CN117609769A - Method, system, equipment and medium for generating power grid new equipment starting scheme - Google Patents
Method, system, equipment and medium for generating power grid new equipment starting scheme Download PDFInfo
- Publication number
- CN117609769A CN117609769A CN202410094573.9A CN202410094573A CN117609769A CN 117609769 A CN117609769 A CN 117609769A CN 202410094573 A CN202410094573 A CN 202410094573A CN 117609769 A CN117609769 A CN 117609769A
- Authority
- CN
- China
- Prior art keywords
- power grid
- new equipment
- scheme
- starting
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000004927 fusion Effects 0.000 claims abstract description 33
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims description 36
- 239000013598 vector Substances 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 13
- 230000010354 integration Effects 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000000873 masking effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a method, a system, equipment and a medium for generating a starting scheme of new equipment of a power grid, and relates to the field of new equipment of the power grid; according to the starting range data and the topology information, performing feature extraction by using an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model to obtain target feature data; and generating a new equipment starting generation scheme of the power grid by utilizing a text generation model of the new equipment starting scheme generation model of the power grid according to the target characteristic data. The method and the device can improve the generation speed of the starting scheme of the new equipment of the power grid.
Description
Technical Field
The invention relates to the field of new equipment of a power grid, in particular to a method, a system, equipment and a medium for generating a starting scheme of the new equipment of the power grid.
Background
The start-up of a new plant in the network involves not only the safe commissioning of the plant itself, but also how it works in conjunction with existing power plants and networks. Various factors such as grid stability, control of voltage and frequency, equipment protection, etc. must be fully considered during the start-up of the new equipment. A good starting method not only can ensure the stable operation of the power grid, but also can reduce the influence and risk possibly brought by starting new equipment to the greatest extent. Therefore, the method accurately generates a new equipment starting scheme, and has important significance for safe and stable operation of the power grid.
There are roughly two types of existing start-up schemes in the manner of generation: the first is to manually select all operations, then the system checks whether the scheme is compliant, and then all operations are directly printed out according to the requirements generated by the starting scheme, and the method can be suitable for all types of new equipment to be accessed, but requires extremely high professional literacy of operators; the second is to select the device to be started and then generate a new device start-up scheme by one-touch, which can simply and quickly generate the start-up scheme, but can only generate the start-up scheme of the preset device type. With the rise of new generation artificial intelligence technology represented by deep learning technology, machine learning and artificial intelligence technology are widely used in power systems. These techniques can build a new device start-up scenario model of the grid through analysis of a large amount of historical data. The models can be trained based on the operation range, the starting scheme and the like of the power grid to generate a feasible power grid new equipment starting scheme, the online generation speed of the power grid new equipment starting scheme is improved, and the new equipment starting schemes of various types are generated in a one-key mode only based on simple expert rules of the power grid new equipment starting scheme rather than complex reasoning logic rules, so that how to generate the new equipment starting scheme is important in practical application.
Disclosure of Invention
The invention aims to provide a method, a system, equipment and a medium for generating a power grid new equipment starting scheme, which can improve the generating speed of the power grid new equipment starting scheme.
In order to achieve the above object, the present invention provides the following.
A power grid new equipment starting scheme generation method comprises the following steps: and acquiring starting range data of the target power grid and topology information of the target power grid. And carrying out feature extraction by using an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model according to the starting range data and the topology information to obtain target feature data. And generating a new equipment starting generation scheme of the power grid by utilizing a text generation model of the new equipment starting scheme generation model of the power grid according to the target characteristic data.
Optionally, the expert rule knowledge fusion network comprises an input layer, an encoding layer and a knowledge integration layer which are sequentially connected.
And converting by utilizing the input layer according to the starting range data and the topology information to obtain an embedded vector.
And extracting semantic information by utilizing a coding layer according to the embedded vector.
And extracting target characteristic data by utilizing the knowledge integration layer according to the semantic information.
Optionally, the text generation model includes a multitasking learning layer, a fine tuning layer and an output layer connected in sequence.
Optionally, the training process of the power grid new equipment starting scheme generation model includes: acquiring training data of a target power grid; the training data comprises historical power grid topology information, a historical power grid new equipment starting scheme and historical power grid starting range data. And dividing the training data into a training set and a testing set according to a set proportion. And performing offline training on the model by using the training set and the testing set, taking historical power grid topology information and historical power grid starting range data as model input, taking the historical power grid new equipment starting scheme as model output and taking a minimized loss function as a target to obtain a power grid new equipment starting scheme generating model.
Optionally, the loss function is a cross entropy loss function.
The invention also provides a system for generating the starting scheme of the new equipment of the power grid, which comprises the following steps: the acquisition module is used for acquiring starting range data of the target power grid and topology information of the target power grid. And the feature extraction module is used for carrying out feature extraction by utilizing an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model according to the starting range data and the topology information to obtain target feature data. And the generating module is used for generating a new equipment starting generating scheme of the power grid by utilizing a text generating model of the new equipment starting scheme generating model of the power grid according to the target characteristic data.
The present invention also provides an electronic device including: one or more processors; a storage device having one or more programs stored thereon; the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the methods as described.
The invention also provides a computer storage medium having stored thereon a computer program which, when executed by a processor, implements a method as described.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the method comprises the steps of obtaining starting range data of a target power grid and topology information of the target power grid; according to the starting range data and the topology information, performing feature extraction by using an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model to obtain target feature data; and generating a new equipment starting generation scheme of the power grid by utilizing a text generation model of the new equipment starting scheme generation model of the power grid according to the target characteristic data. The method and the system can generate the power grid new equipment starting generation scheme directly through the starting range data and the topology information of the target power grid by utilizing the power grid new equipment starting scheme generation model, so that the generation speed of the power grid new equipment starting generation scheme is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for generating a power grid new device starting scheme.
Fig. 2 is a schematic diagram of a method for generating a new device start scheme of a power grid in practical application.
Fig. 3 is a model structure diagram of a method for generating a new device start scheme of a power grid.
Fig. 4 is a specific model diagram of a method for generating a new device start-up scheme of a power grid.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a method, a system, equipment and a medium for generating a power grid new equipment starting scheme, which can improve the generating speed of the power grid new equipment starting scheme.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1 to 4, the method for generating the power grid new equipment starting scheme provided by the invention comprises the following steps.
Step 101: and acquiring starting range data of the target power grid and topology information of the target power grid.
Step 102: and carrying out feature extraction by using an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model according to the starting range data and the topology information to obtain target feature data. The power grid new equipment starting scheme generating model is an ERNIE-based power grid new equipment starting scheme generating model based on fusion expert rules.
The expert rule knowledge fusion network comprises an input layer, a coding layer and a knowledge integration layer which are sequentially connected; converting by using the input layer according to the starting range data and the topology information to obtain an embedded vector; extracting semantic information by using a coding layer according to the embedded vector; and extracting target characteristic data by utilizing the knowledge integration layer according to the semantic information. The expert rule knowledge fusion network is used for reading the target starting range and the topology information of the target power grid at the input layer, and combining the expert rules at the knowledge fusion layer to perform feature extraction so as to obtain target feature data.
Step 103: and generating a new equipment starting generation scheme of the power grid by utilizing a text generation model of the new equipment starting scheme generation model of the power grid according to the target characteristic data. And the text generation model reads the characteristics and outputs a starting scheme of the target.
The text generation model comprises a multi-task learning layer, a fine adjustment layer and an output layer which are sequentially connected.
The training process of the power grid new equipment starting scheme generation model comprises the following steps: acquiring training data of a target power grid; the training data comprise historical power grid topology information, a historical power grid new equipment starting scheme and historical power grid starting range data; dividing the training data into a training set and a testing set according to a set proportion; and performing offline training on the model by using the training set and the testing set, taking historical power grid topology information and historical power grid starting range data as model input, taking the historical power grid new equipment starting scheme as model output and taking a minimized loss function as a target to obtain a power grid new equipment starting scheme generating model. The loss function is a cross entropy loss function.
Specifically, a power grid new equipment starting scheme generating model is constructed; an ERNIE-based power grid new equipment starting scheme generation model based on fusion expert rules comprises the following steps: expert rule knowledge fuses the network and the trained text to generate a model.
And inputting the topology information, the historical power grid new equipment starting scheme data and the corresponding starting range data training set into the expert rule knowledge fusion network for reading, encoding through an encoding layer and linking an entity in an input text to an entity in an expert rule at a knowledge fusion layer, so that the text information and the knowledge information are combined to obtain target feature data.
Taking the target characteristic data as input of the text generation model, and training parameters of the text generation model by taking the minimum cross entropy loss predicted by the minimized language model as a target to obtain a trained text generation model; the parameters include: text generation accuracy.
And taking the test set and the corresponding new equipment starting scheme as input of a trained text generation model, and adjusting parameters of the trained text generation model to obtain the trained text generation model.
Inputting the power grid new equipment starting scheme data and the corresponding topology information into an expert rule knowledge fusion network for feature extraction, and obtaining starting scheme training feature data based on expert rule entity link at a knowledge fusion layer, wherein the method specifically comprises the following steps: inputting the training starting scheme data and the corresponding power grid topology information data into the model, and performing text preprocessing and text embedding on the data; mapping into a high-dimensional space in a Transformer architecture of the coding layer based on the text embedding results; based on expert rules in a knowledge fusion layer, splicing the expert rules with a new equipment starting scheme, and introducing the expert rules into a high-dimensional space; vectors in high-dimensional space may capture relationships between word senses, grammar structures, and text, thereby determining training feature data.
The loss function adopts a cross entropy function for minimizing language model prediction; the cross entropy loss function is expressed as follows.
Wherein,is cross entropy loss; />Is the length of the sequence; />Is the size of the vocabulary; />Is a one-hot vector representing the actual next word. In this vector, the>The element 1 indicates that the actual next word is the first word in the vocabularyEach word, the rest elements are 0; />The next word to be model predicted is +.>Probability of individual words. i represents a position in the sequence.
As shown in fig. 2, in practical application, the method for generating the new equipment starting scheme of the power grid includes the following operation steps.
S1: and the historical power grid new equipment starting scheme data corresponds to the power grid topology information of the corresponding equipment, and the historical power grid new equipment starting scheme corresponds to the corresponding expert rule statement in a physical link mode. The starting scheme comprises a starting range, starting conditions and starting steps. The new devices must be in a network topology where the start-up must be, all new devices must correspond to a topology. For each start condition, the sentence of the start scheme is related to or corresponds to one or more sentences of the rule.
S2: and constructing an ERNIE-based power grid new equipment starting scheme generating model based on fusion expert rules.
Specifically, ERNIE's design inspiration comes from BERT (Bidirectional Encoder Representations from Transformers), but it improves the quality of the text representation by integrating additional knowledge. ERINE adopts heterogeneous corpus to pretrain, adopts mixed corpus of Chinese wikipedia, baiyaoki news and Baiyaoki bar, so that the text feature extraction capacity and the text prediction generation capacity are stronger compared with other foreign models such as BERT when Chinese text is processed. In the aspect of model pre-training, the mask set by ERINE for Chinese can better realize the Chinese generation prediction capability compared with the large pre-training models such as BERT.
The ERNIE model consists of an input layer, a coding layer, a knowledge integration layer, a multitasking learning layer, a fine tuning layer and an output layer.
The input layer reads the starting scheme of the new device and the corresponding power grid topology information, converts the text information and the topology information together into the form of embedded vectors, and inputs the text information and the topology information together for further processing and transmission to the encoding layer. The coding layer encodes the input text using a transform's Self-attention mechanism (Self-Attention Mechanism). This layer includes a plurality of encoders stacked together, each including a Multi-Head Attention (Multi-Head Attention) and a feed-forward neural network (Feedforward Neural Network) to better account for capturing semantic information and relationships in text, while also being able to handle context at different locations. The knowledge integration layer links additional expert rules with the text information in a physical link and maps the information passed by the input layer into a high-dimensional space and extracts information features in the encoding layer. The multi-task learning layer carries out joint training in different downstream tasks of the model by transmitting information obtained from the coding layer so as to improve the generalization capability of the model, and the fine tuning layer is positioned at the top of the model and is used for carrying out fine tuning on model parameters of a text generation task required by a set data of the model, finally determining the parameters of the model and transmitting a generation result to the output layer at the part of the fine tuning layer for completing decoding; the output layer will perform text sequence generation for the set text generation task of the text until complete text is generated.
The multi-headed self-attention mechanism of the transducer architecture is used in the "attention layer" of the transducer, rather than being provided in the feed-forward network.
The output of the self-attention mechanism would be directly connected to the input of the feed forward network. In each encoder and decoder layer of the transducer, a self-attention mechanism is used to capture the dependencies between different positions in the input sequence. Each position of this self-attention output is associated with all other positions in the input sequence in order to capture global information.
The output of this self-attention mechanism is fed to a feed forward network to non-linearly transform the representation of each location. Wherein the self-attention mechanism is multi-head self-attention.
The multi-headed self-attentive input data is typically a sequence of text, such as a sentence or a piece of text, all given in terms of word-embedded vectors, and the output is a set of weighted combined representations. The multi-headed self-attention layer captures the dependencies inside the input sequence by calculating the degree of association (attention weight) between each location and all other locations. This is achieved by dot-integrating each position of the input sequence with all other positions and then normalizing the score. This assigns each location an attention weight vector.
Input data of the feed-forward network: the input to the feed forward network is the output of the multi-headed self-attention, i.e. a representation that has been weighted combined.
The treatment process comprises the following steps: the feed forward network performs a nonlinear transformation of the representation at each location independently, mapping the original representation to a new, higher dimensional representation space, allowing the model to capture more complex features.
Outputting data: the output of the feed forward network is a new representation of each location that has been transformed by a nonlinear transformation.
The input layer receives the original text data and topology information data, an embedded vector is obtained through an embedded form, the coding layer carries out information extraction according to the embedded vector so as to capture the context information of the text and the relation with knowledge in the knowledge integration layer, the output of the information integration layer is a semantic information representation of a higher level after being processed by a self-attention mechanism and a feedforward neural network, and the knowledge integration layer maps additional expert rules with the text data in a physical link mode to a high-dimensional space together with the output result of the coding layer for further information feature extraction. The input of the multi-task learning layer is the data and the labels of various tasks to finally generate the representation and the output of different tasks, and the invention only aims at a single text generation task, so that the parameter fine adjustment of the text generation task is carried out only aiming at the fine adjustment layer; the fine tuning layer is positioned at the top of the model and is used for carrying out fine tuning on model parameters of a text generation task required by the set data set of the model, and finally determining the parameters of the model. After the model parameters are determined, the model will ultimately generate a text hidden state vector for the predicted outcome of the generating task, which contains the current state of the model generating text and context information. This hidden state vector is the primary input to the output layer to select the next word or subword according to the current state. The output layer decodes from this vector to obtain the final generated text.
S3: dividing the power grid graph sample set into a training set and a test set according to the ratio of 8:2 to generate a model, performing offline training on the model, and outputting a power grid new equipment starting scheme under different types of equipment.
As shown in fig. 3, the device starting scheme and the corresponding topology information sample are divided into a training set and a testing set according to the ratio of 8:2, the sample is input into an ERNIE-based power grid new device starting scheme generating model based on fusion expert rules, text information and topology information are mapped into a high-dimensional space by using the ERNIE model, the expert rules are input into a knowledge fusion layer to be in physical link with the starting scheme, so that the text is introduced into the high-dimensional space, the relation among word meaning, grammar structure and text is captured in the high-dimensional space to obtain sample feature data, the sample feature data is read by using a text generating model, and finally the generated text is realized. Training the model, outputting power grid starting schemes under different new devices, and checking the generating capacity of the model by using a test set.
In order to realize the online generation of the deep learning-based power grid starting scheme integrating expert rules and simultaneously improve the generation speed and accuracy of a new equipment starting scheme of a power grid, the method provided by the invention considers the influence of machine learning on the fact that the type of new equipment which does not appear in a sample is difficult to generate a correct result, and combines the expert rules of the new equipment starting scheme, thereby realizing the mutual combination of the advantages of quick generation of the machine learning and the advantages of the rigor of the expert rules; and the constructed ERNIE-based power grid new equipment starting scheme generating model is based on fusion expert rules, the expert rule knowledge fusion network in the model reads the topology information, historical power grid new equipment starting scheme data and corresponding starting range data training set input, the data are encoded through an encoding layer and the entities in the input text are linked to the entities in the expert rules in a knowledge fusion layer, so that the text information and the knowledge information are combined to obtain target characteristic data, and the text generating model based on cross entropy loss predicted by the minimized language model is used for outputting a power grid new equipment starting scheme generating result. According to the method, the corresponding relation between different equipment and corresponding topology information and a starting scheme is learned through an offline training model, new equipment of the power grid and the corresponding topology information are directly input into the model during online application, the starting scheme of the new equipment of the power grid is obtained, and the generation of the starting scheme of the new equipment of the power grid from end to end is realized.
In the expert rule knowledge fusion network, as shown in fig. 4, firstly, a starting scheme is matched with a corresponding starting scheme and topology information to be input as an input layer, and the text information enters an encoding layer after being processed by an attack input layer; the encoding layer masks through the masking strategies of the three layers of the basic level mask, the phrase level mask and the entity level mask, maps the masking strategies into a high-dimensional space together with additional expert rules in the knowledge fusion layer, extracts information features and transmits the results to the text generator.
The text generator selects a text generation task in the multi-task learning layer, aims at text generation in the fine tuning layer, aims at minimum loss function and trains, and after model training is finished, the text generator outputs the text according to the output result of the expert rule knowledge fusion network, namely, generates a new equipment starting scheme of the power grid.
The invention also provides a system for generating the starting scheme of the new equipment of the power grid, which comprises the following steps: the acquisition module is used for acquiring starting range data of the target power grid and topology information of the target power grid. And the feature extraction module is used for carrying out feature extraction by utilizing an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model according to the starting range data and the topology information to obtain target feature data. And the generating module is used for generating a new equipment starting generating scheme of the power grid by utilizing a text generating model of the new equipment starting scheme generating model of the power grid according to the target characteristic data.
The present invention also provides an electronic device including: one or more processors; a storage device having one or more programs stored thereon; the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the methods as described.
The invention also provides a computer storage medium having stored thereon a computer program which, when executed by a processor, implements a method as described.
The method comprises the steps of obtaining a starting range of a target power grid; inputting the target power grid starting range and corresponding power grid topology information into an ERNIE-based power grid new equipment starting scheme generating model based on fusion expert rules to obtain a new equipment starting scheme of the target power grid; expert rule knowledge fusion network, and trained text generation model; the expert rule knowledge fusion network is used for reading the target starting scheme and the topology information of the target power grid at the input layer, and combining the expert rules at the knowledge fusion layer to perform feature extraction so as to obtain target feature data; the trained text generation model reads the characteristics and outputs a starting scheme of the target; the method and the device can improve the generation speed and the accuracy of the new equipment starting scheme.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (8)
1. The method for generating the power grid new equipment starting scheme is characterized by comprising the following steps of:
acquiring starting range data of a target power grid and topology information of the target power grid;
according to the starting range data and the topology information, performing feature extraction by using an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model to obtain target feature data;
and generating a new equipment starting generation scheme of the power grid by utilizing a text generation model of the new equipment starting scheme generation model of the power grid according to the target characteristic data.
2. The power grid new equipment starting scheme generation method according to claim 1, wherein the expert rule knowledge fusion network comprises an input layer, a coding layer and a knowledge integration layer which are sequentially connected;
converting by using the input layer according to the starting range data and the topology information to obtain an embedded vector;
extracting semantic information by using a coding layer according to the embedded vector;
and extracting target characteristic data by utilizing the knowledge integration layer according to the semantic information.
3. The method for generating a new power grid equipment starting scheme according to claim 1, wherein the text generation model comprises a multi-task learning layer, a fine tuning layer and an output layer which are sequentially connected.
4. The power grid new equipment start-up scenario generation method according to claim 1, wherein the training process of the power grid new equipment start-up scenario generation model comprises:
acquiring training data of a target power grid; the training data comprise historical power grid topology information, a historical power grid new equipment starting scheme and historical power grid starting range data;
dividing the training data into a training set and a testing set according to a set proportion;
and performing offline training on the model by using the training set and the testing set, taking historical power grid topology information and historical power grid starting range data as model input, taking the historical power grid new equipment starting scheme as model output and taking a minimized loss function as a target to obtain a power grid new equipment starting scheme generating model.
5. The method of generating a new plant start-up scheme for a power grid of claim 4, wherein the loss function is a cross entropy loss function.
6. A power grid new equipment start-up scheme generation system, comprising:
the acquisition module is used for acquiring starting range data of the target power grid and topology information of the target power grid;
the feature extraction module is used for carrying out feature extraction by utilizing an expert rule knowledge fusion network of a power grid new equipment starting scheme generation model according to the starting range data and the topology information to obtain target feature data;
and the generating module is used for generating a new equipment starting generating scheme of the power grid by utilizing a text generating model of the new equipment starting scheme generating model of the power grid according to the target characteristic data.
7. An electronic device, comprising:
one or more processors;
a storage device having one or more programs stored thereon;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-5.
8. A computer storage medium, having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410094573.9A CN117609769A (en) | 2024-01-24 | 2024-01-24 | Method, system, equipment and medium for generating power grid new equipment starting scheme |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410094573.9A CN117609769A (en) | 2024-01-24 | 2024-01-24 | Method, system, equipment and medium for generating power grid new equipment starting scheme |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117609769A true CN117609769A (en) | 2024-02-27 |
Family
ID=89958355
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410094573.9A Pending CN117609769A (en) | 2024-01-24 | 2024-01-24 | Method, system, equipment and medium for generating power grid new equipment starting scheme |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117609769A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109408152A (en) * | 2018-11-01 | 2019-03-01 | 国网江苏省电力有限公司扬州供电分公司 | A kind of generation method and generating means of equipment starting step |
| CN111524031A (en) * | 2020-04-24 | 2020-08-11 | 内蒙古电力(集团)有限责任公司包头供电局 | Implementation method and system for intelligently generating operation steps based on analysis operation tasks |
| CN113300340A (en) * | 2021-06-01 | 2021-08-24 | 合肥工业大学 | Automatic compilation method for power grid new equipment relay protection starting scheme |
| CN114266427A (en) * | 2021-09-01 | 2022-04-01 | 国网浙江省电力有限公司绍兴供电公司 | Topology analysis-based new equipment commissioning starting scheme generation method |
-
2024
- 2024-01-24 CN CN202410094573.9A patent/CN117609769A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109408152A (en) * | 2018-11-01 | 2019-03-01 | 国网江苏省电力有限公司扬州供电分公司 | A kind of generation method and generating means of equipment starting step |
| CN111524031A (en) * | 2020-04-24 | 2020-08-11 | 内蒙古电力(集团)有限责任公司包头供电局 | Implementation method and system for intelligently generating operation steps based on analysis operation tasks |
| CN113300340A (en) * | 2021-06-01 | 2021-08-24 | 合肥工业大学 | Automatic compilation method for power grid new equipment relay protection starting scheme |
| CN114266427A (en) * | 2021-09-01 | 2022-04-01 | 国网浙江省电力有限公司绍兴供电公司 | Topology analysis-based new equipment commissioning starting scheme generation method |
Non-Patent Citations (1)
| Title |
|---|
| 若年封尘: "详细介绍ERNIE 3.0: Large-scale Knowledge Enhanced Pre-training for Language Understanding and Generation", HTTPS://BLOG.CSDN.NET/ZAG666/ARTICLE/DETAILS/128146473, 5 December 2022 (2022-12-05), pages 1 - 27 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113987209B (en) | Natural language processing method, device, computing equipment and storage medium based on knowledge-guided prefix fine adjustment | |
| CN112084841B (en) | Cross-mode image multi-style subtitle generating method and system | |
| CN111325323A (en) | Power transmission and transformation scene description automatic generation method fusing global information and local information | |
| CN114612891B (en) | Image description generation method and medium based on contrast learning and self-adaptive attention | |
| CN112989796A (en) | Text named entity information identification method based on syntactic guidance | |
| CN107463928A (en) | Word sequence error correction algorithm, system and its equipment based on OCR and two-way LSTM | |
| CN114462018A (en) | Password guessing system and method based on Transformer model and deep reinforcement learning | |
| CN115730070A (en) | Man-machine common situation conversation method, system, electronic equipment and medium | |
| Huang et al. | Personalized dialogue generation with persona-adaptive attention | |
| CN115719072A (en) | Chapter-level neural machine translation method and system based on mask mechanism | |
| CN117609769A (en) | Method, system, equipment and medium for generating power grid new equipment starting scheme | |
| CN117875395A (en) | Training method, device and storage medium of multi-mode pre-training model | |
| CN116432663B (en) | Controllable diversity professional text generation method and system based on element diagram | |
| Li et al. | 3M: Multi-style image caption generation using Multi-modality features under Multi-UPDOWN model | |
| CN116108856B (en) | Emotion recognition method and system based on long and short loop cognition and latent emotion display interaction | |
| CN112364659A (en) | Unsupervised semantic representation automatic identification method and unsupervised semantic representation automatic identification device | |
| Xue et al. | Dual path modeling for semantic matching by perceiving subtle conflicts | |
| CN115496134B (en) | Traffic scene video description generation method and device based on multi-mode feature fusion | |
| CN113469260B (en) | Visual description method based on convolutional neural network, attention mechanism and self-attention converter | |
| CN114638227A (en) | Named entity identification method, device and storage medium | |
| Zou et al. | Associated lattice-bert for spoken language understanding | |
| JP2017182277A (en) | Coding device, decoding device, discrete series conversion device, method and program | |
| CN117787293B (en) | Personalized dialogue generation method and system based on large language model | |
| Latha et al. | Visual audio summarization based on NLP models | |
| CN110472242B (en) | Text processing method, device and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |