Power dispatching technology development prediction method and device, computer equipment and program product

Abstract

The invention relates to a power dispatching technology development prediction method and device, computer equipment and a computer program product. The method comprises the following steps: acquiring source domain data and target domain data in the technical field of power dispatching, and extracting technical development time sequence data and technical semantic data from the source domain data; constructing a collaborative migration model; extracting a time sequence feature vector and a semantic feature vector; performing fusion to obtain fusion features of the source domain data; extracting target time sequence data, and calculating a similarity matrix between the target time sequence data and the fusion features; determining a mapping relation and an alignment feature between the source domain data and the target domain data; aligning the development trend of the target domain data with the development trend of the technical development characteristics of the source domain data, and calculating to obtain integration characteristics of the target domain data; and inputting into the trained prediction network, and predicting to obtain a development prediction index in the technical field of power dispatching. By adopting the method, the prediction accuracy can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting the development of power dispatching technology. Background Technology

[0002] Power dispatching technology forecasting is a technical method that predicts future power dispatching strategies and technical parameters based on historical dispatching data and relevant influencing factors. By analyzing the temporal variation patterns, operational mode characteristics, and dispatching periodic patterns of dispatching decisions, and combining external variables such as power grid topology, equipment operating status, and load distribution changes, mathematical models are constructed to predict the optimal dispatching scheme and key dispatching parameters within a specific time period.

[0003] In related technologies, traditional power dispatch technology prediction mainly adopts the transfer learning method, which transfers the development model of a certain technology field to the power dispatch field in one go. However, it ignores the differences in the characteristics of technological evolution at different levels and the difficulty of transfer, which leads to a decline in the predictive performance of the model for emerging dispatch technologies, resulting in a negative transfer effect. This seriously affects the accuracy of power dispatch technology development prediction under the condition of scarce technical data. Summary of the Invention

[0004] Therefore, it is necessary to provide a power dispatching technology development forecasting method, device, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of development forecasting in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for predicting the development of power dispatching technology, including:

[0006] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0007] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0008] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0009] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0010] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0011] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0012] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0013] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0014] In one embodiment, the technological development characteristics include: technological development trajectory characteristics, technological innovation characteristics, and technological field characteristics; based on the mapping relationship and the alignment characteristics, the development trend of the target domain data is aligned with the development trend of the technological development characteristics of the source domain data in descending order of dimensions to calculate the integrated characteristics of the target domain data, including:

[0015] Extract the low-frequency components within a first preset frequency range from the alignment features. Based on the mapping relationship and the low-frequency components, align the source domain data and the target domain data with the technology development trajectory features, and determine the high-frequency adaptation features.

[0016] Extract the intermediate frequency component within the second preset frequency range from the alignment feature; based on the high frequency adaptation feature and the intermediate frequency component, align the source domain data and the target domain data with the technical innovation feature, and determine the intermediate frequency adaptation feature;

[0017] Extract the high-frequency components within the third preset frequency range from the alignment features. Based on the mid-frequency adaptation features and the high-frequency components, align the source domain data and the target domain data using technical field features, and determine the low-frequency adaptation features.

[0018] The high-frequency adaptation features, the mid-frequency adaptation features, and the low-frequency adaptation features are weighted and summed to obtain the integrated features of the target domain data.

[0019] In one embodiment, the step of extracting semantic features from the technical semantic data through the large model module to obtain a technical semantic feature vector includes:

[0020] Extract the technical description text from the technical semantic data, and perform word segmentation, stop word removal, and terminology standardization on the technical description text to obtain the processed technical description text;

[0021] The processed technical description text is encoded, key semantic features are extracted, and a technical semantic feature vector of preset dimensions is output.

[0022] In one embodiment, fusing the temporal feature vector and the technical semantic feature vector to obtain the fused features of the source domain data includes:

[0023] Project the temporal feature vector and the technical semantic feature vector onto a preset feature space, and calculate the cross-attention weight matrix between the temporal feature vector and the technical semantic feature vector;

[0024] Based on the cross-attention weight matrix, the temporal feature vector and the technical semantic feature vector are dynamically weighted and fused to obtain the fused features of the source domain data.

[0025] In one embodiment, the formula for calculating each similarity in the similarity matrix includes:

[0026] ;

[0027] in, To integrate the similarity between the i-th technological development segment and the j-th technological development segment of the target time series data, To fuse the feature vector of the i-th technological development segment, Let be the feature vector of the j-th technological development segment of the target time series data.

[0028] In one embodiment, the development forecast indicators include: the innovation index of intelligent dispatching algorithms in the field of power dispatching technology, the maturity of distributed dispatching technology, and the intelligence level of dispatching systems.

[0029] Secondly, this application also provides a power dispatching technology development prediction device, comprising:

[0030] The acquisition module is used to acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time series data and technology semantic data from the source domain data;

[0031] A building module is used to construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module.

[0032] The extraction module is used to extract the temporal features from the technology development time series data through the LSTM module to obtain a temporal feature vector, and to extract the semantic features from the technology semantic data through the large model module to obtain a technology semantic feature vector.

[0033] The fusion module is used to fuse the temporal feature vector and the technical semantic feature vector to obtain the fused features of the source domain data;

[0034] The calculation module is used to extract target time-series data from the target domain data and calculate the similarity matrix between the target time-series data and the fused features;

[0035] The determination module is used to determine the mapping relationship and alignment features between the source domain data and the target domain data based on the similarity matrix;

[0036] The calculation module is used to align the development trend of the target domain data with the development trend of the technological development characteristics of the source domain data in order from high to low dimensions based on the mapping relationship and the alignment features, and calculate the integrated features of the target domain data.

[0037] The prediction module is used to input the integrated features into the trained prediction network to predict the development indicators in the field of power dispatching technology.

[0038] 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:

[0039] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0040] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0041] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0042] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0043] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0044] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0045] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0046] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0047] 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:

[0048] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0049] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0050] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0051] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0052] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0053] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0054] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0055] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0056] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0057] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0058] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0059] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0060] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0061] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0062] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0063] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0064] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0065] The aforementioned method, device, computer equipment, computer-readable storage medium, and computer program product for predicting the development of power dispatching technology first acquire source domain data and target domain data in the field of power dispatching technology. From the source domain data, time-series data and technical semantic data of technological development are extracted. A collaborative transfer model is constructed. Time-series features in the time-series data of technological development are extracted using an LSTM module to obtain time-series feature vectors. Semantic features in the technical semantic data are extracted using a large model module to obtain technical semantic feature vectors. The time-series feature vectors and technical semantic feature vectors are fused to obtain fused features of the source domain data. Target time-series data is extracted from the target domain data, and a similarity matrix between the target time-series data and the fused features is calculated. Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined. Based on the mapping relationship and alignment features, the development trend of the target domain data is aligned with the development trend of the technical development features of the source domain data in descending order of dimensions to calculate the integrated features of the target domain data. The integrated features are input into the trained prediction network to predict the development forecast indicators in the field of power dispatching technology. In this way, by capturing the temporal dependencies and innovation cycle patterns of technological development through the LSTM module in the collaborative transfer model, and then learning the semantic similarities and transfer patterns between different technological fields through the large model module, and finally performing dynamic weighted fusion through the attention mechanism, the prediction model can give full play to the accuracy of the temporal modeling of technological evolution and the flexibility of the understanding of technological semantics, thus significantly improving the accuracy of power dispatch technology development prediction. Attached Figure Description

[0066] 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.

[0067] Figure 1 This is an application environment diagram of the power dispatching technology development prediction method in one embodiment;

[0068] Figure 2 This is a flowchart illustrating a power dispatching technology development prediction method in one embodiment;

[0069] Figure 3 This is a structural block diagram of a power dispatching technology development prediction device in one embodiment;

[0070] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0071] 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.

[0072] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0073] The power dispatching technology development prediction method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. 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.

[0074] In one exemplary embodiment, such as Figure 2 As shown, a method for predicting the development of power dispatching technology is provided, and this method is applied to... Figure 1 Taking terminal 102 as an example, the explanation includes the following steps 202 to 216. Wherein:

[0075] Step 202: Obtain source domain data and target domain data in the field of power dispatching technology, and extract technology development time series data and technology semantic data from the source domain data.

[0076] Among them, source domain data consists of historical development data from different fields related to the target domain (power dispatching technology field), such as development data from fields like artificial intelligence and the Internet of Things, specifically the development trajectory of artificial intelligence and the evolution mode of Internet of Things technology; target domain data consists of development data directly related to power dispatching technology, such as the development of intelligent dispatching algorithms and the evolution of distributed dispatching technology; technology development time series data consists of data that can reflect the changes and development trends of technology development over time, such as the time points of technological breakthroughs and the time points of technological maturity; and technology semantic data consists of semantic information data that can describe the specific content and characteristics of technology development, usually existing in the form of text, such as technical patent abstracts, academic paper descriptions, and technical reports.

[0077] For example, historical development data related to the target domain (power dispatching technology field) but from different fields and development data directly related to power dispatching technology are acquired. From the source domain data, data that can reflect the changes and development trends of technology development over time and semantic information data that can describe the specific content and characteristics of technology development are extracted.

[0078] Step 204: Construct a collaborative transfer model, which includes an LSTM module and a large model module.

[0079] Optionally, the collaborative transfer model includes an LSTM (Long Short-Term Memory) module and a large model module. The LSTM module is constructed using a multi-layer LSTM network architecture, including an input layer, three LSTM hidden layers, and a temporal feature output layer.

[0080] In one embodiment, the first LSTM hidden layer has 128 hidden units to capture short-term technology change patterns, the second LSTM hidden layer has 256 hidden units to model medium-term technology development cycles, and the third LSTM hidden layer has 512 hidden units to learn long-term technology evolution trends. A dropout layer is added after each LSTM hidden layer to prevent overfitting, and finally, the time series features are output through a time series feature output layer (fully connected layer).

[0081] In another embodiment, a large-scale model semantic understanding module is constructed. This module is adapted to the technical field based on a pre-trained large language model. It includes a technical text preprocessing module, a large-scale model encoder, a domain adaptation layer, and a semantic feature extraction layer. The technical text preprocessing module is responsible for cleaning and standardizing technical description text. The large-scale model encoder uses a Transformer architecture to perform deep semantic encoding on the technical text. The domain adaptation layer adapts the general semantic representation to the technology development prediction task through a fine-tuning mechanism. The semantic feature extraction layer extracts key semantic features from the encoding results through an attention pooling mechanism.

[0082] Step 206: Extract the temporal features from the time series data of technological development using the LSTM module to obtain the temporal feature vector, and extract the semantic features from the semantic data of technology using the large model module to obtain the semantic feature vector of technology.

[0083] For example, the collaborative transfer model splits the input source domain data. The LSTM module accepts quantified technology development time-series data, which includes monthly sequences of scheduling algorithm patent application data, quarterly sequences of intelligent scheduling system deployment numbers, and annual sequences of scheduling accuracy improvement indicators. After standardization, the above data is input into the LSTM module. Through gating mechanisms and memory units, the time-series dependencies of technology development are captured, and time-series patterns such as the S-shaped curve characteristics of technology maturity evolution, the periodic fluctuation pattern of R&D investment, and the phased leap patterns of industrialization are identified. Finally, a time-series feature vector containing time-series evolution information is output.

[0084] Optionally, the large model module receives technical semantic data, including unstructured text information such as text summaries of scheduling technology development reports, descriptions of technical patents, and technical solutions described in academic papers. The large model first performs semantic encoding on the above text, learns the commonalities of technical principles such as the commonalities of the mathematical foundations of optimization algorithms and the architectural design concepts of distributed systems, identifies the correlational characteristics of application scenarios such as the technical integration mode of smart grids and smart cities, and the technical synergy of new energy access and energy storage scheduling, captures the semantic patterns of development driving factors such as the impact of policy guidance on technical routes and the traction of market demand on innovation directions, and outputs a technical semantic feature vector containing rich semantic associations.

[0085] Step 208: The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data.

[0086] Optionally, the feature vectors of the two modules are then processed collaboratively through a cross-attention mechanism to calculate the correlation strength between each technological development stage in the temporal feature vector and each technological concept in the semantic feature vector. This identifies which technological semantic concepts are more important in a specific temporal stage. For example, the importance weight of the concept of "algorithm innovation" will increase significantly during the technological breakthrough period, and the weight of the concept of "system integration" will increase during the industrialization stage. Through this dynamic weight adjustment mechanism, the temporal evolution law and semantic understanding information are organically integrated, ultimately generating a fused feature that contains both accurate temporal development patterns and rich technological semantic associations.

[0087] Step 210: Extract target time series data from target domain data and calculate the similarity matrix between target time series data and fused features.

[0088] For example, target time series data is extracted from target domain data, and a similarity matrix between the target time series data and the fused features is calculated.

[0089] Step 212: Based on the similarity matrix, determine the mapping relationship and alignment features between the source domain data and the target domain data.

[0090] Optionally, the optimal alignment path is calculated based on the similarity matrix. DTW_align (a time series alignment algorithm used to find the optimal matching path between two unequal time series) is employed. Through dynamic programming, a path from the top left to the bottom right corner of the similarity matrix is ​​found, maximizing the sum of similarity scores (or minimizing the sum of distances) of all points on the path. This path is the optimal alignment path, i.e., the mapping relationship between the source domain data and the target domain data. The optimal alignment path is a coordinate sequence, for example, [(i1, j1), (i2, j2), ..., (i...]. k j k )], where each coordinate pair (i, j) represents the matching of the i-th time point in the source domain with the j-th time point in the target domain.

[0091] In one embodiment, the mapping relationship between source domain data and target domain data includes [(1,1), (2,1), (3,2), (4,3), ...], indicating that the first and second time points in the source domain data correspond to the first time point in the target domain data, and the third time point in the source domain data corresponds to the second time point in the target domain data.

[0092] Optionally, the alignment feature is obtained by rearranging and mapping the fused features based on the mapping relationship.

[0093] In one embodiment, the mapping relationship is assumed to be [(i1, j1), (i2, j2), ..., (i k j k The alignment feature is to reorganize the corresponding feature vectors in the fusion feature according to the path. For example, if the path shows that the third time point in the source domain data corresponds to the first time point in the target domain data, then the sub-vector corresponding to the first time point in the alignment feature is the same as the sub-vector corresponding to the third time point in the fusion feature. In the case of one-to-many (multiple source domain data time points correspond to one target domain data time point), it can be processed by weighted averaging or other fusion strategies. This application embodiment does not limit this.

[0094] Step 214: Based on the mapping relationship and alignment features, align the development trend of the target domain data with the development trend of the technological development features of the source domain data in each dimension from high to low, and calculate the integrated features of the target domain data.

[0095] For example, the time-domain signal is converted into a frequency-domain signal by Fourier transform, and low-pass, band-pass, and high-pass filters are used to obtain frequency components in different ranges. Then, the signal is converted back to the time domain by inverse Fourier transform. Based on the mapping relationship and alignment characteristics between the source domain data and the target domain data, the development trend of the target domain data is aligned with the development trend of the technological development characteristics of the source domain data in each dimension from high to low, and the integrated characteristics of the target domain data are calculated.

[0096] Step 216: Input the integrated features into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0097] Among them, the development forecast indicators include the intelligent scheduling algorithm innovation index, the maturity of distributed scheduling technology, and the intelligence level of the scheduling system.

[0098] For example, by inputting integrated features into the trained prediction network, the innovation index of intelligent scheduling algorithms, the maturity of distributed scheduling technology, and the intelligence level of scheduling systems in the field of power dispatching technology can be predicted.

[0099] The aforementioned method for predicting the development of power dispatching technology involves acquiring source domain data and target domain data in the field of power dispatching technology. From the source domain data, time-series data and technical semantic data of technological development are extracted. A collaborative transfer model is constructed. A time-series feature vector is obtained by extracting time-series features from the time-series data of technological development using an LSTM module, and a technical semantic feature vector is obtained by extracting semantic features from the technical semantic data using a large model module. The time-series feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data. Target time-series data is extracted from the target domain data, and a similarity matrix between the target time-series data and the fused features is calculated. Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined. Based on the mapping relationship and alignment features, the development trend of the target domain data is aligned with the development trend of the technical development features of the source domain data from high to low dimensions, and the integrated features of the target domain data are calculated. The integrated features are input into the trained prediction network to predict the development forecast indicators in the field of power dispatching technology. In this way, by capturing the temporal dependencies and innovation cycle patterns of technological development through the LSTM module in the collaborative transfer model, and then learning the semantic similarities and transfer patterns between different technological fields through the large model module, and finally performing dynamic weighted fusion through the attention mechanism, the prediction model can give full play to the accuracy of the temporal modeling of technological evolution and the flexibility of the understanding of technological semantics, thus significantly improving the accuracy of power dispatch technology development prediction.

[0100] In an exemplary embodiment, the technological development characteristics include: technological development trajectory characteristics, technological innovation characteristics, and technological field characteristics. Based on the mapping relationship and alignment characteristics, the development trend of the target domain data is aligned with the development trend of the technological development characteristics of the source domain data in sequence from high to low dimensions to calculate the integrated characteristics of the target domain data. This includes: extracting low-frequency components within a first preset frequency range from the alignment characteristics; aligning the source domain data and target domain data with technological development trajectory characteristics based on the mapping relationship and the low-frequency components, and determining high-frequency adaptation characteristics; extracting mid-frequency components within a second preset frequency range from the alignment characteristics; aligning the source domain data and target domain data with technological innovation characteristics based on the high-frequency adaptation characteristics and the mid-frequency components, and determining mid-frequency adaptation characteristics; extracting high-frequency components within a third preset frequency range from the alignment characteristics; aligning the source domain data and target domain data with technological field characteristics based on the mid-frequency adaptation characteristics and the high-frequency components, and determining low-frequency adaptation characteristics; and weighted summing the high-frequency adaptation characteristics, mid-frequency adaptation characteristics, and low-frequency adaptation characteristics to obtain the integrated characteristics of the target domain data.

[0101] In practice, a low-pass filter is used to filter out high-frequency noise and short-term fluctuations, while retaining low-frequency trend information, i.e., low-frequency components, that reflect the overall trend of technological development. The source domain data and target domain data are aligned with the characteristics of technological development trajectory to determine high-frequency adaptation features.

[0102] One embodiment focuses on the long-term development trend and innovation cycle pattern of dispatching technology, reflected through historical quantitative data on the development of power dispatching technology. This includes annual time series of dispatching algorithm patent applications, quarterly statistics of installed capacity of intelligent dispatching systems, historical records of dispatching accuracy improvement indicators, time series data of the number of dispatching technology standards released, and annual changes in the scale of investment in dispatching systems—all macro-level development indicators. The decomposition process employs a technology evolution sequence decomposition technique. First, the original dispatching technology development data is preprocessed and standardized. Then, the empirical mode decomposition method is used to decompose the technology development time series into three components: trend component, periodic component, and noise component. The trend component reflects the long-term evolution of dispatching technology from traditional manual dispatching to intelligent automatic dispatching; the periodic component reflects the regular changes in technological innovation, such as the generational upgrade cycle of dispatching technology every 5-7 years; and the noise component represents short-term random fluctuations. Next, wavelet transform technology is used to further extract frequency domain features at different time scales, identifying multi-level periodic patterns in technology development. Finally, high-frequency adaptation features are obtained, including the slope parameter of the technology maturity curve, the time point of industrial development stage transition, and the frequency characteristics of the innovation cycle.

[0103] The necessity and feasibility of alignment are demonstrated by comparing macro-level characteristics such as the shape parameters of technology maturity curves, the standards for classifying industrial development stages, and the length of technological innovation cycles across different technological fields. The specific steps for adjusting domain alignment include: first, calculating the probability distributions of the technological development trajectories of the source and target domain data; constructing the probability density function of technological development using kernel density estimation; then, using the Wasserstein distance to measure the degree of difference between the two technological development distributions; next, finding the optimal mapping function that minimizes the distribution distance using optimal transmission theory; solving for the optimal transmission plan using the Sinkhorn algorithm; and finally, optimizing the mapping parameters using gradient descent and updating the weights of the domain adaptation network through backpropagation. This ensures that the macro-level technological characteristics of the source domain are accurately mapped to the corresponding feature space of the target domain, achieving cross-domain technological innovation feature alignment.

[0104] In one embodiment, a bandpass filter is used to extract signals in the mid-frequency range to reflect the periodic fluctuations and phased characteristics of technological development. The focus is on the selection of specific technical routes and the adaptation of key technological breakthroughs. The frequency domain characteristics of technological innovation are extracted through spectrum analysis of technological development, including meso-level patterns such as R&D cycle, technology iteration frequency, and application promotion rhythm. The periodic probability distribution alignment technology of technological development is used to ensure the effective migration of technological path patterns.

[0105] The adaptation of specific technical routes and key technological breakthroughs is reflected through meso-level technology development data, including the time series of R&D investment for different scheduling technical routes, the occurrence time records of technological breakthrough events, the trajectory data of key technical indicator improvements, and the promotion process of technology applications from the laboratory to engineering. The feature extraction process first obtains the development history of each technical route from the technology path evolution database, including input-output data for different routes such as centralized scheduling, distributed scheduling, and intelligent scheduling. Then, the technology development spectrum analysis method is used to perform Fourier transform on these time-series data, converting the time-domain signal into a frequency-domain representation, identifying the main frequency components of technological innovation, and extracting the low-frequency components corresponding to the R&D cycle, the mid-frequency components corresponding to the technology iteration frequency, and the high-frequency components corresponding to the market promotion rhythm. Next, wavelet packet decomposition is used to further refine the frequency-domain features, separating the technology development patterns at different time scales, and finally obtaining the mid-frequency adaptation features that describe the rules of technical route selection and the occurrence patterns of breakthroughs.

[0106] The criteria for effective transfer learning are that the correlation between the source domain data technology path and the target domain data technology path in the frequency domain exceeds a threshold of 0.7, and the prediction accuracy after transfer is improved by more than 15% compared to the baseline method. The specific steps of the transfer learning include: first, analyzing the cyclical patterns of technological development in the source and target domain data, calculating the frequency statistics of technological breakthrough events occurring in different time intervals; then, obtaining the cyclical probability distribution of technological development by fitting historical data, determining the probability distribution parameters using the maximum likelihood estimation method; next, aligning the probability distribution using the KL divergence minimization method to ensure consistency between the cyclical patterns of the source and target domains; and finally, applying the aligned cyclical knowledge to the prediction of the target domain using a transfer learning algorithm. Here, the cyclical probability of technological development refers to the probability density function of technological innovation events occurring within a specific time period.

[0107] In one embodiment, a high-frequency signal is extracted using a high-pass filter to reflect rapidly changing details such as short-term fluctuations in technological development, the impact of sudden events, and rapid market responses. This process handles the refined characteristics of specific technical parameters and implementation details, including micro-level aspects such as algorithm optimization details, system integration schemes, and engineering application parameters.

[0108] In one embodiment, the process of refining the technical parameters and implementation details includes the following steps: First, collect micro-level technical implementation data, including performance parameters such as the computational complexity, convergence time, and memory usage of the scheduling algorithm; technical details such as the interface specifications, data formats, and communication protocols of system integration; and implementation parameters such as the hardware configuration, software version, and deployment environment of engineering applications. Then, standardize these multi-dimensional heterogeneous technical parameters by using Z-score standardization to eliminate dimensional differences. Next, perform feature engineering to extract key features such as algorithm efficiency indicators, system compatibility scores, and implementation complexity levels. Finally, perform nonlinear feature transformation and fusion through a deep neural network, and use an autoencoder to learn the low-dimensional representation of the technical parameters to obtain a refined feature vector containing rich micro-level technical implementation information. These features can accurately describe the specific details and engineering difficulties of the technical implementation.

[0109] To achieve fine-grained technical feature alignment, an adversarial technology adaptation mechanism is adopted. The specific steps include: First, a technology discriminator network is constructed. This network is a multilayer perceptron structure that can determine the source domain of the input technical features. Then, a feature extractor network is designed with the goal of generating domain-independent technical feature representations. Next, an adversarial training process is carried out. The feature extractor strives to generate features that can "deceive" the discriminator, making it unable to accurately identify the source domain of the features. Meanwhile, the discriminator strives to improve the accuracy of domain classification. Through this game process, the two networks promote each other's optimization. When a Nash equilibrium is reached, the features generated by the feature extractor have good alignment effects across different technology fields. Finally, a gradient inversion layer technique is used to achieve stable optimization of the adversarial training, ensuring the effective mapping and alignment of fine-grained technical features across domains, i.e., technology field feature alignment.

[0110] In practice, the high-frequency adaptation features, mid-frequency adaptation features, and low-frequency adaptation features are weighted and summed to obtain the integrated features of the target domain data.

[0111] In the above embodiments, by gradually aligning the source domain data and the target domain data from components in three different frequency ranges, the fine-grained adaptation of the transition technology features is achieved, effectively avoiding the negative migration effect that may be caused by traditional overall technology migration.

[0112] In an exemplary embodiment, semantic features are extracted from technical semantic data through a large model module to obtain a technical semantic feature vector, including: extracting technical description text from the technical semantic data; performing word segmentation, stop word removal, and terminology standardization on the technical description text to obtain processed technical description text; encoding the processed technical description text; extracting key semantic features; and outputting a technical semantic feature vector of a preset dimension.

[0113] In practice, the technical description text is extracted from the technical semantic data, and then processed by word segmentation, stop word removal, and terminology standardization to obtain the processed technical description text. The processed technical description text is then encoded, key semantic features are extracted, and a technical semantic feature vector of a preset dimension is output.

[0114] In the above embodiments, extracting semantic feature vectors through large model modules can not only improve the generalization ability and prediction accuracy of the model, but also enhance the effect of cross-domain transfer learning.

[0115] In an exemplary embodiment, the temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data, including: projecting the temporal feature vector and the technical semantic feature vector onto a preset feature space, calculating the cross-attention weight matrix between the temporal feature vector and the technical semantic feature vector; and dynamically weighting and fusing the temporal feature vector and the technical semantic feature vector based on the cross-attention weight matrix to obtain the fused features of the source domain data.

[0116] In practice, the feature vectors of the two modules are processed collaboratively through a cross-attention mechanism during the fusion phase. The system calculates the correlation strength between each technological development stage in the temporal feature vector and each technological concept in the semantic feature vector, identifying which technological semantic concepts are more important in a specific temporal stage. For example, the importance weight of the concept of "algorithm innovation" will increase significantly during the technological breakthrough period, and the weight of the concept of "system integration" will increase during the industrialization stage. Through this dynamic weight adjustment mechanism, the temporal evolution law and semantic understanding information are organically integrated, ultimately generating a fused feature that contains both accurate temporal development patterns and rich technological semantic associations.

[0117] In the above embodiments, the two features are fully fused through an attention mechanism to obtain fused features, which provide a high-quality feature foundation for subsequent technology development prediction.

[0118] In an exemplary embodiment, the formula for calculating each similarity in the similarity matrix is ​​shown in (1).

[0119] (1)

[0120] in, To integrate the similarity between the i-th technological development segment and the j-th technological development segment of the target time series data, To fuse the feature vector of the i-th technological development segment, Let be the feature vector of the j-th technological development segment of the target time series data.

[0121] In the above embodiments, by calculating the similarity matrix, the similarities between different technical fields are identified, cross-domain alignment is achieved, and the accuracy of transfer learning is improved.

[0122] In one exemplary embodiment, the development forecast indicators include: the innovation index of intelligent dispatching algorithms in the field of power dispatching technology, the maturity of distributed dispatching technology, and the intelligence level of dispatching systems.

[0123] In practice, development forecast indicators include: the innovation index of intelligent dispatching algorithms in the field of power dispatching technology, the maturity of distributed dispatching technology, and the level of intelligence of dispatching systems.

[0124] The innovation index of intelligent scheduling algorithm, the maturity of distributed scheduling technology, and the intelligence level of scheduling system can be set by the user according to different specific circumstances, and this application embodiment does not limit this.

[0125] For example, the innovation index of intelligent scheduling algorithms can be [8.2, 8.7, 9.1, 9.4....], the maturity of distributed scheduling technology can be [6.8, 7.2, 7.6, 8.0....], and the intelligence level of scheduling systems can be [7.5, 7.9, 8.3, 8.7....].

[0126] In the above embodiments, by predicting the development forecast indicators in the field of power dispatch, the development level and trend of power dispatch technology can be comprehensively evaluated, providing an important reference for the planning and development of power dispatch technology.

[0127] To illustrate the power dispatching technology development prediction method in this application in detail, an embodiment is provided below. For example, this application describes a power dispatching technology development prediction method in a specific scenario.

[0128] First, acquire historical development data related to the target domain (power dispatching technology field) but from different fields, as well as development data directly related to power dispatching technology. From the source domain data, extract data that can reflect the changes and development trends of technology development over time, and semantic information data that can describe the specific content and characteristics of technology development.

[0129] The collaborative transfer model includes an LSTM (Long Short-Term Memory) module and a large model module. The LSTM module is constructed using a multi-layer LSTM network architecture, which includes an input layer, three LSTM hidden layers, and a temporal feature output layer.

[0130] A large-scale semantic understanding module is constructed, which is based on a pre-trained large language model for technical field adaptation. It includes a technical text preprocessing module, a large-scale encoder, a domain adaptation layer, and a semantic feature extraction layer. The technical text preprocessing module is responsible for cleaning and standardizing technical description texts. The large-scale encoder uses the Transformer architecture to perform deep semantic encoding on technical texts. The domain adaptation layer adapts the general semantic representation to the technology development prediction task through a fine-tuning mechanism. The semantic feature extraction layer extracts key semantic features from the encoding results through an attention pooling mechanism.

[0131] The LSTM module extracts the temporal features from the time series data of technological development to obtain the temporal feature vector, and the large model module extracts the semantic features from the semantic data of technology to obtain the semantic feature vector of technology.

[0132] The feature vectors of the two modules are processed collaboratively through a cross-attention mechanism to calculate the correlation strength between each technological development stage in the temporal feature vector and each technological concept in the semantic feature vector. This identifies which technological semantic concepts are more important in a specific temporal stage. For example, the importance weight of the concept of "algorithm innovation" will increase significantly during the technological breakthrough period, and the weight of the concept of "system integration" will increase during the industrialization stage. Through this dynamic weight adjustment mechanism, the temporal evolution law and semantic understanding information are organically integrated, ultimately generating a fused feature that contains both accurate temporal development patterns and rich technological semantic associations.

[0133] Extract target time series data from the target domain data and calculate the similarity matrix between the target time series data and the fused features.

[0134] The optimal alignment path is calculated based on the similarity matrix. DTW_align (a time series alignment algorithm used to find the optimal matching path between two unequal time series) is employed. Through dynamic programming, a path from the top left corner to the bottom right corner is found on the similarity matrix such that the sum of the similarity scores (or the sum of the distances) of all points on the path is maximized (or minimized). This path is the optimal alignment path, i.e., the mapping relationship between the source domain data and the target domain data. The alignment feature is obtained by rearranging and mapping the fused features based on the mapping relationship.

[0135] The time-domain signal is converted into a frequency-domain signal by Fourier transform. Low-pass, band-pass, and high-pass filters are used to obtain frequency components in different ranges. Then, the signal is converted back to the time domain by inverse Fourier transform. Based on the mapping relationship and alignment characteristics between the source domain data and the target domain data, the development trend of the target domain data is aligned with the development trend of the technological development characteristics of the source domain data in each dimension from high to low, and the integrated characteristics of the target domain data are calculated.

[0136] By integrating the features and inputting them into the trained prediction network, the innovation index of intelligent dispatching algorithms, the maturity of distributed dispatching technology, and the intelligence level of dispatching systems in the field of power dispatching technology are predicted.

[0137] It should be understood that although the steps in the flowcharts of the embodiments described above 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 embodiments described above 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 in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0138] Based on the same inventive concept, this application also provides a power dispatching technology development forecasting device for implementing the power dispatching technology development forecasting method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more power dispatching technology development forecasting device embodiments provided below can be found in the limitations of the power dispatching technology development forecasting method described above, and will not be repeated here.

[0139] In one exemplary embodiment, such as Figure 3 As shown, a power dispatching technology development prediction device is provided, comprising: an acquisition module 301, a construction module 302, an extraction module 303, a fusion module 304, a calculation module 305, a determination module 306, and a prediction module 307, wherein:

[0140] The acquisition module is used to acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data.

[0141] A building module is used to construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module.

[0142] The extraction module is used to extract the temporal features from the time-series data of technological development through the LSTM module to obtain a temporal feature vector, and to extract the semantic features from the semantic data of technology through the large model module to obtain a semantic feature vector of technology.

[0143] The fusion module is used to fuse the temporal feature vector and the technical semantic feature vector to obtain the fused features of the source domain data.

[0144] The calculation module is used to extract target time-series data from the target domain data and calculate the similarity matrix between the target time-series data and the fused features.

[0145] The determination module is used to determine the mapping relationship and alignment features between the source domain data and the target domain data based on the similarity matrix.

[0146] The calculation module is used to align the development trend of the target domain data with the development trend of the technological development characteristics of the source domain data in sequence from high to low dimensions based on the mapping relationship and the alignment features, and calculate the integrated features of the target domain data.

[0147] The prediction module is used to input the integrated features into the trained prediction network to predict the development indicators in the field of power dispatching technology.

[0148] In one exemplary embodiment, the above-mentioned calculation module is further configured to extract low-frequency components within a first preset frequency range from the alignment features, and based on the mapping relationship and the low-frequency components, align the source domain data and the target domain data with technology development trajectory features, and determine high-frequency adaptation features.

[0149] Extract the intermediate frequency component within the second preset frequency range from the alignment feature; based on the high frequency adaptation feature and the intermediate frequency component, align the source domain data and the target domain data with the technical innovation feature, and determine the intermediate frequency adaptation feature;

[0150] Extract the high-frequency components within the third preset frequency range from the alignment features. Based on the mid-frequency adaptation features and the high-frequency components, align the source domain data and the target domain data using technical field features, and determine the low-frequency adaptation features.

[0151] The high-frequency adaptation features, the mid-frequency adaptation features, and the low-frequency adaptation features are weighted and summed to obtain the integrated features of the target domain data.

[0152] In one exemplary embodiment, the extraction module is further configured to extract the technical description text from the technical semantic data, and perform word segmentation, stop word removal and terminology standardization on the technical description text to obtain the processed technical description text.

[0153] The processed technical description text is encoded, key semantic features are extracted, and a technical semantic feature vector of preset dimensions is output.

[0154] In one exemplary embodiment, the fusion module is further configured to project the temporal feature vector and the technical semantic feature vector onto a preset feature space, and calculate the cross-attention weight matrix between the temporal feature vector and the technical semantic feature vector;

[0155] Based on the cross-attention weight matrix, the temporal feature vector and the technical semantic feature vector are dynamically weighted and fused to obtain the fused features of the source domain data.

[0156] In one exemplary embodiment, the above-described calculation module is further configured to calculate each similarity in the similarity matrix:

[0157] ;

[0158] in, To integrate the similarity between the i-th technological development segment and the j-th technological development segment of the target time series data, To fuse the feature vector of the i-th technological development segment, Let be the feature vector of the j-th technological development segment of the target time series data.

[0159] In one exemplary embodiment, the development forecast indicators include: the innovation index of intelligent dispatching algorithms in the field of power dispatching technology, the maturity of distributed dispatching technology, and the intelligence level of dispatching systems.

[0160] Each module in the aforementioned power dispatching technology development and forecasting 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.

[0161] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, this computer device includes a processor, memory, input / output interfaces (I / O), a communication interface, a display unit, and input devices. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface, display unit, and input devices are also connected to the system bus via the I / O 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, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores source domain data and target domain data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a power dispatching technology development prediction method.

[0162] The display unit of this computer device 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 this computer device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad set on the casing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0163] 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.

[0164] 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 perform the following steps:

[0165] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0166] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0167] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0168] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0169] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0170] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0171] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0172] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0173] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0174] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0175] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0176] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0177] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0178] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0179] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0180] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0181] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0182] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0183] Acquire source domain data and target domain data in the field of power dispatching technology, and extract technology development time-series data and technology semantic data from the source domain data;

[0184] Construct a collaborative transfer model; the collaborative transfer model includes an LSTM module and a large model module;

[0185] The LSTM module extracts the temporal features from the technology development time series data to obtain a temporal feature vector, and the large model module extracts the semantic features from the technology semantic data to obtain a technology semantic feature vector.

[0186] The temporal feature vector and the technical semantic feature vector are fused to obtain the fused features of the source domain data;

[0187] Extract target time-series data from the target domain data, and calculate the similarity matrix between the target time-series data and the fused features;

[0188] Based on the similarity matrix, the mapping relationship and alignment features between the source domain data and the target domain data are determined;

[0189] Based on the mapping relationship and the alignment features, the development trend of the target domain data is aligned with the development trend of the technological development features of the source domain data in each dimension from high to low, and the integration features of the target domain data are calculated.

[0190] The integrated features are input into the trained prediction network to predict the development indicators of the power dispatching technology field.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0192] 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, and when executed, it 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.

[0193] 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.

[0194] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent 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 the development of power dispatching technology, characterized by, The method comprises: obtaining source domain data and target domain data in the technical field of power dispatch, and extracting technical development time sequence data and technical semantic data from the source domain data; constructing a collaborative migration model; the collaborative migration model comprises an LSTM module and a large model module; extracting time sequence features in the technical development time sequence data through the LSTM module to obtain a time sequence feature vector, and extracting semantic features in the technical semantic data through the large model module to obtain a technical semantic feature vector; fusing the time sequence feature vector and the technical semantic feature vector to obtain fused features of the source domain data; extracting target time sequence data from the target domain data, and calculating a similarity matrix between the target time sequence data and the fused features; determining a mapping relationship and an alignment feature between the source domain data and the target domain data based on the similarity matrix; based on the mapping relationship and the alignment feature, aligning a development trend of the target domain data with a development trend of technical development features of the source domain data in each dimension from high to low in turn, and calculating integrated features of the target domain data; inputting the integrated features into a trained prediction network to predict development prediction indexes in the technical field of power dispatch.

2. The method of claim 1, wherein, The technical development features comprise technical development trajectory features, technical innovation features, and technical field features; based on the mapping relationship and the alignment feature, aligning the development trend of the target domain data with the development trend of the technical development features of the source domain data in each dimension from high to low in turn, and calculating the integrated features of the target domain data, comprising: extracting low-frequency components in a first preset frequency range in the alignment feature, aligning technical development trajectory features of the source domain data and the target domain data based on the mapping relationship and the low-frequency components, and determining high-frequency adaptation features; extracting medium-frequency components in a second preset frequency range in the alignment feature, aligning technical innovation features of the source domain data and the target domain data based on the high-frequency adaptation features and the medium-frequency components, and determining medium-frequency adaptation features; extracting high-frequency components in a third preset frequency range in the alignment feature, aligning technical field features of the source domain data and the target domain data based on the medium-frequency adaptation features and the high-frequency components, and determining low-frequency adaptation features; weighting and summing the high-frequency adaptation features, the medium-frequency adaptation features, and the low-frequency adaptation features to obtain the integrated features of the target domain data.

3. The method of claim 1, wherein, The extraction of the semantic features in the technical semantic data through the large model module to obtain the technical semantic feature vector comprises: extracting technical description texts in the technical semantic data, performing word segmentation, stop word removal, and term standardization processing on the technical description texts to obtain processed technical description texts; encoding the processed technical description texts, extracting key semantic features, and outputting technical semantic feature vectors in a preset dimension.

4. The method of claim 1, wherein, The fusion of the time sequence feature vector and the technical semantic feature vector obtains a fusion feature of the source domain data. The time sequence feature vector and the technical semantic feature vector are projected to a preset feature space, and a cross-attention weight matrix between the time sequence feature vector and the technical semantic feature vector is calculated. Based on the cross-attention weight matrix, the time sequence feature vector and the technical semantic feature vector are dynamically weighted and fused to obtain the fusion feature of the source domain data.

5. The method of claim 1, wherein, The calculation formula of each similarity in the similarity matrix comprises: ； wherein, is a similarity between the i-th technical development segment of the fusion feature and the j-th technical development segment of the target time series data, is a feature vector of the i-th technical development segment of the fusion feature, is a feature vector of the j-th technical development segment of the target time series data.

6. The method of claim 1, wherein, The development prediction index comprises an intelligent scheduling algorithm innovation index, a distributed scheduling technology maturity, and an intelligent level of a scheduling system in the power scheduling technical field.

7. An electric power dispatching technology development prediction device characterized by comprising: The device comprises: An acquisition module is configured to acquire source domain data and target domain data in the power scheduling technical field, and extract technical development time sequence data and technical semantic data from the source domain data; A construction module is configured to construct a collaborative migration model, wherein the collaborative migration model comprises an LSTM module and a large model module; An extraction module is configured to extract time sequence features in the technical development time sequence data by using the LSTM module to obtain a time sequence feature vector, and extract semantic features in the technical semantic data by using the large model module to obtain a technical semantic feature vector; A fusion module is configured to fuse the time sequence feature vector and the technical semantic feature vector to obtain a fusion feature of the source domain data; A calculation module is configured to extract target time sequence data from the target domain data, and calculate a similarity matrix between the target time sequence data and the fusion feature; A determination module is configured to determine a mapping relationship and an alignment feature between the source domain data and the target domain data based on the similarity matrix; The calculation module is configured to align a development trend of the target domain data with a development trend of the technical development feature of the source domain data from high to low in each dimension based on the mapping relationship and the alignment feature, and calculate an integrated feature of the target domain data; A prediction module is configured to input the integrated feature into a trained prediction network to predict a development prediction index in the power scheduling technical field.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method in any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 6.

Images