechanism, which can effectively filter key information and significantly reduce computational complexity, making it particularly good at handling long sequence input and capturing long-term dependencies. Through this branch, the model can accurately grasp the overall trend and periodicity of the power load. The operation logic of the self-attention mechanism is converted into a probability expression, which focuses on key information and efficiently processes it through a probability formula; , where Q*, K, and V are query matrix, key matrix, and value matrix, respectively. Step S302: The high-frequency modal component (representing short-term random fluctuations and local features) obtained by VMD decomposition in step S202 is input into the BiLSTM model, as shown in Figure 3 , the BiLSTM network model is composed of a forward LSTM and a backward LSTM. This bidirectional structure allows it to use past and future context information to analyze data at the current time point, so it performs well in capturing short-term, fast-changing local dependencies in sequences. Through this branch, the model can accurately predict short-term fluctuations in load caused by weather changes, user behavior, etc.

[0031] Step S4: Model training and result fusion The outputs of the two parallel branches are fused by weighting or through a fully connected layer to obtain a final power load prediction result. In the model training stage, an Adam optimizer is adopted, appropriate learning rate, batch size and training rounds are set, a seven-day rolling prediction strategy is adopted, that is, the data of the past seven days are used to predict one or more days in the future, and the prediction results are rolled into the historical window to realize long-term prediction, and the control error is accumulated, and after a certain number of iterations, the model is retrained using the real load data for correction.

[0032] The method of the present application first adopts CEEMDAN combined with wavelet threshold method to perform first decomposition and denoising on the original load sequence, and then uses VMD to perform secondary fine decomposition on the denoised signal to obtain a group of stationary modal components; then, all the components are synchronously input into a parallel hybrid model composed of Informer and BiLSTM to extract global and local features respectively; finally, the prediction result is output through feature splicing and full connection layer fusion, which significantly improves the accuracy and generalization ability of load prediction.

[0033] Example 2: In order to verify the effectiveness of the method of the present application, two groups of public power load data sets with different characteristics are used for experiments.

[0034] Dataset 1: Daily average load data of a certain region from 2012 to 2015, a total of 1107 groups, the original load data curve is as shown in Figure 5 ; Dataset 2: Hourly load data of a certain region from 2016 to 2019, a total of 128544 groups, the original load data curve is as shown in Figure 6 .

[0035] The CEEMDAN-VMD-Informer-BiLSTM model is compared with multiple benchmark models, including single models (BiLSTM, Informer) and different combination models (VMD-Informer, Informer-BiLSTM, etc.), and the mean absolute percentage error (MAPE), root mean square error (RMSE) and determination coefficient (R 2 ) are used as evaluation indexes.

[0036] In order to verify the key role of the secondary modal decomposition CEEMDAN-VMD proposed in the present application in improving the prediction accuracy, an ablation experiment is designed. First, the effectiveness is proved by decomposition result visualization, as shown in Figure 7 and Figure 8 , after the first decomposition of the original sequence by CEEMDAN, the noise is effectively separated into the high-frequency IMF component; further, as shown in Figure 9 and Figure 10As shown, after the VMD secondary decomposition of the denoising signal, the modal component with clear frequency limit and significantly enhanced stationarity is obtained. On this basis, the prediction effect of the complete model CEEMDAN-VMD-Informer-BiLSTM of the present application is compared with the benchmark model Informer-BiLSTM without modal decomposition and the model VMD-Informer-BiLSTM with only one modal decomposition, and the comparison results on the two data sets are shown in Table 1 and Table 2, wherein Table 1 is the precision comparison result of the three modal decomposition prediction models under dataset 1; Table 2 is the precision comparison result of the three modal decomposition prediction models under dataset 2: Table 1 Table 2 The original data is preliminarily decomposed by introducing VMD, and compared with directly using the Informer-BiLSTM model, each error index is obviously decreased, and the R 2 value is also improved. On the basis of VMD, the CEEMDAN is used for pre-decomposition, and the secondary modal decomposition strategy formed brings a performance leap. The present application adopts the secondary decomposition model, on the first group of data sets, the MAPE is reduced by 23.199% and 12.57% respectively, the goodness of fit R 2 is improved by 0.125 and 0.099 respectively; on the second group of data sets, the MAPE is reduced by 26.831% and 17.805% respectively, and the goodness of fit R 2 is improved by 0.163 and 0.134 respectively. As shown in Figure 11 and Figure 12 , the prediction curve of the model of the present application is better fitted with the true value curve. The experiment proves that the secondary modal decomposition, which is a decomposition strategy of first rough and then fine, can focus more on the internal law of each component, so as to realize the great improvement of prediction accuracy and goodness of fit.

[0037] To verify the overall prediction effect and generalization ability of the Informer-BiLSTM multi-feature power load prediction model based on secondary modal decomposition proposed in the application, first, the model is used to predict the load of two data sets, and the prediction effects of single models BiLSTM, Informer, combined models VMD-Informer, VMD-BiLSTM, Informer-BiLSTM, CEEMDAN-VMD-Informer, and CEEMDAN-VMD-BiLSTM are used as control groups, and the prediction results are shown in Tables 3 and 4, wherein Table 3 is the precision comparison result of the CEEMDAN-VMD-Informer-BiLSTM model and multiple benchmark models under dataset 1; Table 4 is the precision comparison result of the CEEMDAN-VMD-Informer-BiLSTM model and multiple benchmark models under dataset 2; and the prediction curve is shown in Figure 13 and Figure 14 The prediction curve of the model of the application is better fitted with the true value curve: on the two data sets, the CEEMDAN-VMD-Informer-BiLSTM model proposed in the application achieves the best effect, compared with the worst single BiLSTM model, on the first data set, the MAPE of the model of the application is reduced by about 92.8%, and the R 2 is increased by about 24.6%, even compared with the suboptimal CEEMDAN-VMD-Informer model, the MAPE of the model of the application is reduced by about 83.6%, and the R 2 is also significantly improved; on the second data set, similar to the results of the first data set, the model of the application has a overwhelming advantage over all control models in the four evaluation indexes, which fully shows that the model proposed in the application has strong prediction performance and good generalization ability, and can adapt to different data characteristics of power load prediction tasks.

[0038] Table 3 Table 4 Example 3 The electronic device of the application includes a central processing unit (CPU) which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or computer program instructions loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM, and RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0039] A number of components in the device are connected to the I / O interface, including: input units, such as a keyboard, a mouse, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as a magnetic disk, an optical disk, etc.; and communication units, such as a network card, a modem, a wireless communication transceiver, etc. The communication units allow the device to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0040] The processing unit performs various methods and processes described above, such as the methods S1-S4. For example, in some embodiments, the methods S1-S4 can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit. In some embodiments, part or all of the computer program can be loaded and / or installed on the device via the ROM and / or the communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of the methods S1-S4 described above can be performed. Alternatively, in other embodiments, the CPU can be configured to perform the methods S1-S4 by any other suitable means, such as by means of firmware.

[0041] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip systems (SOCs), complex programmable logic devices (CPLDs), etc.

[0042] Program code for carrying out methods of the present application can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program code, when executed by the processor or controller, causes the functions / operations specified in the flow charts and / or block diagrams to be implemented. The program code can execute entirely on a machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0043] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0044] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0045] It should be understood that the various forms of flow shown above can be reordered, additional or deleted steps can be used. For example, the steps described in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the technical solutions of the present disclosure can be achieved, which are not limited herein.

[0046] The above description is merely that of a specific implementation of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements shall be encompassed within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be subject to the scope of protection of the claims.

Claims

1. A power load forecasting method based on quadratic modal decomposition and double model parallelism, characterized in that, The method comprises the following steps: Firstly, the original power load sequence is decomposed and denoised by the first modal decomposition to obtain a denoised reconstructed signal; Secondly, the denoised reconstructed signal is input into a variational modal decomposition algorithm for secondary modal decomposition to obtain a group of modal components with different center frequencies; all the modal components obtained by the secondary decomposition are synchronously and parallelly input into a hybrid prediction model composed of an Informer model and a bidirectional long short-term memory network; the Informer model uses a multi-head probabilistic sparse self-attention mechanism to extract global long-term dependency features from all the modal components, and the bidirectional long short-term memory network extracts local short-term dynamic features from all the modal components; the extracted global long-term dependency features and local short-term dynamic features are spliced in the feature dimension to form a fusion feature vector; and the fusion feature vector is mapped to a final power load prediction value through a full connection layer.

2. The power load forecasting method based on quadratic modal decomposition and double model parallelism according to claim 1, characterized in that, The first modal decomposition and denoising specifically comprises the following steps: The original power load sequence is decomposed into a group of intrinsic modal function (IMF) components and a residual term by using an adaptive noise complete empirical modal decomposition algorithm; High-frequency IMF components containing noise are identified, and the wavelet threshold method is used to denoise the high-frequency IMF components; The denoised high-frequency IMF components, the low-frequency IMF components without processing and the residual term are accumulated and reconstructed to obtain the denoised reconstructed signal.

3. The power load forecasting method based on quadratic modal decomposition and double model parallelism according to claim 2, characterized in that, When the wavelet threshold method is used for denoising, the threshold λ is adaptively determined according to the noise level, and the calculation method is as follows: ， Wherein, σ is the standard deviation of noise, n is the length of the signal, and ln is the natural logarithm function.

4. The power load forecasting method based on quadratic modal decomposition and double model parallelism according to claim 1, characterized in that, The number K of modal components in the variational modal decomposition algorithm is adaptively determined by the center frequency method, specifically comprising the following steps: taking the separation degree of the center frequency of the modal component as an evaluation index, when the minimum difference between the center frequency of the newly added modal component and the center frequency of the existing modal component is lower than a preset threshold, the current K value is determined as the optimal decomposition number.

5. The method of claim 1, wherein, The extraction of the global long-term dependency features based on the Informer model specifically comprises the following steps: All the modal components are integrated into a multi-channel time series tensor to represent the composite fluctuations of the power load at different time scales; The multi-head probabilistic sparse self-attention mechanism is used to filter the key time points and modal components that contribute most to the long-term load prediction from the multi-channel time series tensor information; Based on the filtered results, a fixed-dimensional global long-term dependency feature vector is output to represent the long-term trend and periodicity in the power load sequence.

6. The power load forecasting method based on quadratic modal decomposition and double model parallelism according to claim 1, characterized in that, The bidirectional long short-term memory network is a BiLSTM model, and the extraction of the local short-term dynamic features based on the BiLSTM model specifically comprises the following steps: The total modal components are input into the forward and backward long short-term memory network layers in parallel; For each modal component sequence, the past and future context information is cyclically fused through the gating mechanism to capture the transient fluctuation mode of each time point in its local context. The bidirectional hidden states of all modal components at the final time step are aggregated to output a fixed-dimensional local short-term dynamic feature vector representing details of short-term fluctuations and local mutations in the power load sequence.

7. The method of claim 1, wherein the method is characterized by, The method further comprises a multi-feature screening step before the original power load sequence is subjected to the first modal decomposition and denoising: the linear correlation strength between each relevant feature and the power load sequence is evaluated using Pearson correlation analysis, and features with a linear correlation strength greater than a preset threshold are screened out for subsequent model training.

8. The power load forecasting method based on quadratic modal decomposition and double model parallelism according to claim 1, characterized in that, The method further comprises a dynamic iteration process, which adopts a multi-step rolling prediction strategy, comprising: iterative prediction through a sliding time window to generate a future load prediction sequence; During the iterative prediction process, the hybrid prediction model is retrained or fine-tuned using newly acquired actual load data to dynamically correct the prediction error of the model.

9. An electronic device comprising a memory and a processor, said memory having stored thereon a computer program, characterized in that, The processor executes the program to implement the method of any one of claims 1-8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the method of any one of claims 1-8. The processor executes the program to implement the method of any one of claims 1-8.

Images