Power generation prediction method, device, equipment, medium and product
Patent Information
- Application Number
- CN202610839902.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-11
- Publication Date
- 2026-09-22
AI Technical Summary
尽管它们可以在一定程度上揭示数据间的线性相关性,但仍然无法有效处理复杂非线性关系以及数据中的动态变化特性
[0055]如此,通过监测水电站运行过程中的待预测发电数据,动态捕捉因果特征,自动聚焦对发电量影响显著的关键因素,从而更准确地反映不同因素在不同时期对发电量的实际影响,而通过对因果特征和待预测发电数据融合得到的融合数据进行时序特征提取,从而以提取得到的时序特征作为发电量预测的依据,提高了水电站的发电量预测准确性。
Smart Images

Figure CN122801210A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of big data technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting power generation. Background Technology
[0002] With the continuous rise in my country's energy demand and the accelerated construction of new power systems, hydropower stations, with their rapid response and flexible regulation capabilities, have become key peak-shaving power sources supporting energy security, ensuring stable system operation, and improving economic efficiency. Faced with increasingly severe power load fluctuations and the pressure of real-time power generation and consumption balance, especially for cascade hydropower stations, due to their complex cascade hydraulic coupling characteristics, variable hydrological and meteorological conditions, and multiple scheduling constraints, coordinating efficient water resource utilization with dynamic grid demand presents significant challenges. Therefore, achieving accurate prediction of their power generation, especially their short- and medium-term generation capacity, has become a core link and important technical support for improving the efficiency of hydropower resource utilization, optimizing scheduling decisions, and ensuring reliable power supply.
[0003] In the field of hydropower generation prediction, current mainstream technologies still rely on traditional statistical methods and basic machine learning models (SVMA, RIMA, etc.). These methods typically utilize historical power generation data, combined with meteorological and hydrological information, to construct predictive models through linear regression analysis, thereby inferring future power generation. While they can reveal linear correlations between data to some extent, they still cannot effectively handle complex nonlinear relationships and dynamic changes in the data. Furthermore, although some time-series-based prediction models (grey prediction models, LSTM models, etc.) can play a role in processing single data sequences, their prediction accuracy often falls short when dealing with interactions between multiple variables and the complex operating conditions of hydropower stations. They fail to meet the complex and ever-changing needs of hydropower production scenarios and struggle to deeply uncover the underlying patterns in the data, thus hindering refined management and optimized scheduling decisions for hydropower stations. In summary, this results in relatively low accuracy in hydropower generation prediction. Summary of the Invention
[0004] Therefore, it is necessary to provide a power generation prediction method, device, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of power generation prediction for hydropower stations, addressing the aforementioned technical problems.
[0005] In a first aspect, this application provides a method for predicting power generation, including:
[0006] Obtain the predicted power generation data of the hydropower station;
[0007] Dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, and the causal features are fused with the power generation data to be predicted to obtain fused data;
[0008] Temporal features are extracted from the fused data to obtain temporal features;
[0009] Based on the aforementioned time-series characteristics, the power generation of the hydropower station is predicted to obtain the target power generation.
[0010] In one embodiment, the step of extracting dynamic causal features from the power generation data to be predicted to obtain causal features includes:
[0011] Determine the amount of data change corresponding to the power generation data to be predicted;
[0012] Based on the amount of data change corresponding to the power generation data to be predicted, determine the causal relationship change information of the power generation data to be predicted;
[0013] Based on the causal relationship change information, attention weight information is constructed;
[0014] Based on the attention weight information, dynamic causal features are extracted from the power generation data to be predicted to obtain causal features.
[0015] In one embodiment, the step of extracting temporal features from the fused data to obtain temporal features includes:
[0016] The fused data is then weighted to obtain weighted data;
[0017] Short-term fluctuation features are extracted from the weighted data to obtain the first feature;
[0018] The second feature is obtained by extracting intermediate trend features from the first feature and the weighted data;
[0019] Periodic features are extracted from the second feature and the weighted data to obtain time-series features.
[0020] In one embodiment, the weighting process of the fused data to obtain weighted data includes:
[0021] Determine the feature importance information of each feature in the fused data at the current time step;
[0022] The feature importance information of each feature at the current time step is normalized and fused to obtain the feature weight coefficient of the fused data at the current time step;
[0023] The fused data is weighted according to the feature weight coefficients to obtain weighted data.
[0024] In one embodiment, the process of predicting power generation is performed using a pre-trained power generation prediction model, the training process of which includes:
[0025] Obtain historical power generation data corresponding to the hydropower station;
[0026] Determine the correlation between each data dimension in the historical power generation data and the power generation, and filter target power generation data from the historical power generation data based on each correlation.
[0027] The target power generation data is converted into supervised learning data;
[0028] The dynamic causal attention network to be trained and the hidden Markov tree model to be trained are fused to obtain the fused model to be trained.
[0029] Based on the supervised learning data, the fusion model to be trained is iteratively trained to obtain a power generation prediction model, wherein the fusion model to be trained is obtained by fusing a dynamic causal attention network to be trained and a hidden Markov tree model to be trained.
[0030] In one embodiment, converting the target power generation data into supervised learning data includes:
[0031] Get the parameters of the sliding window;
[0032] The target power generation data is traversed through the sliding window corresponding to the sliding window parameters to obtain supervised learning data.
[0033] Secondly, this application also provides a power generation prediction device, comprising:
[0034] The acquisition module is used to acquire the predicted power generation data of the hydropower station;
[0035] The feature extraction module is used to perform dynamic causal feature extraction on the power generation data to be predicted, and obtain causal features;
[0036] The feature fusion module fuses the causal features with the power generation data to be predicted to obtain fused data;
[0037] The feature extraction module is also used to extract temporal features from the fused data to obtain temporal features;
[0038] The prediction module is used to predict the power generation of the hydropower station based on the time-series characteristics, and obtain the target power generation.
[0039] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0040] Obtain the predicted power generation data of the hydropower station;
[0041] Dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, and the causal features are fused with the power generation data to be predicted to obtain fused data;
[0042] Temporal features are extracted from the fused data to obtain temporal features;
[0043] Based on the aforementioned time-series characteristics, the power generation of the hydropower station is predicted to obtain the target power generation.
[0044] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:
[0045] Obtain the predicted power generation data of the hydropower station;
[0046] Dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, and the causal features are fused with the power generation data to be predicted to obtain fused data;
[0047] Temporal features are extracted from the fused data to obtain temporal features;
[0048] Based on the aforementioned time-series characteristics, the power generation of the hydropower station is predicted to obtain the target power generation.
[0049] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:
[0050] Obtain the predicted power generation data of the hydropower station;
[0051] Dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, and the causal features are fused with the power generation data to be predicted to obtain fused data;
[0052] Temporal features are extracted from the fused data to obtain temporal features;
[0053] Based on the aforementioned time-series characteristics, the power generation of the hydropower station is predicted to obtain the target power generation.
[0054] The aforementioned power generation prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product acquire power generation data to be predicted from a hydropower station; extract dynamic causal features from the power generation data to be predicted to obtain causal features; fuse the causal features with the power generation data to be predicted to obtain fused data; extract time-series features from the fused data to obtain time-series features; and predict the power generation of the hydropower station based on the time-series features to obtain the target power generation.
[0055] In this way, by monitoring the power generation data to be predicted during the operation of the hydropower station, the causal characteristics are dynamically captured, and the key factors that have a significant impact on power generation are automatically focused on. This allows for a more accurate reflection of the actual impact of different factors on power generation at different times. Furthermore, by extracting time-series features from the fused data obtained by fusing causal characteristics and power generation data to be predicted, the extracted time-series features can be used as the basis for power generation prediction, thereby improving the accuracy of power generation prediction for hydropower stations. Attached Figure Description
[0056] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0057] Figure 1 This is a diagram illustrating the application environment of a power generation prediction method in one embodiment.
[0058] Figure 2 This is a flowchart illustrating a power generation prediction method in one embodiment;
[0059] Figure 3 This is a structural block diagram of a power generation prediction device in one embodiment;
[0060] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0062] It should be noted that all information and data involved in this application (including but not limited to data used for analysis, stored data, and displayed data) are information and data authorized by the user or fully authorized by all parties, and the acquisition, transmission, storage, use, and processing of related data comply with the relevant provisions of national laws and regulations. Users can refuse or easily refuse content pushed to them (e.g., target power generation). In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, and their purpose is merely to illustrate the feasibility of implementing the technical solution of this application, but does not mean that the applicant has already used or necessarily used such a solution.
[0063] The power generation prediction method provided in this application can be applied to, for example, Figure 1 In the application environment shown, the hydropower station 102 communicates with the server 104 via a network. A data storage system can store the data that the server 104 needs to process. The data storage system can be integrated onto the server 104, or it can be located in the cloud or on another network server. The server 104 acquires the hydropower station's power generation data to be predicted; it performs dynamic causal feature extraction on the power generation data to be predicted to obtain causal features, and fuses the causal features with the power generation data to be predicted to obtain fused data; it extracts time-series features from the fused data to obtain time-series features; and based on the time-series features, it predicts the power generation of the hydropower station to obtain the target power generation. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.
[0064] In one exemplary embodiment, such as Figure 2 As shown, a power generation prediction method is provided, which can be applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 202 to 208. Wherein:
[0065] Step 202: Obtain the predicted power generation data of the hydropower station.
[0066] The power generation data to be predicted in step 202 includes at least one of the following: power generation capacity, inflow, temperature, water level, and precipitation.
[0067] Step 204: Dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, and the causal features are fused with the power generation data to be predicted to obtain fused data.
[0068] As one embodiment, dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, including: determining the amount of data change corresponding to the power generation data to be predicted; determining the causal relationship change information of the power generation data to be predicted based on the amount of data change corresponding to the power generation data to be predicted; constructing attention weight information based on the causal relationship change information; and performing dynamic causal feature extraction on the power generation data to be predicted based on the attention weight information to obtain causal features.
[0069] As one embodiment, fusing the causal features with the power generation data to be predicted to obtain fused data includes: linearly correcting the causal features and the power generation data to be predicted to obtain linearly corrected data, and performing regression activation processing on the linearly corrected data to obtain fused data.
[0070] Specifically, the causal relationship change information of the power generation data to be predicted can be determined by jointly implementing a first-layer fully connected neural network containing 128 neurons and using ReLU as the activation function, and a second-layer fully connected neural network containing 64 neurons and using Tanh as the activation function.
[0071] Thus, the first layer of the fully connected neural network with 128 neurons can fully extract the complex correlation information between "time step and the power generation data to be predicted"; the subsequent compression to 64 neurons can filter redundant information, avoid overfitting, and retain the core features of causal relationships, adapting to the correlation complexity of hydropower station data (capturing the causal logic between features without deeper networks). ReLU can alleviate the gradient vanishing problem during the first layer feature extraction, ensuring the effective transmission of feature information; the output range of Tanh is [-1,1], which can adapt to the representation requirements of "causal relationships" (causal relationships can be reflected as positive or negative correlations, and their strength).
[0072] Specifically, by linearly correcting the causal features and the power generation data to be predicted, the linearly corrected data can be obtained through a fully connected neural network structure containing 128 neurons and using ReLU as the activation function. The linearly corrected data is then subjected to regression activation processing to obtain fused data, which can be obtained through a fully connected neural network structure containing 64 neurons and using Sigmoid as the activation function.
[0073] Thus, maintaining the same number of layers and neurons as the function preserves the continuity of feature dimensions, avoids loss of fusion information caused by abrupt changes in feature dimensions, and balances feature expressiveness with computational efficiency. ReLU is used for feature fusion extraction before the second layer to ensure efficient information transmission; the output range of Sigmoid is [0,1], which can normalize the fusion result, adapt to the input requirements of subsequent model modules (such as the hierarchical memory Transformer), and facilitate the weight allocation of features by the attention mechanism.
[0074] Step 206: Extract time-series features from the fused data to obtain time-series features.
[0075] As an embodiment, step 206 includes: weighting the fused data to obtain weighted data; extracting short-term fluctuation features from the weighted data to obtain a first feature; extracting medium-term trend features from the first feature and the weighted data to obtain a second feature; and extracting periodic features from the second feature and the weighted data to obtain a time-series feature.
[0076] Specifically, short-term fluctuation features are extracted from the weighted data to obtain the first feature, which is implemented through a bottom-level temporal feature layer. The bottom-level temporal feature layer contains three Transformer encoder modules, each containing a multi-head self-attention mechanism (with eight self-attention heads) and a feedforward neural network. Mid-term trend features are extracted from the first feature and the weighted data to obtain the second feature, which is implemented through a middle-level temporal feature layer. The middle-level temporal feature layer contains three Transformer encoder modules, with a structure similar to the bottom layer, and six self-attention heads. Periodic features are extracted from the second feature and the weighted data to obtain the temporal feature, which is implemented through a high-level temporal feature layer. The high-level temporal feature layer contains two Transformer encoder modules and four self-attention heads.
[0077] As one embodiment, the fused data is weighted to obtain weighted data, including: determining the feature importance information of each feature in the fused data at the current time step; normalizing and fusing the feature importance information of each feature at the current time step to obtain the feature weight coefficient of the fused data at the current time step; and weighting the fused data according to the feature weight coefficient to obtain weighted data.
[0078] Step 208: Based on the time series characteristics, predict the power generation of the hydropower station to obtain the target power generation.
[0079] Optionally, the process of predicting power generation is performed through a pre-trained power generation prediction model. The training process of the power generation prediction model includes: acquiring historical power generation data corresponding to the hydropower station; determining the correlation between each data dimension in the historical power generation data and the power generation, and selecting target power generation data from the historical power generation data according to each correlation; converting the target power generation data into supervised learning data; fusing the dynamic causal attention network to be trained and the hidden Markov tree model to be trained to obtain a fused model to be trained; and iteratively training the fused model to be trained according to the supervised learning data to obtain the power generation prediction model, wherein the fused model to be trained is obtained by fusing the dynamic causal attention network to be trained and the hidden Markov tree model to be trained.
[0080] Historical power generation data includes at least one of the following: historical power generation, historical power output, historical inflow, historical temperature, historical precipitation, and historical water level.
[0081] It is understandable that historical power generation data may contain anomalies. To mitigate these issues, the aforementioned methods also include outlier detection and processing of historical power generation data. Outlier detection methods include calculating the data mean, calculating the standard deviation, and using box plots. Outlier processing methods include correcting outliers based on specific circumstances or making reasonable estimates using domain knowledge to correct them.
[0082] Optionally, the above method further includes: performing standardization operations on historical power generation data. These standardization operations include, but are not limited to, Z-score standardization and Min-Max standardization.
[0083] As an example, the correlation between each data dimension in the historical power generation data and the power generation can be determined by calculating the Pearson correlation coefficient and the Spearman correlation coefficient. The target power generation data includes power generation characteristic data and power generation. The power generation characteristic data includes at least one of the following: historical end-of-period water level, historical temperature, historical precipitation, historical power generation, and historical inflow.
[0084] Further, converting the target power generation data into supervised learning data includes: obtaining sliding window parameters; and traversing the target power generation data through the sliding window corresponding to the sliding window parameters to obtain supervised learning data.
[0085] Among them, the sliding window parameter is used to characterize the sliding window size and prediction step size.
[0086] As one embodiment, the target power generation data is traversed through the sliding window corresponding to the sliding window parameter to obtain supervised learning data, including: extracting continuous power generation feature data from the target power generation data that matches the size of the sliding window, and determining the continuous power generation feature data as the input feature data in the supervised learning data; selecting the power generation that matches the prediction step size after the sliding window, and determining it as the true label in the supervised learning data; traversing the target power generation data to obtain multiple sets of supervised learning data.
[0087] Thus, input features and corresponding labels are generated by sliding a fixed-length window. This method allows the model to use data from past time points to predict data from future time points. In this way, the sequential information and temporal dependencies of the time series are preserved, and the model can learn dynamic patterns and trends in the data.
[0088] As one embodiment, the process of iteratively training the fusion model to be trained based on the supervised learning data to obtain a power generation prediction model includes: performing cluster analysis on the supervised learning data to obtain multiple clusters, dividing each cluster into a training set and a validation set, and iteratively training the fusion model to be trained based on the training set and the validation set to obtain a power generation prediction model.
[0089] In the aforementioned power generation prediction method, the following steps are taken: First, predictable power generation data from a hydropower station is acquired. Then, dynamic causal feature extraction is performed on this data to obtain causal features. These causal features are then fused with the predicted power generation data to obtain fused data. Next, time-series feature extraction is performed on the fused data to obtain time-series features. Finally, based on these time-series features, the power generation of the hydropower station is predicted to obtain the target power generation. In this way, by monitoring the predicted power generation data during the operation of the hydropower station, causal features are dynamically captured, and key factors significantly affecting power generation are automatically focused on. This allows for a more accurate reflection of the actual impact of different factors on power generation at different times. Furthermore, by extracting time-series features from the fused data obtained by fusing causal features and the predicted power generation data, the accuracy of power generation prediction for hydropower stations is improved.
[0090] As a detailed embodiment, the following steps are taken: First, predictable power generation data of a hydropower station is acquired. Second, the amount of data change corresponding to the predictable power generation data is determined. Third, causal relationship change information of the predictable power generation data is determined based on the amount of data change. Fourth, attention weight information is constructed based on the causal relationship change information. Fifth, dynamic causal feature extraction is performed on the predictable power generation data based on the attention weight information to obtain causal features. The causal features are then fused with the predictable power generation data to obtain fused data. Sixth, the feature importance information of each feature in the fused data at the current time step is determined. Seventh, the feature importance information of each feature at the current time step is normalized and fused to obtain the feature weight coefficient of the fused data at the current time step. Sixth, the fused data is weighted based on the feature weight coefficient to obtain weighted data. Seventh, short-term fluctuation features are extracted from the weighted data to obtain a first feature. Eighth, medium-term trend features are extracted from the first feature and the weighted data to obtain a second feature. Eighth, periodic features are extracted from the second feature and the weighted data to obtain a time-series feature. Finally, the power generation of the hydropower station is predicted based on the time-series feature to obtain a target power generation.
[0091] Thus, by acquiring the power generation data to be predicted from the hydropower station; dynamically extracting causal features from the power generation data to obtain causal features; fusing the causal features with the power generation data to obtain fused data; extracting time-series features from the fused data to obtain time-series features; and predicting the power generation of the hydropower station based on the time-series features to obtain the target power generation. In this way, by monitoring the power generation data to be predicted during the operation of the hydropower station, dynamically capturing causal features, and automatically focusing on key factors that significantly affect power generation, the actual impact of different factors on power generation at different times can be more accurately reflected. Furthermore, by extracting time-series features from the fused data obtained by fusing causal features and the power generation data to be predicted, the accuracy of power generation prediction for the hydropower station is improved.
[0092] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0093] Based on the same inventive concept, this application also provides a power generation prediction device for implementing the power generation prediction method described above. The solution provided by this device is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more power generation prediction device embodiments provided below can be found in the limitations of the power generation prediction method described above, and will not be repeated here.
[0094] In one exemplary embodiment, such as Figure 3 As shown, a power generation prediction device 300 is provided, including: an acquisition module 302, a feature extraction module 304, a feature fusion module 306, and a prediction module 308, wherein:
[0095] The acquisition module 302 is used to acquire the predicted power generation data of the hydropower station.
[0096] The feature extraction module 304 is used to perform dynamic causal feature extraction on the power generation data to be predicted, and obtain causal features.
[0097] The feature fusion module 306 fuses the causal features with the power generation data to be predicted to obtain fused data.
[0098] The feature extraction module 304 is also used to extract time-series features from the fused data to obtain time-series features.
[0099] The prediction module 308 is used to predict the power generation of the hydropower station based on the time series characteristics to obtain the target power generation.
[0100] In one embodiment, the feature extraction module 304 is further configured to determine the amount of data change corresponding to the power generation data to be predicted; determine the causal relationship change information of the power generation data to be predicted based on the amount of data change corresponding to the power generation data to be predicted; construct attention weight information based on the causal relationship change information; and perform dynamic causal feature extraction on the power generation data to be predicted based on the attention weight information to obtain causal features.
[0101] In one embodiment, the feature extraction module 304 is further configured to perform weighted processing on the fused data to obtain weighted data; extract short-term fluctuation features from the weighted data to obtain a first feature; extract medium-term trend features from the first feature and the weighted data to obtain a second feature; and extract periodic features from the second feature and the weighted data to obtain a time-series feature.
[0102] In one embodiment, the feature fusion module 306 is further configured to determine the feature importance information of each feature in the fused data at the current time step; normalize and fuse the feature importance information of each feature at the current time step to obtain the feature weight coefficient of the fused data at the current time step; and perform weighted processing on the fused data according to the feature weight coefficient to obtain weighted data.
[0103] In one embodiment, the process of predicting power generation is executed through a pre-trained power generation prediction model. The power generation prediction device 300 further includes: a training module, used to acquire historical power generation data corresponding to the hydropower station; determine the correlation between each data dimension in the historical power generation data and the power generation, and filter target power generation data from the historical power generation data according to each correlation; convert the target power generation data into supervised learning data; fuse the dynamic causal attention network to be trained and the hidden Markov tree model to be trained to obtain a fused model to be trained; and iteratively train the fused model to be trained according to the supervised learning data to obtain a power generation prediction model, wherein the fused model to be trained is obtained by fusing the dynamic causal attention network to be trained and the hidden Markov tree model to be trained.
[0104] In one embodiment, the training module is further configured to obtain sliding window parameters; and to traverse the target power generation data through the sliding window corresponding to the sliding window parameters to obtain supervised learning data.
[0105] Each module in the aforementioned power generation prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.
[0106] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a power generation prediction method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0107] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0108] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0109] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.
[0110] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0111] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0112] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0113] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for predicting power generation, characterized in that, The method includes: Obtain the predicted power generation data of the hydropower station; Dynamic causal feature extraction is performed on the power generation data to be predicted to obtain causal features, and the causal features are fused with the power generation data to be predicted to obtain fused data; Temporal features are extracted from the fused data to obtain temporal features; Based on the aforementioned time-series characteristics, the power generation of the hydropower station is predicted to obtain the target power generation.
2. The method according to claim 1, characterized in that, The step of extracting dynamic causal features from the power generation data to be predicted to obtain causal features includes: Determine the amount of data change corresponding to the power generation data to be predicted; Based on the amount of data change corresponding to the power generation data to be predicted, determine the causal relationship change information of the power generation data to be predicted; Based on the causal relationship change information, attention weight information is constructed; Based on the attention weight information, dynamic causal features are extracted from the power generation data to be predicted to obtain causal features.
3. The method according to claim 1, characterized in that, The step of extracting time-series features from the fused data to obtain time-series features includes: The fused data is then weighted to obtain weighted data; Short-term fluctuation features are extracted from the weighted data to obtain the first feature; The second feature is obtained by extracting intermediate trend features from the first feature and the weighted data; Periodic features are extracted from the second feature and the weighted data to obtain time-series features.
4. The method according to claim 3, characterized in that, The weighting process of the fused data to obtain weighted data includes: Determine the feature importance information of each feature in the fused data at the current time step; The feature importance information of each feature at the current time step is normalized and fused to obtain the feature weight coefficient of the fused data at the current time step; The fused data is weighted according to the feature weight coefficients to obtain weighted data.
5. The method according to claim 1, characterized in that, The process of predicting power generation is performed through a pre-trained power generation prediction model, the training process of which includes: Obtain historical power generation data corresponding to the hydropower station; Determine the correlation between each data dimension in the historical power generation data and the power generation, and filter target power generation data from the historical power generation data based on each correlation. The target power generation data is converted into supervised learning data; The dynamic causal attention network to be trained and the hidden Markov tree model to be trained are fused to obtain the fused model to be trained. Based on the supervised learning data, the fusion model to be trained is iteratively trained to obtain a power generation prediction model, wherein the fusion model to be trained is obtained by fusing a dynamic causal attention network to be trained and a hidden Markov tree model to be trained.
6. The method according to claim 5, characterized in that, The process of converting the target power generation data into supervised learning data includes: Get the parameters of the sliding window; The target power generation data is traversed through the sliding window corresponding to the sliding window parameters to obtain supervised learning data.
7. A power generation prediction device, characterized in that, The device includes: The acquisition module is used to acquire the predicted power generation data of the hydropower station; The feature extraction module is used to perform dynamic causal feature extraction on the power generation data to be predicted, and obtain causal features; The feature fusion module fuses the causal features with the power generation data to be predicted to obtain fused data; The feature extraction module is also used to extract temporal features from the fused data to obtain temporal features; The prediction module is used to predict the power generation of the hydropower station based on the time-series characteristics, and obtain the target power generation.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.