Recommendation method, device, equipment and product of new energy system design scheme

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a new energy system design scheme recommendation method and device, equipment and a product, and the method comprises the steps: extracting design demand features of a design demand description text of a new energy system through a large language model; determining design demand feature values of the design demand features, and obtaining a design demand feature combination according to the design demand features and the design demand feature value combination of the design demand features; converting the design demand feature combination into a feature vector combination based on a semantic embedding model, and determining a target new energy system design scheme according to the similarity between the feature vector combination and a preset feature vector combination corresponding to each preset new energy system design scheme in a database; and outputting the target new energy system design scheme, thereby improving the new energy system design efficiency.

Description

Recommended methods, devices, equipment and products for new energy system design. Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to recommended methods, equipment and products for designing new energy systems. Background Technology

[0002] With the urgent global demand for renewable energy, new energy, as a highly promising renewable resource, is being widely utilized. In this process, the economic viability of new energy needs to be considered, and precise and efficient new energy system design is a key link in realizing the economic viability of new energy.

[0003] In related technologies, a series of new energy system design software have been developed. These software programs provide designers with a powerful tool platform, facilitating the design of new energy systems.

[0004] However, when using these software programs to design new energy systems, users need to manually input a series of key parameters, including but not limited to geographical coordinates, meteorological files, and specific configuration parameters of the new energy system such as photovoltaic module type, installation angle, and inverter specifications, which affects the design efficiency of the new energy system. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, and product for recommending new energy system design schemes. It aims to improve the efficiency of new energy system design by combining a large language model and a semantic embedding model to quickly recommend new energy system design schemes based on the input new energy system design requirement description text.

[0006] To achieve the above objectives, this application proposes a method for recommending a new energy system design scheme. The method includes: extracting design requirement features from the design requirement description text of the new energy system using a large language model; determining the design requirement feature values ​​of the design requirement features, and obtaining a combination of design requirement features based on the design requirement features and their design requirement feature values; converting the combination of design requirement features into a feature vector based on a semantic embedding model, and determining the target new energy system design scheme based on the similarity between the feature vector and the preset feature vectors corresponding to various preset new energy system design schemes in the database; and outputting the target new energy system design scheme.

[0007] In one embodiment, determining the design requirement feature value of a design requirement feature includes: searching for a preset feature corresponding to the design requirement feature in a database and obtaining the preset feature value of the preset feature; and obtaining the design requirement feature value of the design requirement feature based on the preset feature value of the preset feature.

[0008] In one embodiment, before the step of searching for the preset feature corresponding to the design requirement feature from the database and obtaining the preset feature value of the preset feature, the method further includes: obtaining a new energy system design scheme sample; performing feature encoding on the preset features of the new energy system design scheme sample to obtain the preset feature value of the preset feature; establishing a correspondence between the preset feature and the preset feature value respectively, and storing the correspondence in the database.

[0009] In one embodiment, the step of encoding preset features of a new energy system design scheme sample to obtain preset feature values ​​includes: when the preset feature is a categorical feature, encoding the categorical feature using raw data to obtain the preset feature value of the categorical feature; when the preset feature is a numerical feature, encoding the numerical feature using discrete interval values ​​to obtain the preset feature value of the numerical feature; when the preset feature is an address feature, encoding the address feature using administrative regions or geographical regions to obtain the preset feature value of the address feature.

[0010] In one embodiment, when there are multiple combinations of design requirement features, the steps of converting the combinations of design requirement features into feature vectors based on a semantic embedding model, and determining the target new energy system design scheme based on the similarity between the feature vectors and the preset feature vector combinations corresponding to each preset new energy system design scheme in the database include: converting each combination of design requirement features into feature vectors based on a semantic embedding model, and combining the feature vectors to obtain a feature vector combination; determining the weight of each feature vector in the feature vector combination; performing a weighted average calculation based on each feature vector and its weight to obtain a weighted average vector corresponding to the feature vector combination; and determining the target new energy system design scheme based on the similarity between the weighted average vector and the preset weighted average vector of each preset feature vector combination.

[0011] In one embodiment, the step of determining the target new energy system design scheme based on the similarity between the weighted average vector and each preset weighted average vector includes: obtaining a preset new energy system design scheme with a similarity greater than a preset similarity; and determining the preset new energy system design scheme as the target new energy system design scheme.

[0012] In one embodiment, before the steps of converting design requirement feature combinations into feature vectors based on a semantic embedding model, and determining the target new energy system design scheme based on the similarity between the feature vectors and the preset feature vectors corresponding to various preset new energy system design schemes in the database, the method further includes: obtaining feature combination sample pairs, wherein the feature combination sample pairs include positive feature combination sample pairs and negative feature combination sample pairs; using an initial semantic embedding model, calculating the feature vectors corresponding to each of the two feature combinations in the feature combination sample pairs; determining the similarity between the feature vectors of the two feature combinations; determining a contrast loss value based on the similarity, and using the contrast loss value to iteratively train the initial semantic embedding model until the obtained contrast loss value reaches a preset value, at which point the iterative training stops and the semantic embedding model is output.

[0013] In one embodiment, after determining the design requirement feature value of the design requirement feature and obtaining the design requirement feature combination based on the design requirement feature and the design requirement feature value of the design requirement feature, the method further includes: when null values ​​are detected in the design requirement feature combination, generating query text based on the null value feature and outputting the query text; receiving the response text of the query text and extracting the target feature value of the null value feature from the response text through a large language model; updating the null value with the target feature value to obtain the updated design requirement feature combination, and converting the updated design requirement feature combination into a feature vector combination based on a semantic embedding model.

[0014] Furthermore, to achieve the above objectives, this application also proposes a device for recommending a new energy system design scheme, comprising: an extraction module for extracting design requirement features from the design requirement description text of the new energy system using a language model; a determination module for determining the design requirement feature values ​​of the design requirement features and obtaining a combination of design requirement features based on the design requirement features and their design requirement feature values; converting the combination of design requirement features into a feature vector combination based on a semantic embedding model, and determining a target new energy system design scheme based on the similarity between the feature vector combination and the preset feature vector combinations corresponding to various preset new energy system design schemes in the database; and an output module for outputting the target new energy system design scheme.

[0015] In addition, to achieve the above objectives, this application also proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the recommended method for the new energy system design scheme described above.

[0016] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the recommended method for the new energy system design scheme described above.

[0017] Compared to the manual design of new energy systems using related technologies, this application employs a large language model to extract design requirement features from the design requirement description text of the new energy system. Then, a semantic embedding model is used to calculate the similarity between these design requirement features and the design requirement features of existing preset new energy system design schemes in the database. This allows for the identification and recommendation of target new energy system design schemes that match the design requirements. By combining a large language model and a semantic embedding model, new energy system design schemes can be quickly recommended based on the input design requirement description text, thereby improving the efficiency of new energy system design. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 is a flowchart illustrating the first embodiment of the recommended method for the new energy system design scheme of this application; Figure 2 is a flowchart illustrating the second embodiment of the recommended method for the new energy system design scheme of this application; Figure 3 is a detailed flowchart illustrating the second embodiment of the recommended method for the new energy system design scheme of this application; Figure 4 is a flowchart illustrating the third embodiment of the recommended method for the new energy system design scheme of this application; Figure 5 is a flowchart illustrating the fourth embodiment of the recommended method for the new energy system design scheme of this application; Figure 6 is a schematic diagram illustrating the modular structure of the recommended device for the new energy system design scheme of this application; Figure 7 is a schematic diagram illustrating the structure of the electronic device of this application.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0023] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0024] Currently, a series of new energy system design software programs have been developed, providing designers with powerful tools and facilitating the design of new energy systems. However, when using these software programs for new energy system design, users need to manually input a series of key parameters, including but not limited to geographical coordinates, meteorological files, and specific configuration parameters of the new energy system such as photovoltaic module type, installation angle, and inverter specifications, which affects the efficiency of new energy system design.

[0025] To address the aforementioned issues, this application proposes a method for recommending new energy system design schemes. The main technical solutions include: extracting design requirement features from the design requirement description text of the new energy system using a large language model; determining the design requirement feature values ​​of the design requirement features, and obtaining a combination of design requirement features based on the design requirement features and their design requirement feature values; converting the combination of design requirement features into feature vectors based on a semantic embedding model, and determining the target new energy system design scheme based on the similarity between the feature vectors and the preset feature vectors corresponding to various preset new energy system design schemes in the database; and outputting the target new energy system design scheme.

[0026] Compared to the manual design of new energy systems using related technologies, this application employs a large language model to extract design requirement features from the design requirement description text of the new energy system. Then, a semantic embedding model is used to calculate the similarity between these design requirement features and the design requirement features of existing preset new energy system design schemes in the database. This allows for the identification and recommendation of target new energy system design schemes that match the design requirements. By combining a large language model and a semantic embedding model, new energy system design schemes can be quickly recommended based on the input design requirement description text, thereby improving the efficiency of new energy system design.

[0027] Furthermore, this application transforms existing new energy system design parameters into semantic vectors and stores them in a database. It then uses a large language model and a semantic embedding model to match and recommend preset new energy system design schemes with high similarity from the database. The large language model and semantic embedding model assist manual operation, thereby improving the convenience of new energy system design and the accuracy of new energy system design schemes.

[0028] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions. The following description uses an electronic device as an example to illustrate this embodiment and the subsequent embodiments.

[0029] The following is a brief introduction to the large language model and semantic embedding model used in this application: First, the large language model (LLM) has natural language understanding capabilities, which can extract information from the irregular natural language input by the user (i.e., the design requirement description text of this application) and actively ask the user questions when information is missing.

[0030] The core technologies of large language models include pre-training and fine-tuning: Pre-training: In the pre-training stage, large language models are trained using large-scale text datasets to learn the patterns, structure and semantic information of language.

[0031] Fine-tuning: In the fine-tuning stage, the pre-trained large language model is applied to a specific task or dataset, and the model parameters are adjusted through supervised learning to adapt them to the requirements of the specific task.

[0032] Second, the semantic embedding model is a natural language processing model based on deep learning technology. Its core function is to transform the design requirement features of this application into continuous vector representations, enabling computers to understand and compare the semantic similarity between texts. The main idea of ​​the semantic embedding model is to map words or phrases in natural language into a high-dimensional vector space, so that the vector representations in this space can reflect the semantic information in the design requirement description text. This vector representation can be regarded as the encoding of semantic information, and the design scheme of the target new energy system is determined by comparing the distance and similarity between different vectors.

[0033] The core technologies of semantic embedding models include word vectors and the combined encoding of design requirement features.

[0034] Word vectors are the process of converting words in natural language into vectors. Common word vector models include Word2Vec and GloVe. These models use deep learning techniques, employing neural networks trained on large amounts of text data to learn the semantic relationships between words and represent each word as a high-dimensional vector.

[0035] Design requirement feature combination encoding is the process of converting combinations of design requirement features into feature vectors. Common encoding models include BERT, ELMo, and FastText.

[0036] It should be noted that the new energy system in this application can be a wind power generation system, an energy storage system, a photovoltaic power generation system, a hydrogen production system, etc. This application takes a photovoltaic power generation system as an example of a new energy system.

[0037] Based on this, this application provides a method for recommending a new energy system design scheme. Referring to Figure 1, Figure 1 is a flowchart of the first embodiment of the method for recommending a new energy system design scheme of this application.

[0038] In this embodiment, the method for recommending a new energy system design scheme includes steps S10 to S40: Step S10, extracting the design requirement features of the design requirement description text of the new energy system through a large language model.

[0039] First, a dialog box is designed in the front-end system. Users input the design requirements description text for the new energy system in the dialog box. Here, the front-end system refers to the interface and related technical implementation that directly interacts with the user; it is responsible for receiving user input and displaying information. Next, a large language model is invoked to extract design requirement features from the user-input design requirement description text.

[0040] It should be noted that the large language model can not only receive user input of design requirement description text for new energy systems and extract design requirement features, but also receive input information of various types, including voice, documents, images, and files related to the design requirements of new energy systems, and extract design requirement features based on this input information.

[0041] It should be noted that the design requirements for each new energy system are different. For example, the design requirements for wind power generation systems include, but are not limited to: site location, converter model, rated output power of wind turbine, wind shear coefficient, wake model, and wind farm layout parameters.

[0042] The design requirements for energy storage systems include, but are not limited to: installed capacity, converter model, cell model, cell capacity, maximum state of charge (SOC), and minimum state of charge (SOC).

[0043] The design requirements for a hydrogen production system include, but are not limited to: electrolyzer model, rated hydrogen production capacity, rated oxygen production capacity, rated operating voltage, rated operating current, and operating pressure.

[0044] The design requirements for photovoltaic power generation systems include, but are not limited to: address, power plant type, installed capacity, module type, module model, inverter type, and inverter model.

[0045] It should be noted that design requirement characteristics can also be divided into: categorical characteristics, numerical characteristics, and address-based characteristics.

[0046] Step S20: Determine the design requirement feature values ​​of the design requirement features, and obtain the design requirement feature combination based on the design requirement features and their design requirement feature values.

[0047] It should be noted that there may be one or more design requirement features.

[0048] In one feasible implementation, when there is only one design requirement feature, the design requirement feature can be encoded to obtain the corresponding design requirement feature value. Finally, the design requirement feature and its corresponding design requirement feature value are combined to obtain the design requirement feature combination. The feature encoding process is equivalent to assigning a value to the design requirement feature to obtain the design requirement feature value.

[0049] In another feasible implementation, when multiple design requirement features exist, each design requirement feature can be encoded separately to obtain a corresponding design requirement feature value. Each design requirement feature is then combined with its corresponding design requirement feature value to obtain multiple design requirement feature combinations. The more design requirement features there are, the more accurate the final recommended target new energy system design scheme will be.

[0050] For example, enter the design requirements description text of the new energy system in the dialog box, such as: "I want to build a floating photovoltaic power station in County C, City B, Province A, with an installed capacity of about 90MW, using B1 type inverters and A1 type components."

[0051] The large language model is invoked to extract design requirement features from the user-input text describing the design requirements of the new energy system. Each design requirement feature is then coded to obtain multiple sets of design requirement feature combinations. For example, a design requirement feature combination could be: “Address”: “H Province”, “Power Station Type”: “Floating Power Station”, “Installed Capacity”: “10-100 MW”, “Component Type”: “Null”, “Component Model”: “A1 Model”, “Inverter Type”: “Null”, “Inverter Model”: “B1 Model”.

[0052] Among them, address, power plant type, installed capacity, component type, component model, inverter type, and inverter model are all design requirement characteristics. For example, in Province H, for a floating power plant, the capacity is 10-100 MW, and the specifications are Null, A1 model, and B1 model.

[0053] Step S30: Based on the semantic embedding model, the design requirement feature combination is converted into a feature vector, and the target new energy system design scheme is determined according to the similarity between the feature vector and the preset feature vector corresponding to each preset new energy system design scheme in the database.

[0054] Semantic embedding models can include Word2Vec, GloVe, BERT, ELMo, etc. These semantic embedding models can transform the combination of design requirement features into dense feature vectors in a high-dimensional space, so that semantically similar texts are closer in the feature vector space.

[0055] For example, for semantic embedding models such as Word2Vec or GloVe, the word vector table in the model can be directly queried to obtain the feature vector representation of each design requirement feature combination. If the design requirement feature value in the design requirement feature combination contains multiple words, weighted average, TF-IDF weighting, or other methods can be used to combine the vectors of multiple words into a single feature vector.

[0056] For context-sensitive models such as BERT and ELMo, since these models can generate word vectors based on context changes, the entire design requirement feature or its key parts need to be taken as input, and the model will output one or more feature vectors representing the design requirement feature.

[0057] The feature vectors of design requirement feature combinations can be obtained by processing them in the manner described above. After obtaining the feature vectors corresponding to each design requirement feature combination, the feature vectors of each design requirement feature combination are combined into a feature vector combination.

[0058] The cosine similarity between the feature vector combination and the preset feature vector combinations of various preset new energy system design schemes in the backend database can be directly calculated: After determining the similarity, the feature vectors of the preset new energy system design schemes with the highest similarity or the top few in similarity can be selected, and their numbers i can be used to retrieve the target new energy system design scheme from the new energy system design scheme database and recommend it to the user.

[0059] Step S40: Output the target new energy system design scheme.

[0060] The design scheme of the target new energy system is output and displayed on the display interface of the front-end system.

[0061] It should be noted that the output can include one or more target new energy system design schemes, which the user can choose from. When multiple target new energy system design schemes exist, new energy system simulation software can be used to simulate and verify each scheme separately. The scheme that passes the simulation verification is determined as the final target new energy system design scheme. When multiple target new energy systems pass the simulation verification, the user can select one as the final target new energy system design scheme. This improves the accuracy of the determined target new energy system design scheme. The new energy system simulation software is used to verify the feasibility of the target new energy system design scheme and / or the absence of errors.

[0062] This embodiment provides a method for recommending new energy system design schemes. Compared to related technologies that involve manual design of new energy systems, this method uses a large language model to extract design requirement features from the design requirement description text of the new energy system. Then, a semantic embedding model is used to calculate the similarity between the design requirement features and the design requirement features of existing preset new energy system design schemes in the database, thereby identifying and recommending target new energy system design schemes that match the design requirements. By combining the processing of large language models and semantic embedding models, new energy system design schemes can be quickly recommended based on the input design requirement description text of the new energy system, improving the efficiency of new energy system design.

[0063] Based on the first embodiment of this application, in the second embodiment of this application, the same or similar content as the first embodiment can be referred to the above description, and will not be repeated hereafter. Based on this, referring to Figure 2, in order to accurately obtain the design requirement feature combination, step S20 determines the design requirement feature value of each design requirement feature to obtain the design requirement feature combination, which includes: step S11, searching for the preset feature corresponding to the design requirement feature in the database, and obtaining the preset feature value of the preset feature.

[0064] In one feasible implementation, the similarity between the design requirement feature and the preset features in the database can be calculated, and the preset feature with the highest similarity can be determined as the preset feature corresponding to the design requirement feature.

[0065] In another feasible implementation, since there may be preset features in the database that are semantically similar to the design requirement features, in order to retrieve the preset features corresponding to the design requirement features from the database more accurately, the design requirement features can be processed by synonym conversion or fuzzy word conversion, and then the preset features of the design requirement features after the synonym conversion or fuzzy word conversion can be matched from the database.

[0066] Step S12: Obtain the design requirement feature value of the design requirement feature based on the preset feature value of the preset feature.

[0067] The database pre-stores the correspondence between preset features and preset feature values ​​of different new energy system design scheme samples. First, the database can be used to find the preset features corresponding to each design requirement feature. Then, through the preset features and the correspondence between preset features and preset feature values, the preset feature values ​​of the preset features can be obtained. The preset feature values ​​of each preset feature are determined as the design requirement feature values ​​of the corresponding design requirement features, thereby achieving the purpose of assigning values ​​to each design requirement feature.

[0068] In this embodiment, by automatically searching the database for the design requirement feature value corresponding to the design requirement feature, the accurate design requirement feature value can be obtained quickly, avoiding the tedious process of manual searching and calculation, thereby improving the efficiency and accuracy of the design work. This helps to reduce human error and accelerate the design process.

[0069] Furthermore, before searching for the preset features corresponding to the design requirement features in the database and obtaining the preset feature values ​​of the preset features, in order to ensure that each design requirement feature can quickly find its corresponding design requirement feature value, it is necessary to pre-encode each feature and assign a corresponding preset value to each feature. Referring to Figure 3, before step S11, the following steps are included: Step S01, obtaining a sample of new energy system design schemes.

[0070] The sample design schemes for new energy systems can be previously used or designed. Each time a new energy system is designed, the resulting scheme can be stored in a database, from which sample design schemes can be retrieved directly. Alternatively, sample design schemes can be obtained from other sources, such as websites.

[0071] To enrich the preset features and preset feature values, and improve the accuracy of the subsequent determined design requirement feature combinations, it is advisable to obtain as many new energy system design scheme samples as possible.

[0072] Step S02: Encode the preset features of the new energy system design scheme sample to obtain the preset feature values ​​of the preset features.

[0073] Each new energy system design scheme sample may contain multiple preset features. Feature extraction can be performed on each new energy system design scheme sample to obtain the corresponding preset features. Feature encoding can be performed on each preset feature to obtain the preset feature value of each preset feature.

[0074] It should be noted that preset features can exist in different types, and different feature types should have different encoding methods. Using different encoding methods for different feature types improves the accuracy of the preset feature values. The preset feature types can be categorical features, numerical features, and address features.

[0075] Step S03: Establish the correspondence between preset features and preset feature values, and store the correspondence in the database.

[0076] Each new energy system design scheme can be numbered, and a correspondence can be established between each preset feature of a single new energy system design scheme and its corresponding preset feature value. The established correspondence can then be associated with the number of the new energy system design scheme and stored in the database.

[0077] Alternatively, data can be stored based on preset feature types. All preset features of all new energy system design schemes can be clustered based on feature types to obtain preset feature sets of different types. The correspondence between preset features and their preset feature values ​​is then stored in units of these preset feature sets. By clustering feature types, preset features with similar properties or functions can be grouped into the same set, reducing data redundancy and duplicate storage. This method also makes data storage more orderly and structured, facilitating subsequent data retrieval and processing.

[0078] In this embodiment, by encoding the preset features of each preset feature of the new energy system design scheme sample, it is ensured that each preset feature has a corresponding preset feature value, thereby improving the accuracy of the subsequent target new energy system design scheme.

[0079] Furthermore, to improve the accuracy of preset feature values ​​for different feature types, corresponding encoding methods can be set for different feature types, and different encoding methods can be used for different feature types to improve the accuracy of preset feature values ​​for each preset feature. Specifically, step S02 includes: step S021, when the preset feature is a categorical feature, the categorical feature is encoded using the original data to obtain the preset feature value of the categorical feature.

[0080] For categorical features, such as power plant type, component model, and inverter model, raw data can be used as feature encoding. Taking the power plant type parameter as an example, common values ​​for power plant type include "ground-mounted power plant," "floor-mounted power plant," and "rooftop power plant," representing a finite number of categories. Therefore, raw data can be directly used as the encoding when performing feature encoding. For example: Power plant type: Ground-mounted power plant.

[0081] Component type: Double-sided component.

[0082] Component model: A1.

[0083] Inverter type: Centralized.

[0084] Inverter model: B1.

[0085] Step S022: When the preset feature is a numerical feature, the numerical feature is encoded using discrete interval values ​​to obtain the preset feature value of the numerical feature.

[0086] For numerical features, they can be discretized into several intervals. The discrete interval values ​​are then used to encode the numerical features, resulting in preset feature values. Taking installed capacity as an example, the installed capacity of a photovoltaic power station is typically a floating-point number, ranging from less than 1 MW to hundreds of MW. Therefore, for ease of classification, it can be discretized into several intervals, such as less than 1 MW, 1-10 MW, 10-100 MW, and greater than 100 MW. For example: Installed capacity: 1-10 MW.

[0087] Step S023: When the preset feature is an address-type feature, the address-type feature is encoded using administrative regions or geographical regions to obtain the preset feature value of the address-type feature.

[0088] For address-type features, different administrative regions can be used as feature codes; or the coordinates can be clustered, dividing them into several clusters, each cluster representing a geographical region. That is, the address-type feature is encoded using administrative regions or geographical regions to obtain the preset feature value of the address-type feature. For example: Address: Province H.

[0089] It should be noted that, in addition to the categorical features, numerical features, and address features mentioned above, other feature types may also be included.

[0090] In this embodiment, corresponding encoding methods are set for different feature types, and different encoding methods are used for encoding different feature types to improve the accuracy of the preset feature values ​​of each preset feature.

[0091] Based on the first embodiment of this application, in the third embodiment of this application, the same or similar content as the first embodiment can be referred to the above description, and will not be repeated hereafter. On this basis, when there are multiple combinations of design requirement features, referring to Figure 4, step S30 includes: step S31, converting each combination of design requirement features into feature vectors based on a semantic embedding model, and combining the feature vectors to obtain a feature vector combination.

[0092] Based on the semantic embedding model, each design requirement feature combination is converted into a feature vector, as described in the first embodiment, which will not be repeated here.

[0093] After obtaining the feature vectors corresponding to each design requirement feature combination, the feature vectors of each design requirement feature combination are combined into a single feature vector combination. This can be achieved by simply concatenating all feature vectors into a single long vector; this method is simple and easy to implement. Alternatively, each feature vector can be assigned a different weight based on its importance, and then a weighted average vector can be calculated as the feature vector combination. Another approach is to use an attention mechanism to dynamically calculate the weight of each feature vector and combine them accordingly to form the final feature vector combination. This method can more flexibly capture the correlation and importance between features.

[0094] After obtaining the feature vector combination, normalization and dimensionality reduction can be performed on the feature vector combination. Subsequently, similarity calculation can be performed based on the normalized and dimensionality-reduced feature vector combination, which can improve the efficiency of similarity calculation and determine the target new energy system design scheme more quickly.

[0095] Step S32: Determine the weights of each feature vector in the feature vector combination.

[0096] The weights of each feature vector in the feature vector combination can be preset, and each feature vector can be assigned a corresponding weight according to its importance.

[0097] Step S33: Calculate the weighted average based on each feature vector and its weight to obtain the weighted average vector corresponding to the feature vector combination.

[0098] The weighted average vector corresponding to the feature vector combination can be obtained by multiplying each feature vector in the feature vector combination with its weight, adding them together, and then dividing by the number of feature vectors.

[0099] Step S34: Determine the target new energy system design scheme based on the similarity between the weighted average vector and the preset weighted average vector of each preset feature vector combination.

[0100] A pre-constructed weighted average vector for each preset feature vector combination can be pre-built, and the correspondence between the preset feature vector combinations and their corresponding pre-constructed weighted average vectors can be stored. The pre-constructed weighted average vector for the preset feature vector combinations can be calculated using the weighted averaging method described above.

[0101] The similarity between the weighted average vector corresponding to the feature vector combination and the preset weighted average vector of each preset feature vector combination is calculated, and the target system design scheme is determined based on the similarity. Here, the similarity can be cosine similarity, Euclidean distance, Manhattan distance, etc.; this application uses cosine similarity as an example.

[0102] In one feasible implementation, preset new energy system design schemes with a similarity greater than a preset similarity can be obtained; these preset new energy system design schemes are then determined as the target new energy system design scheme. The preset similarity can be set according to actual conditions. When multiple preset new energy system design schemes with a similarity greater than the preset similarity exist, one of them can be selected as the target new energy system design scheme based on the actual situation.

[0103] In another feasible implementation, a preset new energy system design scheme with the highest similarity can be obtained; the preset new energy system design scheme with the highest similarity is determined as the target new energy system design scheme.

[0104] For example, weights are assigned to each feature parameter based on its importance. ; Calculate the similarity between the weighted average vector corresponding to each feature vector combination and the preset weighted average vector corresponding to each preset feature vector combination using weights: ; ; where i above is the number, This represents the n feature vectors in each preset feature vector combination. β represents the weighted average vector corresponding to each preset feature vector combination, and β represents the weighted average vector corresponding to the feature vector combination.

[0105] Calculate the similarity between the weighted average vector corresponding to each feature vector combination and the pre-defined weighted average vector of each pre-defined feature vector combination. For example, using cosine similarity: After determining the similarity, the feature vectors of the preset new energy systems with the highest similarity or the top few in similarity can be selected, and their numbers i can be used to retrieve the target new energy system design scheme from the new energy system design scheme database and recommend it to the user.

[0106] In this embodiment, the weighted average vector can adjust the importance of different dimensions or features according to different weight configurations. This means that when processing complex data, weights can be flexibly allocated according to data characteristics and business needs, thereby better capturing subtle differences and importance between data. Furthermore, when facing noisy or inaccurate data, the weighted average method can reduce the impact of certain inaccurate or misleading features by reasonably setting weights, thereby improving overall stability and robustness.

[0107] Based on the first embodiment of this application, in the fourth embodiment of this application, the same or similar content as the first embodiment can be referred to the above description, and will not be repeated hereafter. On this basis, referring to FIG5, before step S30, the method further includes: step S110, obtaining feature combination sample pairs, wherein the feature combination sample pairs include positive feature combination sample pairs and negative feature combination sample pairs.

[0108] A positive feature combination sample pair consists of two feature combinations with similar feature codes, while a negative feature combination sample pair consists of two feature combinations with dissimilar feature codes. These positive and negative feature combination sample pairs can be constructed by professional new energy system designers, or they can be automatically constructed using large language models. These positive and negative feature combination sample pairs can be used as training data for fine-tuning semantic embedding models.

[0109] Step S120: Using the initial semantic embedding model, calculate the feature vectors corresponding to the two feature combinations in the feature combination sample pair.

[0110] Step S130: Determine the similarity between the feature vectors of the two feature combinations.

[0111] An initial semantic embedding model refers to a model that has not been trained.

[0112] The method for calculating the feature vectors corresponding to the two feature combinations in a feature combination sample pair, as well as the similarity calculation process, are the same as those described in the above embodiments and will not be repeated here.

[0113] Step S140: Determine the contrast loss value based on similarity, and use the contrast loss value to iteratively train the initial semantic embedding model until the obtained contrast loss value reaches the preset value, then stop the iterative training and output the semantic embedding model.

[0114] A contrastive loss function is constructed to calculate the similarity between two feature vectors. The contrastive loss is then calculated, and this loss is used to iterate the initial semantic embedding model, ultimately resulting in a semantic embedding model adept at calculating feature vectors of new energy systems. The preset value can be set to 0 or a very small value, indicating no loss or minimal loss. When the contrastive loss reaches the preset value, the semantic embedding model has stabilized, at which point training can be stopped and the semantic embedding model can be output.

[0115] In this embodiment, the initial semantic embedding model is trained by combining feature samples to obtain a semantic embedding model that is good at calculating feature vectors.

[0116] Based on the first embodiment of this application, in the fifth embodiment of this application, the same or similar content as the first embodiment can be referred to the above description, and will not be repeated hereafter. On this basis, after step S20, steps S210 to S230 are included: Step S210, when a null value is detected in the combination of design requirement features, query text is generated based on the null value feature and the query text is output.

[0117] Step S220: Receive the response text to the query text, and extract the target feature value of null features from the response text using a large language model.

[0118] As obtained from the first embodiment, the design requirement feature combinations can be: "Address": "Province H", "Power Plant Type": "Floating Power Plant", "Installed Capacity": "10-100 MW", "Module Type": "Null", "Module Model": "A1 Model", "Inverter Type": "Null", "Inverter Model": "B1 Model". It can be seen that many design requirement feature combinations correspond to the null value "Null". To improve the accuracy of the subsequently recommended new energy system design schemes, it is necessary to ensure the completeness of the design requirement feature combination information; that is, each design requirement feature in a design requirement feature combination needs to have a corresponding feature value.

[0119] The system generates query text based on null value features and outputs and displays the query text on the display interface. It receives the response text to the query text and extracts the target feature values ​​of the null value features from the response text using a large language model. For example, when certain design requirement feature combinations are missing feature values, the large language model organizes its language and actively requests the user to supplement the relevant information. Taking the above design requirement feature combination as an example, the missing feature values ​​are "detailed address," "component type," and "inverter type." The large language model can ask the user: "At which specific address in H province do you want to build the website? What type of components and inverter do you want to use?" In step S230, the null values ​​are updated using the target feature values ​​to obtain the updated design requirement feature combination. The updated design requirement feature combination is then converted into a feature vector combination based on a semantic embedding model.

[0120] In this embodiment, it is necessary to ensure the completeness of the design requirement feature combination information, that is, the design requirement features in each design requirement feature combination need to have corresponding feature values, so as to improve the accuracy of the subsequent recommended new energy system design scheme.

[0121] In other embodiments, all candidate feature values ​​matching the null feature can be retrieved, and query text can be generated based on all candidate feature values. Specifically, the large language model can also retrieve similar new energy system design schemes based on the current design requirement feature combination, and then obtain the feature values ​​of the features in the feature combination of the similar new energy system design scheme for the user to select. For example, the retrieved similar new energy system design schemes may be: [China, Province A, City B, District C, Street D, Floating Power Station, 10-100MW, Bifacial Modules, Model A1, Centralized Inverter, Model B1, ...], [China, Province A, City B, District E, Street F, Ground Power Station, 10-100MW, Monofacial Modules, Model A1, String Inverter, Model B1, ...]. The large language model can organize the language to prompt the user as follows: "Do you need to use bifacial modules or monofacial modules? Centralized inverter or string inverter?".

[0122] Alternatively, if the user is unable to provide or select a feature value, then an empty value can be used instead.

[0123] Alternatively, input boxes and selection boxes can be designed in the front-end system for users to input and select design requirements; feature values ​​that the user has not filled in can be replaced with empty values.

[0124] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the recommended method of the new energy system design scheme of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0125] Based on the same inventive concept, referring to Figure 6, this application also provides a device for recommending a new energy system design scheme, comprising: an extraction module 10, used to extract design requirement features from the design requirement description text of the new energy system through a language model; a determination module 20, used to determine the design requirement feature values ​​of the design requirement features, and obtain a combination of design requirement features based on the design requirement features and the design requirement feature values ​​of the design requirement features; converting the combination of design requirement features into a combination of feature vectors based on a semantic embedding model, and determining a target new energy system design scheme based on the similarity between the combination of feature vectors and the preset feature vector combinations corresponding to each preset new energy system design scheme in the database; and an output module 30, used to output the target new energy system design scheme.

[0126] Optionally, the determining module 20 is further configured to search for the preset features corresponding to the design requirement features in the database and obtain the preset feature values ​​of the preset features; and obtain the design requirement feature values ​​of the design requirement features based on the preset feature values ​​of the preset features.

[0127] Optionally, before the step of determining module 20 is further configured to search for the preset features corresponding to the design requirement features in the database and obtain the preset feature values ​​of the preset features, the module further includes: obtaining a sample of new energy system design schemes; performing feature encoding on the preset features of the sample of new energy system design schemes to obtain the preset feature values ​​of the preset features; establishing a correspondence between the preset features and the preset feature values ​​respectively, and storing the correspondence in the database.

[0128] Optionally, the determining module 20 is further used to perform feature encoding on the preset features of the new energy system design scheme sample to obtain the preset feature values ​​of the preset features. The steps include: when the preset feature is a categorical feature, the categorical feature is encoded using the original data to obtain the preset feature value of the categorical feature; when the preset feature is a numerical feature, the numerical feature is encoded using discrete interval values ​​to obtain the preset feature value of the numerical feature; when the preset feature is an address feature, the address feature is encoded using administrative regions or geographical regions to obtain the preset feature value of the address feature.

[0129] Optionally, the determining module 20 is further configured to convert each design requirement feature combination into a feature vector based on a semantic embedding model, combine the feature vectors to obtain a feature vector combination; determine the weight of each feature vector in the feature vector combination; perform a weighted average calculation based on each feature vector and its weight to obtain a weighted average vector corresponding to the feature vector combination; and determine the target new energy system design scheme based on the similarity between the weighted average vector and the preset weighted average vector of each preset feature vector combination.

[0130] Optionally, the determining module 20 is also used to obtain a preset new energy system design scheme with a similarity greater than a preset similarity; and to determine the preset new energy system design scheme as the target new energy system design scheme.

[0131] Optionally, the device further includes a training module, which is used to acquire feature combination sample pairs, wherein the feature combination sample pairs include positive feature combination sample pairs and negative feature combination sample pairs; using an initial semantic embedding model, calculate the feature vectors corresponding to the two feature combinations in the feature combination sample pairs; determine the similarity between the feature vectors of the two feature combinations; determine the contrast loss value based on the similarity, and use the contrast loss value to iteratively train the initial semantic embedding model until the obtained contrast loss value reaches a preset value, at which point the iterative training stops and the semantic embedding model is output.

[0132] Optionally, the device further includes an update module, which is used to generate an inquiry text based on the null value when a null value is detected in the design requirement feature combination, and output the inquiry text; receive the reply text of the inquiry text, and extract the target feature value of the null value from the reply text through a large language model; update the null value with the target feature value to obtain the updated design requirement feature combination, and convert the updated design requirement feature combination into a feature vector combination based on a semantic embedding model.

[0133] The new energy system design scheme recommendation device provided in this application adopts the new energy system design scheme recommendation method in the above embodiments. It can quickly recommend new energy system design schemes based on the input new energy system design requirement description text through a combination of a large language model and a semantic embedding model, thereby improving the efficiency of new energy system design. Compared with the prior art, the beneficial effects of the new energy system design scheme recommendation device provided in this application are the same as those of the new energy system design scheme recommendation method provided in the above embodiments, and other technical features in the new energy system design scheme recommendation device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0134] Based on the same inventive concept, this application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the recommended method of the new energy system design scheme in the above embodiments.

[0135] Referring now to Figure 7, a schematic diagram of an electronic device suitable for implementing embodiments of this application is shown. The electronic device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. The electronic device shown in Figure 7 is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0136] As shown in Figure 7, the electronic device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the electronic device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. While electronic devices with various systems are shown in the figures, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0137] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0138] The electronic device provided in this application, employing the recommended method of the new energy system design scheme in the above embodiments, can improve the design efficiency of new energy systems. Compared with the prior art, the beneficial effects of the electronic device provided in this application are the same as those of the recommended method of the new energy system design scheme provided in the above embodiments, and other technical features of the electronic device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0139] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0140] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0141] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the recommended method for the new energy system design scheme described above.

[0142] The computer program product provided in this application can improve the design efficiency of new energy systems. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the recommended method for new energy system design schemes provided in the above embodiments, and will not be repeated here.

[0143] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for recommending a design scheme for a new energy system, characterized in that, The method includes: extracting design requirement features from the design requirement description text of the new energy system using a large language model; determining the design requirement feature values ​​of the design requirement features, and obtaining a design requirement feature combination based on the design requirement features and their design requirement feature values; converting the design requirement feature combination into a feature vector based on a semantic embedding model, and determining a target new energy system design scheme based on the similarity between the feature vector and the preset feature vectors corresponding to various preset new energy system design schemes in the database; and outputting the target new energy system design scheme.

2. The method as described in claim 1, characterized in that, The step of determining the design requirement feature value of the design requirement feature includes: searching for a preset feature corresponding to the design requirement feature in the database and obtaining the preset feature value of the preset feature; and obtaining the design requirement feature value of the design requirement feature based on the preset feature value of the preset feature.

3. The method as described in claim 2, characterized in that, Before the step of searching the database for the preset features corresponding to the design requirement features and obtaining the preset feature values ​​of the preset features, the method further includes: obtaining a sample of new energy system design schemes; performing feature encoding on the preset features of the sample of new energy system design schemes to obtain the preset feature values ​​of the preset features; establishing a correspondence between the preset features and the preset feature values ​​respectively, and storing the correspondence in the database.

4. The method as described in claim 3, characterized in that, The step of encoding the preset features of the new energy system design scheme sample to obtain the preset feature values ​​of the preset features includes: when the preset feature is a categorical feature, encoding the categorical feature using raw data to obtain the preset feature value of the categorical feature; when the preset feature is a numerical feature, encoding the numerical feature using discrete interval values ​​to obtain the preset feature value of the numerical feature; when the preset feature is an address feature, encoding the address feature using administrative regions or geographical regions to obtain the preset feature value of the address feature.

5. The method as described in claim 1, characterized in that, When there are multiple combinations of design requirement features, the step of converting the combinations of design requirement features into feature vectors based on a semantic embedding model, and determining the target new energy system design scheme based on the similarity between the feature vectors and the preset feature vector combinations corresponding to each preset new energy system design scheme in the database, includes: converting each combination of design requirement features into feature vectors based on the semantic embedding model, and combining the feature vectors to obtain the feature vector combination; determining the weight of each feature vector in the feature vector combination; performing a weighted average calculation based on each feature vector and its weight to obtain the weighted average vector corresponding to the feature vector combination; and determining the target new energy system design scheme based on the similarity between the weighted average vector and the preset weighted average vector of each preset feature vector combination.

6. The method as described in claim 5, characterized in that, The step of determining the target new energy system design scheme based on the similarity between the weighted average vector and each of the preset weighted average vectors includes: obtaining preset new energy system design schemes with similarity greater than preset similarity; and determining the preset new energy system design scheme as the target new energy system design scheme.

7. The method as described in claim 1, characterized in that, Before the step of converting the design requirement feature combination into feature vectors based on the semantic embedding model, and determining the target new energy system design scheme based on the similarity between the feature vectors and the preset feature vectors corresponding to each preset new energy system design scheme in the database, the method further includes: obtaining feature combination sample pairs, wherein the feature combination sample pairs include positive feature combination sample pairs and negative feature combination sample pairs; using an initial semantic embedding model, calculating the feature vectors corresponding to each of the two feature combinations in the feature combination sample pairs; determining the similarity between the feature vectors of the two feature combinations; determining a contrast loss value based on the similarity, and using the contrast loss value to iteratively train the initial semantic embedding model until the obtained contrast loss value reaches a preset value, at which point the iterative training stops and the semantic embedding model is output.

8. The method as described in claim 1, characterized in that, After the steps of determining the design requirement feature value of the design requirement feature and obtaining the design requirement feature combination based on the design requirement feature and the design requirement feature value of the design requirement feature, the method further includes: when a null value is detected in the design requirement feature combination, generating an inquiry text based on the null value feature and outputting the inquiry text; receiving the reply text of the inquiry text and extracting the target feature value of the null value feature from the reply text through the large language model; updating the null value with the target feature value to obtain an updated design requirement feature combination, and converting the updated design requirement feature combination into a feature vector combination based on a semantic embedding model.

9. A recommended device for a new energy system design scheme, characterized in that, The device includes: an extraction module for extracting design requirement features from the design requirement description text of a new energy system using a language model; a determination module for determining the design requirement feature values ​​of the design requirement features and obtaining a design requirement feature combination based on the design requirement features and their design requirement feature values; converting the design requirement feature combination into a feature vector combination based on a semantic embedding model, and determining a target new energy system design scheme based on the similarity between the feature vector combination and the preset feature vector combinations corresponding to various preset new energy system design schemes in the database; and an output module for outputting the target new energy system design scheme.

10. An electronic device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the recommended method for the new energy system design scheme as described in any one of claims 1 to 8.

11. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the recommended method for the new energy system design scheme as described in any one of claims 1 to 8.

Images