A method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework

Abstract

The present invention provides a method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework, including: establishing a basic FHI logical reasoning model; repeatedly calling the basic FHI logical reasoning model for the target problems in the photovoltaic business field to obtain an FHI logical reasoning model, and then converting it into an argument fact tree; analyzing and processing each Fact fact sub-node in the argument fact tree to obtain an embedding encoding embedded with hierarchical features in the argument fact tree, and inputting it into the transformer module. The present invention enables the large language model in the photovoltaic business field not only to understand photovoltaic professional knowledge, but also to schedule the business interfaces of the information system based on the reasoning logic chain to achieve long-link reasoning analysis, making the reasoning process of the large language model conform to the business logic, enhancing the understanding depth of the large language model for the photovoltaic business, and thus improving the accuracy and reliability of the prediction and analysis results of the large language model.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data analysis and processing, and particularly relates to a method for realizing the analysis of the photovoltaic business field by a large model based on the FHI logical derivation framework. Background Art

[0002] In the photovoltaic business field, according to application requirements, various large language models are often used. For example, a large language model for evaluating the health status of photovoltaic equipment; a large language model for predicting the power generation power and power generation amount of photovoltaic equipment by combining meteorological and topographic data.

[0003] In the prior art, the large language model in the photovoltaic business field is a general pre-trained large language model: by pre-training on a large amount of data, learning the common knowledge of the language to improve the transfer learning ability and generalization ability on general tasks.

[0004] However, due to the complex equipment link relationship in the photovoltaic business field, when using a general pre-trained large language model, the processing ability for multi-hop tasks with long logical chains is weak, resulting in low accuracy of prediction results. Summary of the Invention

[0005] Aiming at the defects existing in the prior art, the present invention provides a method for realizing the analysis of the photovoltaic business field by a large model based on the FHI logical derivation framework, which can effectively solve the above problems.

[0006] The technical solution adopted by the present invention is as follows:

[0007] The present invention provides a method for realizing the analysis of the photovoltaic business field by a large model based on the FHI logical derivation framework, including the following steps:

[0008] Step S1, establishing an FHI logical reasoning basic model; the FHI logical reasoning basic model is a tree structure, including Fact fact nodes, Hypothesize rational assumption nodes, Investigate system investigation nodes, and Fact fact sub-nodes connected hierarchically from top to bottom;

[0009] Step S2, repeatedly calling the FHI logical reasoning basic model for the target problem in the photovoltaic business field to obtain an FHI logical reasoning model; the method is:

[0010] After establishing the FHI logical reasoning basic model that matches the target problem in the photovoltaic business domain, call the associated business knowledge and business logic, analyze each Fact fact sub-node, and determine whether the Fact fact sub-node is the bottommost node associated with the target problem in the photovoltaic business domain in the business logic. If so, end the re-call of the FHI logical reasoning basic model; if not, use the Fact fact sub-node as the Fact fact node, re-call the FHI logical reasoning basic model, and increase the level of the tree structure. If the loop continues, finally establish the FHI logical reasoning model;

[0011] Step S3, convert the FHI logical reasoning model into an argument fact tree; in the argument fact tree, there are only Fact fact nodes and the Fact fact sub-nodes at each level;

[0012] Step S4, analyze and process each Fact fact sub-node in the argument fact tree to obtain an embedding encoding with the hierarchical features in the argument fact tree embedded;

[0013] Step S5, input the embedding encodings with hierarchical cascade relationships into the transformer module after sorting them by level, and the transformer module outputs the analysis result for the target problem in the photovoltaic business domain.

[0014] Preferably, in step S1, the establishment process of the FHI logical reasoning basic model is as follows:

[0015] Step S1.1, store the semantic description information of the target problem into the Fact fact node;

[0016] Step S1.2, call the business knowledge and business logic associated with the target problem, analyze and reason the semantic description information of the target problem stored in the Fact fact node, and obtain N kinds of rational hypothesis texts associated with the target problem. For each rational hypothesis text, generate a Hypothesize rational hypothesis node and use it as the sub-node of the Fact fact node;

[0017] Step S1.3, for each Hypothesize rational hypothesis node, obtain M associated Investigate system investigation nodes and use them as the sub-nodes of this Hypothesize rational hypothesis node; each Investigate system investigation node is bound with a model agent, and by executing the model agent, text-based argument facts are obtained;

[0018] Step S1.4: The text-based argument facts of each Investigation system investigation node are stored in the Fact fact sub-node that is a sub-node of the Investigation system investigation node.

[0019] Thus, from top to bottom, the FHI logical reasoning basic model with a tree structure formed by Fact fact nodes, Hypothesize rational hypothesis nodes, Investigation system investigation nodes, and Fact fact sub-nodes is established.

[0020] Preferably, the business knowledge and business logic are large language models enhanced based on knowledge graphs and vector retrieval.

[0021] Preferably, the model agent bound to the Investigation system investigation node is a model agent obtained by training with historical data.

[0022] The training method is as follows:

[0023] Use key-value pairs in the form of {"text": (model agent name, key parameter list)} to train the model agent, so that the model agent has the capabilities of semantic understanding, top_N model agent recall, and key parameter extraction.

[0024] Preferably, the model agent has an input processing sub-module, a business interface call sub-module, and an integrated output sub-module.

[0025] The input processing sub-module is used to perform semantic understanding on the rational hypothesis text input from the Hypothesize rational hypothesis node.

[0026] The business interface call sub-module is used to call the associated business database or business data table according to the semantic understanding of the input processing sub-module, and query the key parameters associated with the semantic understanding as the query result.

[0027] The integrated output sub-module is used to form a key parameter table from the multiple key parameters obtained by the business interface call sub-module and output it to the Fact fact sub-node.

[0028] Preferably, in step S3, converting the FHI logical reasoning model into an argument fact tree specifically means:

[0029] Delete each Hypothesize rational hypothesis node and Investigation system investigation node in the FHI logical reasoning model, and keep the hierarchy and cascading relationship of each Fact fact sub-node unchanged to obtain the argument fact tree; among them, the hierarchy and cascading relationship are called the argument derivation hierarchy relationship.

[0030] Preferably, step S4 is specifically as follows:

[0031] The Fact fact sub-node stores text-based argument facts and the argument derivation hierarchy;

[0032] Segment the text-based argument facts. The text-based argument facts have M-dimensional segmentation, and each segmentation is transformed into a digital representation of a semantics, thereby obtaining an M-dimensional text semantic vector v_semantic;

[0033] According to the text position information of each segmentation in the text-based argument facts, through different cycle sine-cosine combinations, obtain an M-dimensional text position vector v_position;

[0034] Relying on the argument fact tree, according to the argument derivation hierarchy of the Fact fact sub-node, including the hierarchy and the order within the layer, construct an initial hierarchical logic vector with dimension D, and then obtain a structure position vector v_structure with dimension M through linear layer mapping;

[0035] Sum up the text semantic vector v_semantic, the text position vector v_position, and the structure position vector v_structure to obtain the embedding encoding embeding of the text-based argument facts.

[0036] Preferably, step S5 is specifically as follows:

[0037] The transformer module includes an encoder, a decoder, a linear layer, and a Softmax layer;

[0038] Embed and encode each of the embeding with a hierarchical cascade relationship, form an embeding embedding encoding sequence after sorting by hierarchy, and input the embeding embedding encoding sequence into the encoder;

[0039] The encoder captures the global dependencies between the elements in the input embeding embedding encoding sequence, generates a series of context-aware representations, and inputs them into the decoder;

[0040] The decoder is used to generate an output sequence according to the context-aware representation;

[0041] The linear layer is used to map the output sequence of the decoder to a vector;

[0042] The Softmax layer is used to convert the vector output by the linear layer into a probability distribution.

[0043] The method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework provided by the present invention has the following advantages:

[0044] The present invention provides a method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework, enabling the large language model in the photovoltaic business field to not only understand photovoltaic professional knowledge, but also, based on the reasoning logic chain, dispatch the business interfaces of the information system to achieve long-link reasoning analysis, making the reasoning process of the large language model conform to the business logic, enhancing the depth of understanding of the photovoltaic business by the large language model, and thus improving the accuracy and reliability of the prediction and analysis results of the large language model. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic diagram of generating an argument fact tree by the FHI logical reasoning model provided by the present invention;

[0046] Figure 2 It is a schematic diagram of enhancing the transformer module based on the argument fact tree provided by the present invention;

[0047] Figure 3 It is an example diagram of generating an FHI logical reasoning model in the photovoltaic business field provided by an embodiment of the present invention;

[0048] Figure 4 It is an example diagram of the generated argument fact tree provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0050] The present invention provides a method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework, enabling the large language model in the photovoltaic business field to not only understand photovoltaic professional knowledge, but also, based on the reasoning logic chain, dispatch the business interfaces of the information system to achieve long-link reasoning analysis, making the reasoning process of the large language model conform to the business logic, enhancing the depth of understanding of the photovoltaic business by the large language model, and thus improving the accuracy and reliability of the prediction and analysis results of the large language model.

[0051] The overall process of the present invention can be divided into the following two parts:

[0052] Relying on the FHI logical reasoning model of fact - rational hypothesis - systematic investigation (FHI), through the cyclic scheduling of semantic understanding and model agent, an argument fact tree is obtained;

[0053] Enhance the logical relationship in the positional encoding by using the hierarchical structure information of the argument fact tree, improve the understanding depth of the transformer - like model for specific photovoltaic business problems, and achieve reasoning and analysis.

[0054] The following is a detailed description of the present invention:

[0055] The present invention provides a method for a large - model to analyze the photovoltaic business field based on the FHI logical derivation framework, including the following steps:

[0056] Step S1, establish an FHI logical reasoning basic model;

[0057] Combined with Figure 1 , the FHI logical reasoning basic model is a tree - shaped structure, including Fact fact nodes, Hypothesize rational hypothesis nodes, Investigate system investigation nodes, and Fact fact sub - nodes connected hierarchically from top to bottom;

[0058] The establishment process of the FHI logical reasoning basic model is as follows:

[0059] Step S1.1, store the semantic description information of the target problem into the Fact fact node;

[0060] Step S1.2, call the business knowledge and business logic associated with the target problem, analyze and reason the semantic description information of the target problem stored in the Fact fact node, obtain N kinds of rational hypothesis texts associated with the target problem, and for each rational hypothesis text, generate a Hypothesize rational hypothesis node and use it as a sub - node of the Fact fact node;

[0061] In this step, the business knowledge and business logic are large - language models enhanced based on knowledge graphs and vector retrieval.

[0062] Step S1.3, for each Hypothesize rational hypothesis node, obtain M associated Investigate system investigation nodes and use them as sub - nodes of this Hypothesize rational hypothesis node; each Investigate system investigation node is bound with a model agent, and by executing the model agent, text - type argument facts are obtained;

[0063] In this step, the model agent bound to the Investigate system investigation node is a model agent trained with historical data;

[0064] The training method is:

[0065] The model agent is trained using key-value pairs in the form of {"text": (model agent name, list of key parameters)} so that the model agent has the capabilities of semantic understanding, top_N model agent recall, and key parameter extraction.

[0066] The model agent has an input processing sub-module, a service interface call sub-module, and an integrated output sub-module;

[0067] The input processing sub-module is used to perform semantic understanding on the rational hypothesis text input from the Hypothesize rational hypothesis node;

[0068] The service interface call sub-module is used to call the associated service database or service data table according to the semantic understanding of the input processing sub-module, and query the key parameters associated with the semantic understanding as the query result;

[0069] The integrated output sub-module is used to form a key parameter table from the multiple key parameters obtained by the service interface call sub-module and output it to the Fact fact sub-node.

[0070] Step S1.4, the text-based evidentiary facts of each Investigate system investigation node are stored in the Fact fact sub-node that is a sub-node of the Investigate system investigation node;

[0071] Thus, from top to bottom, the FHI logical reasoning basic model with a tree structure formed by the Fact fact node, Hypothesize rational hypothesis node, Investigate system investigation node, and Fact fact sub-node is established.

[0072] Therefore, in the present invention, the FHI logical reasoning basic model is a logical reasoning architecture that starts from facts, goes through model semantic understanding and model agent scheduling, and finally returns to facts, including the Fact fact node, Hypothesize rational hypothesis node, Investigate system investigation node, and Fact fact sub-node connected hierarchically from top to bottom.

[0073] The Fact fact node is used to store an objective fact state, including a description of the problem and supporting facts obtained through reasoning and investigation;

[0074] The Hypothesize rational hypothesis node is a rational hypothesis made based on the facts in the parent Fact fact node, combined with business knowledge and logic, and is the ideological guidance for the next verification work;

[0075] The Investigate system investigation node is a verification of the parent Hypothesize rational hypothesis node. In this node, it involves the scheduling selection of the proprietary model agent and the logical concatenation of the internal sub-steps of the model agent.

[0076] When the Investigate system investigation node realizes the interaction with the real world and the real business information system through the proprietary model agent, the final result will be stored in its child Fact fact node.

[0077] One Fact fact node can correspond to one or more Hypothesize rational hypothesis nodes, one Hypothesize rational hypothesis node can correspond to one or more Investigate system investigation nodes, and one Investigate system investigation node can only correspond to one Fact fact child node. The root node of the complete Fact-Hypothesize-Investigate (FHI) logical reasoning architecture must be a Fact fact node, and all terminal nodes (nodes without child nodes) must be Fact fact child nodes.

[0078] From the Fact fact node to the Hypothesize rational hypothesis node, it is an assumption based on the parent Fact fact node under business knowledge and logical constraints. The implementation process can be achieved through a large model efficiency improvement method based on knowledge graph and vector retrieval enhancement. Specifically, this method generates a vector expert library of photovoltaic domain expertise, generates a knowledge graph of the spatio-temporal associations of photovoltaic devices and states, and improves the application effect of the general large language model in the photovoltaic power field. Through this step, "generating rational hypothesis texts that conform to business logic according to factual semantics" is achieved.

[0079] From the Hypothesize rational hypothesis node to the Investigate system investigation node, the purpose is to select one or more appropriate model agents based on the semantic understanding of the rational hypothesis text of the parent node. The implementation process is as follows: constructing key-value pairs of {"text": (model agent name, list of key parameters)} from historical scheduling records for model agent training, enabling the model to have the capabilities of hypothesis semantic understanding, top_N model agent recall, and key parameter extraction.

[0080] From the Investigate system investigation node to the Fact fact child node, it relies on the specific implementation and actual execution of each model agent. In the model agent construction part, each model agent includes three types of sub-modules: "input processing", "business interface call", and "integrated output"; a complete model agent is composed of a chained concatenation of one or more sub-modules; the model agent construction part is implemented relying on manually written code. In the actual execution part, the list of key parameters is used as the output of the model agent and is automatically run to achieve.

[0081] Step S2. For the target problem in the photovoltaic business domain, repeatedly call the FHI logical reasoning basic model to obtain the FHI logical reasoning model. The method is as follows:

[0082] After establishing the FHI logical reasoning basic model that matches the target problem in the photovoltaic business domain, call the associated business knowledge and business logic, analyze each Fact fact sub-node, and determine whether the Fact fact sub-node is the bottommost node associated with the target problem in the photovoltaic business domain in the business logic. If so, end the re-call of the FHI logical reasoning basic model; if not, use the Fact fact sub-node as the Fact fact node, re-call the FHI logical reasoning basic model, and increase the level of the tree structure. If the loop continues, finally establish the FHI logical reasoning model;

[0083] As an implementation method, by continuously calling the FHI logical reasoning basic model, a logical reasoning tree FHI-TREE with continuously increasing levels is obtained. When the Hypothesize rational hypothesis is exhausted or a closed loop appears in the logical reasoning tree FHI-TREE or the set maximum iteration depth is reached, stop generating to obtain the FHI logical reasoning model.

[0084] Step S3. Convert the FHI logical reasoning model into an argument fact tree; combined with Figure 1 in the argument fact tree, there are only Fact fact nodes and the Fact fact sub-nodes at each level;

[0085] In this step, converting the FHI logical reasoning model into an argument fact tree is specifically as follows:

[0086] Delete each Hypothesize rational hypothesis node and Investigate system investigation node in the FHI logical reasoning model, and keep the level number and cascade relationship of each Fact fact sub-node unchanged to obtain the argument fact tree; among them, the level number and cascade relationship are called the argument derivation level relationship.

[0087] When specifically implemented, sequentially remove each Hypothesize rational hypothesis node and Investigate system investigation node in the FHI logical reasoning model, and construct a new connection between the original "grandparent node" and the original "grandson node". When only Fact fact nodes and Fact fact sub-nodes are left in the tree, it is recorded as the "argument fact tree".

[0088] Step S4: Analyze and process each Fact fact sub-node in the argument fact tree to obtain an embedding encoding embedded with the hierarchical features in the argument fact tree.

[0089] This step combines Figure 2 , specifically:

[0090] The Fact fact sub-node stores text-based argument facts and the argument derivation hierarchical relationship.

[0091] Segment the text-based argument facts. The text-based argument facts have M-dimensional segmentation, and each segmentation is transformed into a digital representation of a semantic meaning, thereby obtaining an M-dimensional text semantic vector v_semantic.

[0092] According to the text position information of each segmentation in the text-based argument facts, obtain an M-dimensional text position vector v_position through a combination of sine and cosine in different periods.

[0093] Relying on the argument fact tree, according to the argument derivation hierarchical relationship of the Fact fact sub-node, including the level and the order within the level, construct an initial hierarchical logic vector with dimension D, and then obtain a structure position vector v_structure with dimension M through a linear layer mapping.

[0094] Add the text semantic vector v_semantic, the text position vector v_position, and the structure position vector v_structure to obtain the embedding encoding embeding of the text-based argument facts.

[0095] Specifically, the argument fact tree contains two parts of information. Each node contains text-based argument facts, and the relationship between nodes contains the argument derivation hierarchical relationship.

[0096] The Transformer model uses the attention mechanism to capture long-range dependencies in sequence data, and abandons the recursive and convolutional structures in traditional recurrent neural networks (RNNs) and convolutional neural networks (CNNs). Since the Transformer model does not contain any cyclic structure, it cannot capture the order information of the elements in the sequence itself. Position encoding is added to the embedding encoding embeding so that the Transformer model can understand the position of the elements in the sequence.

[0097] Traditionally, the process of converting text tokenization embeding into Embedding encoded embedding is as follows: First, the text is segmented to obtain tokens; then, the tokens are mapped into multi-dimensional vectors; at the same time, a position information vector is generated according to the formal position of the tokens in the text; finally, the two are superimposed as the final embedded encoding.

[0098] In the present invention, in combination with the argument fact tree derived from the previous model, both the text position information and the logical hierarchical relationship in the derivation process are retained in the process of generating the position information vector.

[0099] Step S5, input the embedding encodings with hierarchical cascade relationships into the transformer module after sorting them hierarchically, and the transformer module outputs the analysis result for the target problem in the photovoltaic business field.

[0100] In this step, in combination with Figure 2 , the transformer module includes an encoder, a decoder, a linear layer, and a Softmax layer;

[0101] The embedding encodings with hierarchical cascade relationships are sorted hierarchically to form an embedding encoding sequence, and the embedding encoding sequence is input into the encoder;

[0102] The encoder captures the global dependencies between the elements in the input embedding encoding sequence and generates a series of context-aware representations, which are input into the decoder;

[0103] The decoder is used to generate an output sequence based on the context-aware representation; further, when generating the output sequence, the decoder considers both the context of the input sequence and the output sequence that has already been generated.

[0104] The linear layer is used to map the output sequence of the decoder to a vector;

[0105] The Softmax layer is used to convert the vector output by the linear layer into a probability distribution.

[0106] Thus, the inference and analysis of the photovoltaic business based on the argument fact tree are realized.

[0107] A method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework provided by the present invention has the following advantages:

[0108] 1. The present invention realizes a vertical domain model solution for the photovoltaic field from semantic understanding to logical derivation analysis combined with business facts;

[0109] 2. The present invention relies on the logical reasoning framework of Fact - Hypothesis - Investigation (FHI), and through the cyclic scheduling of semantic understanding and model agents, obtains an argument fact tree.

[0110] 3. Utilize the hierarchical structure information of the argument fact tree to enhance the logical relationship in the positional encoding, improve the understanding depth of the transformer - like model for specific photovoltaic business problems, and achieve reasoning and analysis.

[0111] The present invention realizes a model agent from semantic understanding to actual analysis and diagnosis, makes the model reasoning process conform to business logic, enhances the understanding depth of the transformer model for photovoltaic business, and thus improves the accuracy and reliability of the prediction and analysis results of the large - language model.

[0112] The following combines Figure 3 and Figure 4 , taking the specific analysis process of "the equipment health status of the photovoltaic inverter at Station A" as an example, to introduce a specific embodiment:

[0113] In this embodiment:

[0114] The basic root node is the Fact node "the equipment health status of the photovoltaic inverter at Station A";

[0115] By making a single call to the FHI logical reasoning basic model, the following logical reasoning tree is generated:

[0116] Relying on the "method for improving the efficiency of large models enhanced by knowledge graph and vector retrieval", four Hypothesize nodes are generated, namely

H1 - 1

H1 - 2

H1 - 3

H1 - 4

[0117] For the Hypothesize node

H1 - 1

I1 - 1

F1 - 1

[0118]

H1-2

I1-2

F1-2

[0119]

H1-3

I1-3

[0120]

F1-3

[0121]

H1-4

I1-4

F1-4

[0122] Based on the business logic, analyze the above Fact fact sub-nodes. The Fact fact sub-nodes

F1-2

F1-3

F1-1

F1-4

[0123] Therefore, after one generation of FHI derivation, the Fact fact sub-node

F1-1

F1-1-1

F1-1-2

[0124] The Fact fact child node

F1-4

F1-4-1

[0125] Based on business logic analysis, the Fact fact child nodes

F1-1-1

F1-1-2

F1-4-1

[0126] Finally obtain Figure 3 The FHI logical reasoning model shown; Figure 3 The FHI logical reasoning model shown is transformed into Figure 4 The argument fact tree shown.

[0127] Utilize the hierarchical structure information of the argument fact tree to enhance the logical connection of the positional encoding, improve the understanding depth of the transformer - like model for specific photovoltaic business problems, and achieve reasoning and analysis.

[0128] In this solution, combined with the argument fact tree derived from the previous model, during the process of generating the position information vector, both the text information position and the logical hierarchical relationship in the derivation process are retained. The specific process is as follows: Each word segmentation in the text is transformed into a digital representation of semantics, obtaining an M - dimensional text semantic vector denoted as v_semantic; According to the text position information, an M - dimensional text position vector is obtained through the combination of sine and cosine in different periods, denoted as v_position; Relying on the argument fact tree, according to the hierarchical level and in - layer order of each node, an initial hierarchical logical vector with dimension D is formed, and then through linear layer mapping, a structural position vector with dimension M is obtained, denoted as v_structure; Finally, these three M - dimensional vectors are added together to obtain the embedding encoding. In this embodiment, M is 512 and D is 2.

[0129] Subsequently, it passes through the encoder (5 layers), decoder (5 layers), linear layer, and Softmax layer in sequence, and finally outputs the probability distribution. Finally, the reasoning and analysis of the photovoltaic business based on the argument fact tree are realized.

[0130] In this embodiment, for "the equipment health status of the photovoltaic inverter at Substation A", the model outputs the following diagnostic results:

[0131] Based on the provided information, perform the following reasoning to evaluate the health status of the photovoltaic inverter equipment at Substation A:

[0132] 1. **Equipment age and failure rate**:

[0133] The inverters in Batch B01 (NB001 and NB002) have been in operation for 2.5 years, with an accumulated failure rate of 5% and a failure rate in the past year of 1%. This indicates that these devices are relatively new, with good overall operating conditions and a low failure rate.

[0134] The inverter in Batch B02 (NB003) has been in operation for 4.5 years, with an accumulated failure rate of 35% and a failure rate in the past year of 15%. This indicates that these devices are older, with a significantly higher failure rate than Batch B01, and there may be problems with aging or design defects.

[0135] 2. **Environmental factors**:

[0136] Since Station A is located in a monsoon climate region and there were rainstorm and high temperature warnings in July, these extreme weather conditions may exert additional stress on the inverter equipment and increase the risk of failure.

[0137] 3. **Device alarm information**:

[0138] The voltage out-of-bounds alarm on July 15, 2024 may indicate problems with the inverter in coping with grid fluctuations or problems with the grid itself.

[0139] The common connection point imbalance and flicker alarm may mean that there are problems with the inverter in terms of output power quality, which may be caused by faults in the inverter itself or external grid problems.

[0140] The abnormal reset of the disconnector may indicate problems with the maintenance or operation of the device and requires further inspection.

[0141] 4. **Grid ownership and group influence**:

[0142] Station A and Station A-2 belong to the same power generation group, and there has also been an imbalance alarm at Station A-2. This may indicate problems with the grid stability within the group, affecting the normal operation of the inverters.

[0143] Based on the above information, the following conclusions are deduced:

[0144] **Inverters in Batch B01 (NB001 and NB002)**: Currently in good health, but continuous monitoring is required, especially under extreme weather conditions, to prevent potential failures.

[0145] **Inverter in Batch B02 (NB003)**: Poor health, high failure rate. It may be necessary to immediately conduct a detailed inspection and maintenance, or consider early replacement to avoid possible long-term downtime.

[0146] **Overall power grid condition**: There may be stability issues, and it is necessary to coordinate with the power grid operator to check the power grid status within the group to reduce the impact on the inverter.

[0147] **Maintenance and monitoring**: It is recommended to strengthen the daily maintenance and monitoring of the equipment, especially during extreme weather events, as well as the timely response and handling of alarm information.

[0148] **Preventive measures**: Considering environmental factors and equipment age, a preventive maintenance plan should be formulated, including regular inspections and replacement of vulnerable parts, to extend the equipment life and maintain efficient operation.

[0149] The present invention provides a method for a large model to analyze the photovoltaic business field based on the FHI logical derivation framework, enabling the large language model in the photovoltaic business field to not only understand photovoltaic professional knowledge, but also, based on the reasoning logic chain, schedule the business interfaces of the information system to achieve long-link reasoning analysis, making the reasoning process of the large language model conform to the business logic, enhancing the understanding depth of the large language model for the photovoltaic business, and thus improving the accuracy and reliability of the prediction and analysis results of the large language model.

[0150] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present invention, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework, characterized in that: The following steps are involved: Step S1, establishing a basic model of FHI logic reasoning; the basic model of FHI logic reasoning is a tree structure, including a Fact node, a Hypothesize node, an Investigate node and a Fact sub-node connected hierarchically from top to bottom; Step S2, for the target problem in the photovoltaic business field, repeatedly call the FHI logical reasoning basic model to obtain the FHI logical reasoning model; the method is: After the FHI logic reasoning basic model matching the target problem in the photovoltaic business field is established, the associated business knowledge and business logic are called, and each of the Fact fact sub-nodes is analyzed to determine whether the Fact fact sub-node is the bottom-level node associated with the target problem in the photovoltaic business field in the business logic. If so, the re-calling of the FHI logic reasoning basic model is terminated; if not, the Fact fact sub-node is used as the Fact fact node, and the FHI logic reasoning basic model is called again to increase the level of the tree structure. If the cycle is continuous, the FHI logic reasoning model is finally established; Step S3, converting the FHI logical reasoning model into an argument fact tree; the argument fact tree only has Fact fact nodes and Fact fact sub-nodes of each level; Step S4, analyzing and processing each of the Fact sub-nodes in the argument fact tree to obtain an embedding code embedded with the hierarchical features in the argument fact tree; Step S4 is specifically as follows: The Fact sub-node stores text-based argument facts and argument derivation hierarchical relationships; Segmenting the text-like argument facts, wherein the text-like argument facts have M-dimensional segmentations, and each segmentation is converted into a digital representation of semantics, thereby obtaining an M-dimensional text semantic vector v_semantic; According to the text position information of each word in the text class argument fact, an M-dimensional text position vector v_position is obtained by combining sine and cosine with different periods; Relying on the argument fact tree, the hierarchical relationship is deduced according to the arguments of the Fact fact sub-node, including the hierarchical level and the order within the level, to form an initial hierarchical logic vector with a dimension of D, and then a structural position vector v_structure with a dimension of M is obtained through linear layer mapping; Add the text semantic vector v_semantic, the text position vector v_position and the structure position vector v_structure to obtain the embedding encoding of the text class argument fact; Step S5, embedding and encoding each of the embeddings having a hierarchical cascade relationship, sorting them by level, and inputting them into a transformer module, and the transformer module outputs an analysis result for the target problem in the photovoltaic business field.

2. According to claim 1, a method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework is characterized in that: In step S1, the process of establishing the FHI logical reasoning basic model is as follows: Step S1.1, storing the semantic description information of the target problem into the Fact node; Step S1.2, calling the business knowledge and business logic associated with the target problem, analyzing and reasoning the semantic description information of the target problem stored in the Fact node, obtaining N rational hypothesis texts associated with the target problem, and for each rational hypothesis text, generating a Hypothesize rational hypothesis node as a child node of the Fact node; Step S1.3, for each of the Hypothesize rational hypothesis nodes, obtain M associated Investigate system investigation nodes and use them as child nodes of the Hypothesize rational hypothesis node; each of the Investigate system investigation nodes is bound to a model agent, and by executing the model agent, text-based argument facts are obtained; Step S1.4, the text-type argument facts of each Investigate system investigation node are stored in the Fact fact child node which is a child node of the Investigate system investigation node; From top to bottom, the FHI logical reasoning basic model is established to obtain a tree structure formed by Fact nodes, Hypothesize nodes, Investigate nodes and Fact sub-nodes.

3. According to claim 2, a method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework is characterized in that: The business knowledge and business logic are a large language model enhanced based on knowledge graph and vector retrieval.

4. According to claim 2, a method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework is characterized in that: The model agent bound to the Investigate system investigation node is a model agent obtained by training with historical data; The training method is: The model agent is trained using a key-value pair in the form of {"text":(model agent name, key parameter list)}, so that the model agent has the capabilities of semantic understanding, top_N model agent recall, and key parameter extraction.

5. According to claim 4, a method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework is characterized in that: The model agent comprises an input processing submodule, a business interface calling submodule and an integrated output submodule; The input processing submodule is used to perform semantic understanding on the rational hypothesis text input from the Hypothesize rational hypothesis node; The business interface calling submodule is used to call the associated business database or business data table according to the semantic understanding of the input processing submodule, and query and obtain key parameters associated with the semantic understanding as query results; The integration output submodule is used to form a key parameter table with the multiple key parameters obtained by the business interface calling submodule, and output it to the Fact fact subnode.

6. The method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework according to claim 1 is characterized in that: In step S3, the FHI logical reasoning model is converted into an argument fact tree, specifically: Delete each Hypothesize rational hypothesis node and Investigate system investigation node in the FHI logical reasoning model, keep the number of levels and cascade relationship of each Fact fact sub-node unchanged, and obtain the argument fact tree; wherein the number of levels and cascade relationship are called argument derivation level relationship.

7. The method for realizing large-scale model analysis of photovoltaic business field based on FHI logic deduction framework according to claim 1 is characterized in that: Step S5 is specifically as follows: The transformer module includes an encoder, a decoder, a linear layer and a Softmax layer; Embedding the embedding codes in a hierarchical cascade relationship, sorting them by level to form an embedding code sequence, and inputting the embedding code sequence into the encoder; The encoder captures the global dependencies between the elements in the embedding coding sequence input, generates a series of context-aware representations, and inputs them to the decoder; The decoder is used to generate an output sequence based on the context-aware representation; The linear layer is used to map the output sequence of the decoder to a vector; The Softmax layer is used to convert the vector output by the linear layer into a probability distribution.

Images