Wind speed prediction method and system

Abstract

The invention discloses a wind speed prediction method and system, and the method comprises the steps: obtaining first historical wind speed data in a target scene, and carrying out the first preprocessing of the first historical wind speed data; presetting a first decomposition strategy, and decomposing the first historical wind speed data after the first preprocessing to obtain a plurality of groups of second data; and taking the plurality of groups of second data as the input of the pre-trained first network architecture, and performing wind speed prediction according to the output of the first network architecture. The accuracy of wind speed prediction can be effectively improved, so that a more reliable reference is provided for wind power generation planning and operation of a power grid. By adopting advanced data processing and network architecture, the complexity and uncertainty of wind speed data can be processed, and technical support is provided for optimization management of wind power generation.

Description

Technical Field

[0001] The present invention relates to the technical field of wind speed prediction, and in particular, to a wind speed prediction method and system. Background Art

[0002] With the increasing prominence of global energy security issues and the exacerbation of environmental problems such as climate change, wind power generation technology has gradually attracted the attention of countries around the world. Natural wind power fluctuates irregularly with weather changes, resulting in unstable power generation.

[0003] At the same time, the proportion of wind power in the national power grid is also non-stationary, and these unstable factors pose challenges to the safe and reliable operation of the power grid.

[0004] Therefore, the ability to accurately predict wind speed is of great significance for the planning and safe operation of power grid wind power generation. Summary of the Invention

[0005] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Simplifications or omissions may be made in this part, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this part, the abstract, and the title, and such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a wind speed prediction method and system, which can solve the problems mentioned in the background art.

[0008] To solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, the present invention provides a wind speed prediction method, including:

[0010] Obtaining first historical wind speed data in a target scenario and performing first preprocessing on the first historical wind speed data;

[0011] Presetting a first decomposition strategy, and decomposing the first preprocessed first historical wind speed data according to the first decomposition strategy to obtain several groups of second data;

[0012] Using the several groups of second data as the input for pre-training a first network architecture, and performing wind speed prediction according to the output of the first network architecture.

[0013] As a preferred solution of the wind speed prediction method according to the present invention, wherein: performing wind speed prediction according to the output of the first network architecture includes:

[0014] The second data of different groups all correspond to the output of a first network architecture;

[0015] Perform a first merge on the outputs of all the first network architectures;

[0016] Perform wind speed prediction according to the first merge result.

[0017] As a preferred solution of the wind speed prediction method of the present invention, wherein: the first network architecture includes: the first network architecture is any network architecture with several groups of second data as input and the wind speed value or relevant parameters that can directly or indirectly obtain the wind speed value as output.

[0018] As a preferred solution of the wind speed prediction method of the present invention, wherein: the first network architecture further includes: several first network architectures all have the same architecture;

[0019] The inputs of several first network architectures are different groups of second data randomly assigned;

[0020] After obtaining the outputs of all the first network architectures, perform a first merge.

[0021] As a preferred solution of the wind speed prediction method of the present invention, wherein: the first preprocessing at least includes determining the time step of the first historical wind speed data.

[0022] As a preferred solution of the wind speed prediction method of the present invention, wherein: the first decomposition strategy includes:

[0023] Perform data parsing on the first historical wind speed data after the first preprocessing to obtain a parsed signal;

[0024] Preset an objective function, which is used to minimize the sum of the bandwidths of the center frequencies of each modal component while ensuring that the sum of all modal components is equal to the original signal;

[0025] By optimizing the objective function and using singular value decomposition, obtain a set of modal functions of variational mode decomposition;

[0026] Use a set of modulation signals with increasing frequencies to demodulate the modal functions to obtain multiple modal components.

[0027] As a preferred solution of the wind speed prediction method of the present invention, wherein: the first merge includes: performing weighted summation on the outputs of all the first network architectures according to a preset weight.

[0028] In a second aspect, the present invention provides a wind speed prediction system, which includes:

[0029] A data acquisition and processing module, configured to acquire first historical wind speed data in a target scenario and perform first preprocessing on the first historical wind speed data;

[0030] A data decomposition module, configured to preset a first decomposition strategy, and decompose the first preprocessed first historical wind speed data according to the first decomposition strategy to obtain several groups of second data;

[0031] A prediction module, configured to use the several groups of second data as inputs for pre-training a first network architecture, and perform wind speed prediction according to the output of the first network architecture.

[0032] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the method described above are implemented.

[0033] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described above are implemented.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a wind speed prediction method and system, which acquire first historical wind speed data in a target scenario and perform first preprocessing on the first historical wind speed data; preset a first decomposition strategy, decompose the first preprocessed first historical wind speed data to obtain several groups of second data; use the several groups of second data as inputs for pre-training a first network architecture, and perform wind speed prediction according to the output of the first network architecture. It can effectively improve the accuracy of wind speed prediction, thereby providing a more reliable reference for the wind power generation planning and operation of the power grid. By adopting advanced data processing and network architectures, this application can handle the complexity and uncertainty of wind speed data, providing technical support for the optimized management of wind power generation. In addition, the prediction system of this application has good scalability and adaptability, and can adapt to the wind speed prediction requirements in different scenarios, further improving the efficiency and stability of wind power generation. Additionally, by combining quantum computing and deep learning technologies, the wind speed prediction model can more accurately capture and predict wind speed changes, providing more reliable decision-making support for related fields, and thus playing an important role in improving energy efficiency, ensuring public safety, and optimizing operation strategies. With the continuous progress of quantum computing technology, future wind speed prediction models will be more accurate and efficient, making greater contributions to social and economic development. Description of the Drawings

[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0036] Figure 1 It is a method flowchart of a wind speed prediction method and system provided by an embodiment of the present invention;

[0037] Figure 2 It is a detailed flowchart of a wind speed prediction method and system provided by an embodiment of the present invention;

[0038] Figure 3 It is a schematic diagram of the VQC architecture of a wind speed prediction method and system provided by an embodiment of the present invention;

[0039] Figure 4 It is a schematic diagram of the QTCN module of a wind speed prediction method and system provided by an embodiment of the present invention;

[0040] Figure 5 It is a schematic diagram of quantum convolution and quantum dilated convolution of a wind speed prediction method and system provided by an embodiment of the present invention;

[0041] Figure 6 It is a schematic diagram of the QBiLSTM module of a wind speed prediction method and system provided by an embodiment of the present invention;

[0042] Figure 7 It is a schematic diagram of the LSTM structure of a wind speed prediction method and system provided by an embodiment of the present invention;

[0043] Figure 8 It is a schematic diagram of the variational quantum circuit of the QBiLSTM of a wind speed prediction method and system provided by an embodiment of the present invention;

[0044] Figure 9 It is a schematic diagram of the QSA module of a wind speed prediction method and system provided by an embodiment of the present invention;

[0045] Figure 10 It is a quantum circuit schematic diagram of the QSA module of a wind speed prediction method and system provided by an embodiment of the present invention;

[0046] Figure 11 It is an internal structure diagram of a computer device of a wind speed prediction method and system provided by an embodiment of the present invention. Detailed implementation manners

[0047] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings of the specification. Apparently, the described embodiments are some embodiments of the present invention, rather than all 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 scope of protection of the present invention.

[0048] Embodiment 1

[0049] Referring to Figures 1 - 11 , which is the first embodiment of the present invention. This embodiment provides a wind speed prediction method and system, including:

[0050] In the existing related technologies, there are some problems, such as the irregular fluctuations of wind speed data and the inaccuracy of prediction models, which pose challenges to the planning of wind power generation and the stable operation of the power grid.

[0051] This application provides a method that can effectively solve the above-mentioned problems. Next, multiple embodiments will be combined to elaborate in detail on how to implement this wind speed prediction method;

[0052] Figure 1 shows a method flowchart of a wind speed prediction method and system, including:

[0053] S101, obtaining first historical wind speed data in a target scenario and performing first preprocessing on the first historical wind speed data;

[0054] In an alternative embodiment, there are many scenarios where wind speed measurement is required. For example, it can be a wind farm, a weather station, the edge of a city, or other areas where wind power generation may be deployed. After obtaining these data, it is necessary to preprocess the data to ensure data quality, such as removing noise, filling in missing values, standardization, etc.

[0055] In another alternative embodiment, the scenarios where wind speed measurement is required can also include pre - predicting the wind speed in areas planned for wind power generation. In these areas, there may be no ready - made wind speed data, so it is necessary to predict through simulation or use data from historical similar scenarios. The preprocessing steps can include corresponding processing of these simulated data or historical data to ensure that they can be used for subsequent wind speed prediction.

[0056] In an alternative embodiment, the first historical wind speed data can include time - series data of wind speed. These data can be wind speed records over a past period of time, and can also be some parameter data related to wind speed, such as wind direction, temperature, humidity, etc. The preprocessing steps can include performing time - series analysis on these data to identify and eliminate possible outliers or trends, ensuring the accuracy and consistency of the data.

[0057] In the embodiments of the present application, the target scenario is not restricted, and those skilled in the art can select according to actual needs.

[0058] In an alternative embodiment, the first preprocessing may include determining the time step of the first historical wind speed data. The selection of the time step is crucial for the accuracy of wind speed prediction, as it determines the data sampling frequency and the resolution of the prediction model. For example, the time step can be set to 10 minutes, 1 hour, or a longer time interval, depending on the prediction requirements and the characteristics of the available data.

[0059] In another alternative embodiment, the first preprocessing may further include normalizing the first historical wind speed data to reduce the impact of different dimensions and numerical ranges on the prediction results. Various methods can be used for normalization, such as min-max normalization, Z-score standardization, etc., to ensure that the data is analyzed on a unified scale. In addition, the preprocessing step may also include handling missing values in the data, for example, by interpolation, deletion, or using statistical methods to estimate the missing values to ensure the integrity of the data.

[0060] In an alternative embodiment, the first preprocessing may further include filtering the noise of the first historical wind speed data to improve the data quality. The noise filtering can be achieved through various digital signal processing techniques. For example, a low-pass filter can be used to remove high-frequency noise, or a median filter can be used to reduce the impact of random noise. Through these preprocessing steps, it can be ensured that the data input into the wind speed prediction model is accurate and reliable, thereby improving the accuracy and efficiency of the prediction.

[0061] In the embodiments of the present application, the first preprocessing at least includes determining the time step of the first historical wind speed data.

[0062] Exemplarily, predicting the wind speed value at a certain future moment usually requires inputting the wind speed data of a continuous historical time period corresponding to that moment. Therefore, the setting of the time step of the historical data is particularly important. In this technology, through historical experience and experiments, the time step is determined to be 5, that is, the wind speed data signals of the past 5 moments are used, and the wind speed value at the 6th moment can be predicted most accurately. In addition, the data to be measured needs to be standardized to ensure that the data has the same scale as the model training data.

[0063] It should be noted that obtaining the first historical wind speed data in the target scenario and performing the first preprocessing on the first historical wind speed data can ensure that the data received by the wind speed prediction model is optimized, thereby improving the accuracy and reliability of the prediction. Through precise time step setting and data standardization, it can be ensured that the model can capture the subtle features of wind speed changes, which is particularly important for predicting short-term fluctuations in wind speed. These preprocessing steps are crucial for improving the overall performance of the wind speed prediction system, enabling the system to better adapt to various complex wind speed change situations, and providing strong technical support for the planning of wind power generation and the stable operation of the power grid.

[0064] S102, preset the first decomposition strategy, and decompose the first historical wind speed data after the first preprocessing according to the first decomposition strategy to obtain several groups of second data;

[0065] It should be noted that in this application, due to the volatility and randomness characteristics of wind speed data, in practical applications, it is usually necessary to combine certain processing methods to obtain relatively stable subsequences. Through these subsequences, the trends and patterns of wind speed changes can be captured more accurately.

[0066] In an optional embodiment, the wind speed data is decomposed into multiple frequency components, and each component represents a different scale of wind speed change. This decomposition method can effectively separate the trend term and the periodic term in the wind speed signal, thereby providing a clearer and more accurate data basis for subsequent wind speed prediction.

[0067] In an optional embodiment, the decomposed data can also be used to identify and analyze the periodic characteristics of wind speed changes, which has important practical significance for power prediction of wind farms and power grid dispatching. Through this decomposition strategy, the wind speed prediction system can better adapt to the complexity and uncertainty of wind speed data, thereby improving the accuracy and reliability of the prediction results.

[0068] In an optional embodiment, the first decomposition strategy can use the EMD algorithm, that is, the empirical mode decomposition method. This method decomposes the wind speed signal into a finite number of intrinsic mode functions (IMFs) and a residual term, thereby realizing the multi-scale decomposition of wind speed data. Each IMF represents an inherent oscillation mode in the signal, and these modes reflect different time scales of wind speed changes. In this way, the EMD algorithm can reveal the internal fluctuation characteristics of wind speed data and provide a more refined data processing result for subsequent wind speed prediction.

[0069] In an alternative embodiment, the first decomposition strategy may use the SSA algorithm, which is the singular spectrum analysis method. It extracts the main components of the signal by constructing a time series matrix of the wind speed signal and performing singular value decomposition on the matrix. The SSA algorithm is particularly suitable for processing wind speed data with non-linear and non-stationary characteristics, and can effectively separate the trend term, periodic term, and random term in the wind speed signal. In this way, the SSA algorithm helps to improve the accuracy of wind speed prediction, especially when the wind speed data exhibits complex change patterns.

[0070] In an alternative embodiment, the first decomposition strategy may use the ITD algorithm, which is the time delay embedding algorithm. It reveals the dynamic characteristics in the data by converting one-dimensional wind speed time series data into a multi-dimensional phase space. The core of this algorithm lies in reconstructing the phase space trajectory of the wind speed signal by introducing time delay and embedding dimension, so that the dynamic information originally hidden in the time series can be revealed. The ITD algorithm is particularly suitable for analyzing and predicting wind speed data with chaotic characteristics, and can help researchers capture the non-linear characteristics of wind speed changes, thus providing a more accurate model basis for wind speed prediction.

[0071] In the embodiment of the present application, the first decomposition strategy uses the VMD algorithm, which is the variational mode decomposition algorithm. It is an adaptive signal processing technology that can decompose a complex signal into a finite number of intrinsic mode functions. By minimizing the bandwidth and mode aliasing of the data, this algorithm effectively decomposes the wind speed signal into multiple sub-signals with different central frequencies. Each sub-signal represents a specific fluctuation pattern in the wind speed data, enabling the wind speed prediction model to be optimized for each mode and improving the overall prediction accuracy. The VMD algorithm performs excellently in processing non-linear and non-stationary signals and is particularly suitable for the analysis and prediction of wind speed data.

[0072] In the embodiment of the present application, the first decomposition strategy includes:

[0073] Perform data parsing on the first preprocessed first historical wind speed data to obtain a parsed signal;

[0074] Preset an objective function, which is used to minimize the sum of the bandwidths of the central frequencies of each mode component while ensuring that the sum of all mode components is equal to the original signal;

[0075] By optimizing the objective function and using singular value decomposition, obtain a set of mode functions for variational mode decomposition;

[0076] Demodulate the mode functions using a set of modulation signals with increasing frequencies to obtain multiple mode components.

[0077] Exemplarily, the present technical solution first uses the variational mode decomposition (VMD) algorithm to decompose the wind speed data sequence. The VMD algorithm is an adaptive and completely non-recursive modal variational and signal decomposition method. Its core idea is to construct and solve a variational problem, that is, assuming that the original data can be decomposed into components, and the VMD decomposition method uses iterative search to find the optimal solution of the variational model to determine the center frequency and bandwidth of each decomposed component, which belongs to a completely non-recursive model. This model searches for a set of modal components and their respective center frequencies, and each mode is smooth after being demodulated to the baseband. When using the VMD algorithm to decompose the wind speed data sequence, it is crucial to select an appropriate number of modes (value). If the value is set too small, it may lead to insufficient decomposition and inability to capture all important modes in the sequence; while when the value is set too large, it may cause modal overlap or introduce additional noise. Both of these situations may reduce the accuracy of the wind power prediction model. Therefore, a reasonable selection of the value is crucial to ensure that the VMD algorithm can effectively decompose the wind speed sequence and improve the performance of the prediction model. The steps of the variational mode decomposition (VMD) algorithm mainly include the following stages:

[0078] Step (1): Hilbert transform. First, perform a Hilbert transform on the input wind speed signal data to obtain an analytic signal. This step is to convert the signal into a complex form for subsequent processing.

[0079] Step (2): Define a regularization term and an objective function. The objective function aims to minimize the sum of the bandwidths of the center frequencies of each modal component while ensuring that the sum of all modal components is equal to the original signal.

[0080] Step (3): By optimizing the objective function and using the singular value decomposition (SVD) method, a set of modal functions of VMD is obtained.

[0081] Step (4): Use a set of modulation signals with increasing frequencies for demodulation to obtain multiple modal components.

[0082] Step (5): Through an optimization problem-solving method, gradually update the modulation signal and the modal components.

[0083] Step (6): The algorithm iterates continuously until the convergence condition is met to obtain the final modal components.

[0084] It should be noted that after VMD processing, the wind speed data is decomposed into each component, and then through a hybrid quantum-classical neural network architecture composed of a quantum time-domain convolutional network (QTCN), a quantum bidirectional long short-term memory network (QBiLSTM), and a quantum self-attention mechanism (QSA), the corresponding wind speed values are predicted from each component respectively. Finally, the predicted values are superimposed to obtain the final wind speed prediction value.

[0085] It should also be noted that by presetting the first decomposition strategy to decompose the first preprocessed historical wind speed data, several groups of second data can be obtained, which can more accurately capture the dynamic characteristics of wind speed changes. The preset decomposition strategies include, but are not limited to, empirical mode decomposition (EMD), wavelet transform (WT), etc. These methods can adaptively decompose the signal into a finite number of intrinsic mode functions (IMFs) or wavelet coefficients according to the inherent characteristics of the wind speed data. In this way, the trend term and periodic term in the wind speed signal can be effectively separated, providing a clearer and more discriminative data input for the subsequent prediction model.

[0086] S103, Use several groups of second data as the input for pre-training the first network architecture, and perform wind speed prediction according to the output of the first network architecture.

[0087] In the embodiment of the present application, performing wind speed prediction according to the output of the first network architecture includes:

[0088] Each group of second data corresponds to an output of the first network architecture;

[0089] Perform a first merge on the outputs of all the first network architectures;

[0090] Perform wind speed prediction according to the first merge result.

[0091] In the embodiment of the present application, the first network architecture includes: The first network architecture is any network architecture with several groups of second data as the input and the wind speed value or relevant parameters that can directly or indirectly obtain the wind speed value as the output.

[0092] In an optional embodiment, the first network architecture can be implemented by using a convolutional neural network (CNN). This network can automatically extract the features of the input data and learn the complex patterns in the data through a multi-layer structure. CNN performs excellently in the field of image recognition and processing and is also applicable to processing time series data such as wind speed data. Through the combination of convolutional layers, pooling layers, and fully connected layers, the network can capture the local features and overall trends of wind speed changes, thereby improving the prediction accuracy. In addition, the network can be trained through the backpropagation algorithm to continuously optimize the network weights to achieve the best prediction effect.

[0093] In an alternative embodiment, the first network architecture can also be implemented by using a Recurrent Neural Network (RNN), which is particularly suitable for processing sequential data. The RNN is capable of remembering previous information and using this information to influence subsequent outputs, which is very useful for predicting time-dependent data such as wind speed. Through its internal recurrent structure, the RNN can capture the dynamic characteristics of wind speed data changing over time, so that the influence of historical data can be taken into account when predicting future wind speeds. In addition, variants of the RNN, such as Long Short-Term Memory networks (LSTMs) and Gated Recurrent Units (GRUs), can better solve the problem of vanishing or exploding gradients in traditional RNNs for long sequence data, further improving the stability and accuracy of predictions.

[0094] In an alternative embodiment, the first network architecture can also be designed by improving a hybrid quantum-classical neural network to utilize the parallel processing power of quantum computing to accelerate the network training process. The hybrid quantum-classical neural network combines the flexibility of classical neural networks and the high efficiency of quantum computing. Through the superposition and entanglement states of qubits, the computing time can be significantly reduced when processing large-scale data sets. This network architecture is particularly suitable for prediction tasks that require processing a large amount of wind speed data, and can provide faster training speed and higher prediction accuracy than traditional networks.

[0095] In the embodiment of the present application, the first network architecture is designed by composing a QTCN module, a QBiLSTM module, and a QSA module. The specific design of the QTCN module is as follows:

[0096] After being processed by VMD, the wind speed data is decomposed into components, and then sent into a Quantum Temporal Convolutional Network (QTCN) as the underlying structure to receive the spatial data of multiple IMF components obtained by VMD decomposition. The reason for choosing the QTCN model is that this model can well preserve the original topological structure of the data, and uses causal convolutions to ensure that only the data at the current time and before is used when predicting the value at the current time. Dilated convolutions are used to expand the receptive field of the convolutional layer, enabling the network to capture long-range sequence dependencies without increasing the number of parameters or computational complexity. At the same time, TCN can process the data of all time steps in parallel, which makes TCN more efficient in processing long sequences. The model structure of QTCN is as Figure 4 shown.

[0097] Among them, it includes a quantum convolutional layer (QConv) with a convolutional kernel and two quantum dilated convolutional layers (QDCN). They are both one-dimensional convolutional layers Conv1D for two-dimensional matrices with the input, and are both implemented by the same variational quantum circuit; the normalization layer uses WeightNorm, which can bring faster convergence speed and stronger learning rate robustness, and has better optimization performance in dynamic networks such as RNN; in addition, the activation function is ReLU, and the Dropout strategy is used simultaneously to optimize the model to avoid overfitting. Quantum dilated convolution is a special convolution that expands the kernel by inserting holes (i.e., points with zero weights) between consecutive kernel elements.

[0098] It should be noted that compared with standard convolution, dilated convolution introduces an additional hyperparameter called the dilation rate, which determines the stride at which input pixels are sampled.

[0099] In an optional embodiment, compared with standard convolution with the same kernel size, dilated convolution can capture a larger receptive field without introducing more learnable parameters. Compared with standard quantum convolution with the same kernel size, quantum dilated convolution does not require more qubits than, and the specific differences and variational quantum circuit design are as Figure 5 shown. The green dashed box is the standard quantum convolution, and the blue dashed box is the quantum dilated convolution.

[0100] In an optional embodiment, the quantum logic gates involved in the quantum circuit are H, RY, RX, RZ, and U 3 gates, X gates, and CNOT gates. The H gate is used to initialize the quantum state of the qubit, changing the ground state to a superposition state; the combination of RY and RX constitutes the angular encoding part of classical data. Angular encoding is a technique used for data representation in quantum machine learning, which uses the rotation of quantum gates (RX, RY, and RZ) to encode classical information. This method encodes N features of classical data as the angles between n input qubits of quantum states.

[0101] In an optional embodiment, N is kept equal to n to enable the quantum network layer to use the maximum size of classical features as much as possible. The quantum state generated by performing angular encoding on the input qubits can be represented by the following formula:

[0102]

[0103] where R(.) can be any one of the RX, RY, and RZ logic gates. In angular encoding, the angles between quantum states can vary continuously to capture complex data. This leads to more accurate and detailed data representation and can improve the performance of certain types of quantum machine learning models.

[0104] It should be noted that although it can only encode one eigenvalue into one qubit, it reduces noise, which makes it particularly advantageous in NISQ computing. The subsequent U 3 gates, RZ gates, and RY gates are all logic gates with optimizable parameters and belong to the variational layer. The X gate realizes the quantum state flip of adjacent qubits, and the CNOT gate realizes quantum entanglement.

[0105] In the embodiment of the present application, the QBiLSTM module is specifically designed as follows:

[0106] Since QTCN can flexibly adjust the receptive field size through dilated convolution to adapt to features of different time scales, multiple IMF components can effectively extract unique local features in each component's data through the QTCN module. Then, these features are processed by the quantum bidirectional long short-term memory network (QBiLSTM) to learn the high-dimensional feature representation of the sequence, so as to enhance the overall feature representation ability of the model.

[0107] It should be noted that the core feature of QBiLSTM is that it has two QLSTM layers, one for processing forward (from past to future) sequence information and the other for processing backward (from future to past) sequence information. This bidirectional processing enables the model to consider the context information of the sequence simultaneously, which is very useful for prediction tasks (such as stock prices, weather forecasts, etc.). The structure of QBiLSTM is as Figure 6 shown.

[0108] Furthermore, the long short-term memory network (LSTM) is a special RNN. As Figure 7 shown in the LSTM structure, it can learn the sequential dependencies over a longer range in the data. It solves the important problem of gradient vanishing in the original RNN: each LSTM unit at time step t has an additional cell state, denoted by c t (referred to as the cell state), which allows the gradient to flow without change and can be regarded as the memory of the LSTM unit (therefore, LSTM has two memory components h t and c t , while the RNN only has c t ). This property makes LSTM numerically more stable during training and more accurate in prediction. LSTM consists of three gates and a cell state, where the mathematical formula: σ represents the sigmoid activation function, f t represents the forget gate, that is, which part of the information the network is going to discard; x t represents the input, that is, the data stream input to the network; i t represents the update gate; represents the cell state; o t represents the output gate; h represents the hidden state.

[0109] In QBiLSTM, the transformation operations of the forget gate, update gate, and output gate are replaced by VQC. There are a total of 6 VQC structures, all of which have the same quantum circuit design. The specific design is as follows Figure 8 shown, Ch 1 , Ch 2 , Ch 3 , Ch 4 are the input eigenvalues. The RY gates within the blue boxes are used for angle encoding, and the RY within the green boxes are logical gates with optimizable parameters. There are also CNOT gates to achieve quantum entanglement.

[0110] In the embodiment of this application, the specific design of the QSA module is as follows:

[0111] The output of QBiLSTM is further processed by the quantum self-attention (QSA) module. The structure of the QSA module is as Figure 9 shown, and the schematic diagram of the quantum circuit of the QSA module is as Figure 10 shown. The attention mechanism can enhance the role of important time steps in QBiLSTM, thereby further reducing the model prediction error. The output vector of the QLSTM hidden layer is used as the input of the attention layer and is trained through three quantum network layers with the same quantum circuit design but different optimizable parameters. Then, the outputs of the quantum network layers are multiplied and normalized using the softmax activation function, which can be specifically expressed as follows:

[0112] S i = VQC(x, W i + b i ) i = 1, 2, 3

[0113] y = softmax(S 1 × S 2 ) × S 3

[0114] where x is the input, W i , b i are the weights and biases, S i is the output of each VQC, and y is the final output, that is, the predicted wind speed value obtained through each VMD component.

[0115] In the embodiment of this application, the first network architecture further includes: several first network architectures all have the same architecture;

[0116] The inputs of several first network architectures are different groups of second data randomly assigned;

[0117] After obtaining the outputs of all first network architectures, a first merge is performed.

[0118] In an alternative embodiment, the first merging can be achieved by weighted averaging, where the output of each first network architecture is assigned a different weight according to its performance. The determination of the weights can be dynamically adjusted based on the prediction error on the validation set to ensure the accuracy of the overall prediction result.

[0119] In an alternative embodiment, the first merging can also be accomplished through a voting mechanism, i.e., each first network architecture outputs a predicted value, and the final prediction result is the median of the predicted values of the majority of the network architectures. This method can reduce the influence of outliers and improve the robustness of the overall prediction. Additionally, other statistical methods, such as the median or mean, can be used to combine the prediction results of different network architectures to achieve the best prediction effect.

[0120] In an alternative embodiment, the first merging can also be realized by an ensemble learning method, which involves taking the prediction results of multiple first network architectures as a basis and combining these results through a specific algorithm in order to obtain better prediction performance than a single model. For example, ensemble learning techniques such as random forest or gradient boosting tree can be adopted, which can effectively combine the predictions of multiple models, thereby improving the generalization ability and prediction accuracy of the overall model. In this way, the first merging can not only utilize the advantages of each network architecture but also alleviate the overfitting problem to a certain extent, further improving the accuracy of wind speed prediction.

[0121] In the embodiment of the present application, the first merging includes: performing a weighted sum on the outputs of all the first network architectures according to preset weights.

[0122] Exemplarily, k VMD decomposition components obtain k predicted wind speed values after passing through a hybrid quantum-classical neural network architecture, and finally, the predicted values are weighted and fused to obtain the final wind speed prediction value.

[0123] During the model training process, assume that the k predicted wind speed values are y 1 , y 2 , y 3 , …, y k , the true wind speed value is y, and the model hopes to perform a weighted average on these prediction results through weights w 1 , w 2 , w 3 , …, w k to obtain the comprehensive prediction result of the model That is

[0124]

[0125] It should be noted that, in order to minimize the error between the comprehensive prediction result of the model and the true value, a loss function is designed during the training process to optimize the weights w 1 , w2 , w 3 , …, w k Optimize to obtain the optimal solution and get the final predicted wind speed value.

[0126] It should be noted that wind speed prediction plays a crucial role in multiple key fields, including renewable energy management, meteorological services, aviation and navigation, and environmental monitoring, etc. In the field of renewable energy, accurate wind speed prediction is crucial for the optimal scheduling of wind power generation and improving the utilization rate of wind energy, which helps to balance the grid load and ensure the stability and economy of energy supply. In meteorological services, wind speed information is an important part of weather prediction models and is of great significance for early warning of extreme weather events and reducing losses caused by natural disasters. For the aviation and navigation industries, wind speed data is also crucial for route planning, flight safety, and navigation efficiency.

[0127] It should also be noted that one of the main motivations for designing a high-precision wind speed prediction model is to improve the accuracy and reliability of the prediction. By introducing quantum computing technology, we can enhance the ability of deep learning models to process complex wind speed patterns. The parallelism and efficiency of quantum computing enable the model to quickly process a large amount of meteorological data and identify the nonlinear relationships and hidden patterns therein. In addition, the advantages of quantum algorithms in feature space mapping and pattern recognition help to capture the subtle differences in wind speed changes, thus improving the model's prediction ability for wind speed change trends.

[0128] It should also be noted that the introduction of quantum computing, especially in the field of quantum machine learning, provides new tools and methods for wind speed prediction models. The application of algorithms such as quantum long short-term memory network (QLSTM) and quantum support vector machine (QSVM) enables the model to effectively learn and generalize in high-dimensional feature spaces. The integration of these quantum algorithms can not only improve the accuracy of wind speed prediction but also reduce the consumption of computing resources during model training and prediction, and improve the real-time response ability of the prediction system.

[0129] It should also be noted that by combining quantum computing and deep learning technologies, the wind speed prediction model can more accurately capture and predict wind speed changes, provide more reliable decision-making support for related fields, and thus play an important role in improving energy efficiency, ensuring public safety, and optimizing operation strategies. With the continuous progress of quantum computing technology, future wind speed prediction models will be more accurate and efficient, making greater contributions to social and economic development.

[0130] In summary, the present invention proposes a wind speed prediction method, which obtains the first historical wind speed data in a target scenario and performs a first preprocessing on the first historical wind speed data; presets a first decomposition strategy, decomposes the first preprocessed first historical wind speed data to obtain several groups of second data; uses the several groups of second data as the input for pre-training the first network architecture, and performs wind speed prediction according to the output of the first network architecture. It can effectively improve the accuracy of wind speed prediction, thereby providing a more reliable reference for the wind power generation planning and operation of the power grid. By adopting advanced data processing and network architecture, the present application can handle the complexity and uncertainty of wind speed data and provide technical support for the optimized management of wind power generation. In addition, the prediction system of the present application has good scalability and adaptability, can adapt to the wind speed prediction requirements in different scenarios, and further improve the efficiency and stability of wind power generation. Additionally, by combining quantum computing and deep learning technologies, the wind speed prediction model can more accurately capture and predict wind speed changes, provide more reliable decision-making support for related fields, and thus play an important role in improving energy efficiency, ensuring public safety, and optimizing operation strategies. With the continuous progress of quantum computing technology, future wind speed prediction models will be more accurate and efficient, making greater contributions to social and economic development.

[0131] Embodiment 2

[0132] In a preferred embodiment, the technical solution realizes the prediction of wind speed data at future moments through a method combining signal decomposition and a deep learning model. The specific framework of the algorithm is as Figure 2 shown. The algorithm mainly includes six parts: data preprocessing, VMD decomposition, QTCN module, QBiLSTM module, QSA module, and prediction value merging. Among them, the three modules of QTCN, QBiLSTM, and QSA are all hybrid quantum-classical neural network modules, which are composed of a classical network layer and a quantum network layer based on a variational quantum circuit (VQC).

[0133] VQC is a quantum circuit containing parameterized quantum logic gates, whose parameters are adjustable and can be iteratively optimized. The general VQC architecture is as Figure 3 shown. Among them, the block is used for state preparation to encode classical data into the quantum state of the circuit and is not affected by optimization. The block represents the variational part with learnable parameters, which will be optimized by the gradient method. Some existing research results show that this kind of circuit has strong robustness to quantum noise, so it is suitable for NISQ devices. VQC has been successfully applied to tasks such as function approximation, classification, generative modeling, deep reinforcement learning, and transfer learning. In addition, VQC is more expressive than classical neural networks and may therefore be better than the latter. Here, the expression ability refers to the ability to represent certain functions or distributions with a finite number of parameters.

[0134] Example 3

[0135] In this embodiment, a wind speed prediction system is further provided, including:

[0136] A data acquisition and processing module, configured to acquire first historical wind speed data in a target scenario and perform first preprocessing on the first historical wind speed data;

[0137] A data decomposition module, configured to preset a first decomposition strategy, decompose the first historical wind speed data after the first preprocessing, and obtain several groups of second data;

[0138] A prediction module, configured to use several groups of second data as the input for pre-training a first network architecture, and perform wind speed prediction according to the output of the first network architecture.

[0139] The above-mentioned unit modules can be embedded in or independent of a processor in a computer device in the form of hardware, or stored in a memory in a computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above respective modules.

[0140] This embodiment further provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in Figure 11 shown. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The computer program, when executed by the processor, implements a wind speed prediction method. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0141] This embodiment further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0142] Acquire first historical wind speed data in a target scenario and perform first preprocessing on the first historical wind speed data;

[0143] Preset a first decomposition strategy to decompose the first preprocessed historical wind speed data to obtain several groups of second data;

[0144] Use several groups of second data as the input for pre-training the first network architecture, and perform wind speed prediction based on the output of the first network architecture.

[0145] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

[0146] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0147] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0148] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0149] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, causing a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process or a plurality of processes and / or blocks. Figure 1 one process or a plurality of processes and / or blocks Figure 1 steps for implementing the functions specified in one block or a plurality of blocks.

[0150] Although the preferred embodiments of the present application have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic creative concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications that fall within the scope of the present application.

[0151] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. A wind speed prediction method, characterized in that: include: Acquire first historical wind speed data in a target scene, and perform first preprocessing on the first historical wind speed data; Preset a first decomposition strategy, and decompose the first preprocessed first historical wind speed data according to the first decomposition strategy to obtain a plurality of groups of second data; The plurality of groups of second data are used as inputs of a pre-trained first network architecture, and wind speed prediction is performed according to an output of the first network architecture.

2. The wind speed prediction method according to claim 1, characterized in that: The performing wind speed prediction according to the output of the first network architecture comprises: Different groups of second data correspond to outputs of a first network architecture; Performing a first merging on all outputs of the first network architecture; Wind speed prediction is performed according to the first merging result.

3. The wind speed prediction method according to claim 2, characterized in that: The first network architecture includes: the first network architecture is any network architecture that takes as input several groups of second data and outputs wind speed values ​​or related parameters of wind speed values ​​that can be directly or indirectly obtained.

4. The wind speed prediction method according to claim 3, characterized in that: The first network architecture also includes: a plurality of first network architectures are all the same architecture; The inputs of the plurality of first network architectures are different sets of randomly assigned second data; After obtaining all outputs of the first network architecture, a first merge is performed.

5. The wind speed prediction method according to claim 4, characterized in that: The first preprocessing includes at least determining a time step of the first historical wind speed data.

6. The wind speed prediction method according to claim 5, characterized in that: The first decomposition strategy includes: Performing data analysis on the first historical wind speed data after the first preprocessing to obtain an analysis signal; Presetting an objective function, the objective function is used to minimize the sum of the bandwidths of the center frequencies of each modal component while ensuring that the sum of all modal components is equal to the original signal; By optimizing the objective function and using singular value decomposition, a set of modal functions of variational modal decomposition is obtained; The modal function is demodulated using a set of modulation signals with increasing frequencies to obtain a plurality of modal components.

7. The wind speed prediction method according to claim 6, characterized in that: The first merging includes: performing weighted summation on the outputs of all the first network architectures according to preset weights.

8. A wind speed prediction system, characterized in that: include: A data acquisition and processing module, used to acquire first historical wind speed data in a target scene, and perform first preprocessing on the first historical wind speed data; A data decomposition module, configured to preset a first decomposition strategy, and decompose the first preprocessed first historical wind speed data according to the first decomposition strategy to obtain a plurality of groups of second data; The prediction module is used to use the plurality of groups of second data as input of the pre-trained first network architecture and perform wind speed prediction according to the output of the first network architecture.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images