Report generation system

Abstract

The invention discloses a report generation system, and the system comprises an import analysis layer which is used for importing a process package corresponding to a green chemical project, and analyzing the process package to obtain production information; the data acquisition layer is used for acquiring evaluation information matched with the production information; the data processing layer is used for compensating the evaluation information and determining influence weights corresponding to different production scenes; wherein the influence weight corresponds to the fluctuation intensity of different production scenes on the evaluation information; and the report generation layer is used for generating evaluation reports of the green chemical project in different production scenes based on the compensated evaluation information, the influence weight and the production information. According to the method and the device, the corresponding evaluation reports of the green chemical project in different production scenes are generated, different factor conditions can be fully considered only by providing the process package of the green chemical project, and finally the evaluation report with an actual production guidance value is generated.

Description

Technical Field

[0001] The present application relates to the field of manufacturing technology, and in particular to a report generation system. Background Art

[0002] Chemical projects such as hydrogen and methanol production can become green chemical projects by using green electricity. Evaluation reports based on production and cost information are needed to guide the actual production of green chemical projects.

[0003] Because green chemical projects involve a wide range of equipment, raw materials, and energy resources, the corresponding production and cost information is extensive, and some of this information is dynamic. Collecting this information requires significant time and effort. Furthermore, the reliability of the collected information cannot be guaranteed. This results in evaluation reports losing their guiding significance and hindering the implementation of green chemical projects.

[0004] In summary, how to obtain an evaluation report with production guidance value is a technical problem that technical personnel in this field urgently need to solve. Summary of the Invention

[0005] The purpose of this application is to provide a report generation system that can obtain evaluation reports with production guidance value, which is conducive to the production implementation of green chemical projects.

[0006] To solve the above technical problems, this application provides the following technical solutions:

[0007] A report generating system comprising:

[0008] The import and parsing layer is used to import the process package corresponding to the green chemical project and parse the process package to obtain production information;

[0009] A data collection layer for acquiring evaluation information matching the production information;

[0010] A data processing layer, configured to compensate the evaluation information and determine the impact weights corresponding to different production scenarios; wherein the impact weights correspond to the intensity of fluctuations in the evaluation information due to different production scenarios;

[0011] The report generation layer is used to generate an evaluation report of the green chemical project in different production scenarios based on the compensated evaluation information, the impact weight and the production information.

[0012] Preferably, the parsing layer is imported, including:

[0013] A parsing sublayer, configured to extract equipment information, raw material information, and energy consumption information from the process package based on semantic parsing and in combination with a parameter extraction rule library, and determine the equipment information, raw material information, and energy consumption information as the production information;

[0014] The derivation sublayer is used to establish a mapping relationship between the production information and the information channel; wherein the information channel is a channel for obtaining the evaluation information.

[0015] Preferably, the data collection layer includes:

[0016] The information capture layer is used to obtain evaluation information corresponding to equipment and raw materials from web pages;

[0017] The interface docking layer is used to connect to the energy supply channel interface and obtain the corresponding energy assessment information;

[0018] The integrated communication layer is used to obtain real-time evaluation information based on artificial intelligence technology and combined with multiple communication technologies.

[0019] Preferably, the integrated communication layer includes:

[0020] The email communication layer is used to determine the email workflow schedule, generate and send target emails, receive and parse reply emails to the target emails, and obtain real-time evaluation information;

[0021] The online program layer is used to simulate manual operation, send a standard evaluation information acquisition template to the target object, and determine the real-time evaluation information based on the response information fed back by the target object;

[0022] The telephone communication layer is used to obtain real-time assessment information in a conversational manner by converting text to speech.

[0023] Preferably, the data collection layer further includes:

[0024] The evaluation and negotiation layer is used to verify the evaluation information without the data being out of the domain.

[0025] Preferably, the data processing layer includes:

[0026] For production factors that are obtained multiple times, obtain the corresponding evaluation information in batches and identify the periodic evaluation information; production factors include raw materials, equipment, and energy consumption;

[0027] After data collection fails, use periodic evaluation information to predict current evaluation information;

[0028] When data collection fails and information prediction fails, substitute evaluation information is obtained through knowledge graph completion inference.

[0029] Preferably, the data processing layer further includes:

[0030] When the missing assessment information reaches the specified proportion, the acquired assessment information will be anomaly checked, and the missing assessment information will be simulated and supplemented within the calculated reasonable range.

[0031] Preferably, the data processing layer includes:

[0032] According to the fluctuation threshold corresponding to the assessment information, the emergency fluctuation weight is determined for the assessment information that exceeds the fluctuation threshold.

[0033] Preferably, the fluctuation threshold corresponds to the raw material.

[0034] Preferably, the data processing layer further includes:

[0035] Obtain deduction information, and use the deduction information to deduct and compensate the evaluation information.

[0036] The system provided in the embodiment of the present application is applied, and an analysis layer is imported to import the process package corresponding to the green chemical project and parse the process package to obtain production information; a data collection layer is used to obtain evaluation information matching the production information; a data processing layer is used to compensate the evaluation information and determine the impact weights corresponding to different production scenarios; wherein the impact weights correspond to the fluctuation intensity of the evaluation information for different production scenarios; a report generation layer is used to generate evaluation reports for green chemical projects in different production scenarios based on the compensated evaluation information, impact weights and production information.

[0037] In this application, the production information can be clarified by parsing the process package corresponding to the green chemical project. After clarifying the production information, the evaluation information corresponding to the production information is obtained through the data acquisition layer. Taking into account the possibility that the evaluation information may be missing, in order to ensure the accuracy of the final evaluation report, the evaluation information can be compensated at the data processing layer. In addition, under different production scenarios, the evaluation information will fluctuate, and the corresponding impact weights under different scenarios can be determined. The impact weights reflect the intensity of the fluctuation of the evaluation information under different production scenarios. Finally, at the report generation layer, based on the compensated evaluation information, impact weights and production information, an evaluation report of the green chemical project under different production scenarios can be generated. That is to say, in this application, to generate the corresponding evaluation report of the green chemical project under different production scenarios, it is only necessary to provide the process package of the green chemical project to fully consider different factors and finally generate an evaluation report with actual production guidance value. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1This is a structural diagram of a report generation system in an embodiment of the present application;

[0040] Figure 2 This is a schematic diagram of a green chemical project in an embodiment of this application. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the embodiments described are only a part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present application.

[0042] Please refer to Figure 1 , Figure 1 This is a structural diagram of a report generation system in an embodiment of the present application, which includes:

[0043] The import and parsing layer 101 is used to import the process package corresponding to the green chemical project and parse the process package to obtain production information;

[0044] Data collection layer 102, used to obtain evaluation information that matches production information;

[0045] The data processing layer 103 is used to compensate the evaluation information and determine the impact weights corresponding to different production scenarios; wherein the impact weights correspond to the intensity of fluctuations in the evaluation information due to different production scenarios;

[0046] The report generation layer 104 is used to generate an evaluation report of the green chemical project in different production scenarios based on the compensated evaluation information, impact weights and production information.

[0047] Green chemical projects refer to chemical projects that adhere to the principles of green chemistry and sustainable development and employ advanced technologies and processes in the design, production, and use of chemical products to reduce or eliminate environmental pollution and resource waste. For example, green electricity-based hydrogen and methanol production.

[0048] It includes the instructions for the corresponding process of the overall green chemical project, the output of the process equipment list of the required related equipment (Aspen / HYSYS, etc.), standardized Excel templates (such as equipment location / material balance sheet), and PDF process flow chart (OCR identification of key parameters).

[0049] Production information can specifically include production equipment, production raw materials, energy required for production, and other production-related information required for green chemical projects. For example, a green chemical project that uses direct biomass combustion coupled with wind, solar, and green hydrogen to produce methanol from carbon dioxide could include production information such as gasification equipment (e.g., a gasifier with a design pressure of 3.2 MPa, a volume of 50 m³, and a calorific value compensation range of 12-18 MJ / kg), methanol synthesis equipment (e.g., a methanol synthesis tower with a design pressure of 5.0 MPa and a steam drum operating pressure of 4.3 MPa), and distillation equipment (e.g., a methanol distillation tower with a design pressure of 2.5 MPa and made of 316L stainless steel); raw materials such as white pellets at 2.5 tons per ton of methanol, raw straw at 3.0 tons per ton of methanol, reagents such as NaoH at 15 kg per ton of methanol, catalysts such as PDS catalyst at 0.1 kg per ton of methanol, and fuel gas supplementation at 120 Nm³ per ton of methanol when the straw ash content exceeds 20%; and energy requirements such as electricity at 800 kWh per ton of methanol. Among them, A unit / unit B means that the production of each unit of B requires the consumption of each unit of A. For example, 3.0 tons of raw straw / ton of methanol means that 3.0 tons of raw straw is needed to produce 1 ton of methanol; 120Nm³ of fuel gas / ton of methanol means that 120 Nm³ of fuel gas is needed to produce one ton of methanol.

[0050] The evaluation information may include evaluation contents such as the value of the object corresponding to the production information, the difficulty of obtaining the object, or the time cost of obtaining the object.

[0051] Considering that the acquired assessment information may contain missing values ​​or inaccurate values, compensation can be applied to the assessment information, such as compensating for missing values ​​or correcting inaccurate values. Furthermore, assessment information may fluctuate across different scenarios. To obtain corresponding assessment reports for multiple scenarios, weights can be set for each scenario. Ultimately, an assessment report is generated based on the production information, the compensated assessment information, and the weights corresponding to each scenario, which can be used to guide production. This assessment report can assess the production difficulty, production costs, or production value, based on the required production information.

[0052] Among them, different scenarios can specifically include baseline scenarios. Above the baseline scenarios, there are some unexpected scenarios, such as severe weather during production, long-term power outages and work stoppages during production, etc.

[0053] In a specific embodiment of the present application, importing the parsing layer includes:

[0054] The parsing sublayer is used to extract equipment information, raw material information, and energy consumption information from the process package based on semantic parsing and combined with the parameter extraction rule library, and determine the equipment information, raw material information, and energy consumption information as production information;

[0055] The derivation sublayer is used to establish a mapping relationship between production information and information channels; among them, information channels are channels for obtaining evaluation information.

[0056] For example: through format verification and the establishment of a parameter extraction rule library, parameters are imported based on semantic analysis of the process package to extract the equipment required for production (different products and production processes correspond to different production equipment), the raw materials required for production (different products correspond to different raw materials), and the energy consumption required for production (such as electricity).

[0057] Among them, a parameter extraction rule library is established. For example, when intelligently analyzing the process package, the raw materials, reagents, catalyst types, and consumption are obtained, the CAS number in the bill of materials is matched by extraction logic, and the equipment name is obtained. The equipment name, model, and requirements of the process equipment list are extracted by extraction logic. If they are not clearly stated in the list, the equipment requirements are further matched according to the equipment name, such as obtaining the reaction pressure requirements, and the pressure controller setting value in the PID diagram is extracted by extraction logic.

[0058] Furthermore, a knowledge graph can be built to embed intelligent deduction, forming an association mapping between the equipment, raw materials, energy consumption and supply channels in the cost elements. For example, a list of compliant suppliers can be obtained, and the equipment, raw materials and energy consumption in the cost elements can be matched and associated with the products supplied by qualified suppliers. If there is no match, the equipment and raw materials can be automatically searched and replaced with existing supply products based on the properties of the equipment and raw materials. For example, 316L stainless steel can be replaced with 2205 duplex steel, or catalysts with the same function can be replaced. At the same time, the process engineer can be prompted to confirm the feasibility. When the replacement is confirmed, a mark is added for comparison with external supply channels. Products that have no alternative supply channels are marked as external supply channels for consultation. External supply channels refer to other supply channels that are not on the list of qualified suppliers and can obtain evaluation information. For example, evaluation information can be obtained through online search and communication, historical business card / flyer email communication, industry B2B platforms, etc. Furthermore, an association is established between the equipment, raw materials and energy consumption in the cost elements and their possible suppliers and the fluctuation pattern of the evaluation value.

[0059] In order to facilitate the acquisition of evaluation information, it is also possible to embed intelligent inference with the help of knowledge graphs, such as establishing a cost factor mapping to map the relationship between equipment, raw materials, energy consumption and supply channels (a channel for obtaining evaluation information) in cost factors.

[0060] In a specific embodiment of the present application, the data collection layer includes:

[0061] The information capture layer is used to obtain evaluation information corresponding to equipment and raw materials from web pages;

[0062] The interface docking layer is used to connect to the energy supply channel interface and obtain the corresponding energy assessment information;

[0063] The integrated communication layer is used to obtain real-time evaluation information based on artificial intelligence technology and combined with multiple communication technologies.

[0064] Among them, the integrated communication layer includes:

[0065] The email communication layer is used to determine the email workflow scheduling, generate and send target emails, receive and parse reply emails to target emails, and obtain real-time evaluation information;

[0066] The online program layer is used to simulate manual operations, send standard evaluation information acquisition templates to the target object, and determine the real-time evaluation information based on the response information fed back by the target object;

[0067] The telephone communication layer is used to obtain real-time assessment information in a conversational manner by converting text to speech.

[0068] The data collection layer also includes:

[0069] The evaluation and consultation layer is used to verify the evaluation information without leaving the domain.

[0070] Since the production factors corresponding to production information are diverse, when obtaining the evaluation information corresponding to the production information, different evaluation information acquisition methods can be adopted according to different production factors.

[0071] The data collection layer can be based on multimodal evaluation and collection. Specifically, for the collection of equipment and raw material evaluation information, an intelligent crawler based on NLP (natural language processing) can be deployed to support adaptive parsing of supply channel official website pages against DOM structure changes, intelligent docking with industry B2B platform APIs, and automatic extraction of email / IM (Instant Messaging) communication records, integrated with OCR recognition of unstructured evaluation forms, thereby obtaining equipment and raw material evaluation information. For the collection of energy evaluation information, multiple energy supply channel interfaces can be constructed, including virtual power plant API docking, digital twin wind and / or photovoltaic power station API docking, and electricity market API docking, to obtain energy evaluation information.

[0072] AI (artificial intelligence) multi-channel intelligent consultation, based on AI and supply channels for automated intelligent consultation, supports intelligent consultation via email, IM instant messaging, and telephone to obtain corresponding evaluation information.

[0073] Among them, the intelligent crawler adopts BERT and dynamic XPath parsing, and the industry B2B platform API interface includes the open interface of 1688 / HC360; email intelligent consultation, email consultation workflow scheduling through LangChain, intelligent generation of consultation emails through GPT-4, monitoring of sending and receiving functions through SMTP protocol, automatic tagging of supplier replies by BERT-based classifiers, and semantic analysis of attachment evaluation information sheets.

[0074] IM instant intelligent consultation simulates manual operations through RPA and combines with the enterprise WeChat / DingTalk open platform to carry out the process of adding supplier contacts, sending standard consultation templates, asking personalized questions based on historical records, intelligent negotiation, and automatically recording the evaluation value of commitments to the CRM. At the same time, anti-crawler strategies are added to simulate typing speeds at random intervals of 50-300ms, and emoticons are added at intervals. Telephone intelligent consultation uses ASR voice recognition, combined with pre-trained language models to analyze voiceprints and text sentiment, and speech templates, and uses TTS text-to-speech to determine the intelligent consultation evaluation value.

[0075] Privacy protection consulting based on federated learning can also be provided. Each supplier can locally deploy lightweight AI agents to complete evaluation information consulting without leaving the data domain.

[0076] In a specific embodiment of the present application, the data processing layer includes:

[0077] For production factors that are obtained multiple times, obtain the corresponding evaluation information in batches and identify the periodic evaluation information; production factors include raw materials, equipment, and energy consumption;

[0078] After data collection fails, use periodic evaluation information to predict current evaluation information;

[0079] When data collection fails and information prediction fails, substitute evaluation information is obtained through knowledge graph completion inference.

[0080] The data processing layer also includes:

[0081] When the missing assessment information reaches the specified proportion, the acquired assessment information will be anomaly checked, and the missing assessment information will be simulated and supplemented within the calculated reasonable range.

[0082] Specifically, for compensation of non-one-time purchase evaluation information, first mark the raw materials, equipment, and energy consumption of non-one-time purchases, then obtain the corresponding continuous supply evaluation information from the supply channel through the data collection layer, and identify the periodic characteristics of the evaluation values ​​in the corresponding evaluation information; when the collection of continuous supply evaluation information fails, the evaluation data with periodic characteristics starts the ARIMA-based time series or Holt-Winters forecast to determine the evaluation value; when the collection of non-periodic evaluation information and one-time purchase evaluation information fails, the knowledge graph is completed and inferred, and the associated substitute supply channel evaluation information or historical evaluation information is used for weighted calculation.

[0083] Compensation for the collection of assessment information: prediction of the assessment value interval when 20%-30% of the data is missing, filtering of abnormal assessment values ​​through the IsolationForest algorithm, calculation of the current reasonable assessment value interval using Huber regression, and Monte Carlo simulation to supplement the missing data.

[0084] The emergency fluctuation weight can also be determined for the assessment information that exceeds the fluctuation threshold according to the fluctuation threshold corresponding to the assessment information.

[0085] Among them, the fluctuation threshold corresponds to the raw material, and the specific size of the fluctuation threshold can be preset and adjusted according to actual conditions, which are not listed here one by one.

[0086] The supplementary cost calculation weights in the intelligent processing layer include:

[0087] Emergency fluctuation processing sets the fluctuation threshold of the corresponding characteristics according to the characteristics of the assessment information. If the fluctuation threshold is exceeded, the weight calculation of the emergency fluctuation is triggered:

[0088] The weight calculation of emergency fluctuations adopts the following modified algorithm:

[0089] , where Ccorrection is the corrected cost under the emergency fluctuation scenario; Cbenchmark is the benchmark cost; ki is the weight coefficient of the i-th emergency fluctuation scenario, reflecting the sensitivity of the fluctuation scenario to the cost; Si is the intensity coefficient of the i-th emergency fluctuation scenario, reflecting the intensity of the fluctuation.

[0090] Among them, the value of K is based on historical data statistics, business rules and risk preference constraints, mathematical model quantitative calculation and optimization iteration; the value of S is quantified through threshold comparison.

[0091] Compensation for raw material quality fluctuations, with weighted compensation set according to raw material standards.

[0092] In one embodiment of the present application, the data processing layer is further configured to obtain offset information and use the offset information to offset the assessment information. The offset information, such as carbon offset compensation, is used to offset the initially acquired assessment information in proportion to the captured CO2 based on the carbon trading assessment information.

[0093] The system provided in the embodiment of the present application is applied, and an analysis layer is imported to import the process package corresponding to the green chemical project and parse the process package to obtain production information; a data collection layer is used to obtain evaluation information matching the production information; a data processing layer is used to compensate the evaluation information and determine the impact weights corresponding to different production scenarios; wherein the impact weights correspond to the fluctuation intensity of the evaluation information for different production scenarios; a report generation layer is used to generate evaluation reports for green chemical projects in different production scenarios based on the compensated evaluation information, impact weights and production information.

[0094] In this application, the production information can be clarified by parsing the process package corresponding to the green chemical project. After clarifying the production information, the evaluation information corresponding to the production information is obtained through the data acquisition layer. Taking into account the possibility that the evaluation information may be missing, in order to ensure the accuracy of the final evaluation report, the evaluation information can be compensated at the data processing layer. In addition, under different production scenarios, the evaluation information will fluctuate, and the corresponding impact weights under different scenarios can be determined. The impact weights reflect the intensity of the fluctuation of the evaluation information under different production scenarios. Finally, at the report generation layer, based on the compensated evaluation information, impact weights and production information, an evaluation report of the green chemical project under different production scenarios can be generated. That is to say, in this application, to generate the corresponding evaluation report of the green chemical project under different production scenarios, it is only necessary to provide the process package of the green chemical project to fully consider different factors and finally generate an evaluation report with actual production guidance value.

[0095] To facilitate those skilled in the art to better understand and implement the report generation system provided in the embodiment of the present application, the report generation system is described in detail below by taking the generation of a cost assessment report implemented on the report generation system as an example.

[0096] Please refer to Figure 2 , Figure 2 This is a schematic diagram of a green chemical project in an embodiment of this application. This green chemical project involves green electricity, such as biomass power generation, photovoltaic power generation, and wind power generation. This green electricity can be directly used in green chemical production of green hydrogen and green alcohol. Green electricity can also be directly fed into the power grid (for commercial and residential use). The production of green hydrogen and green alcohol can also utilize green electricity from distributed power networks. Green chemical projects can also be implemented by integrating carbon capture and storage.

[0097] The specific implementation of generating multi-scenario cost calculation reports within this report generation system is as follows. This report generation system (in this scenario, the green chemical dynamic cost calculation system) includes an AI-driven import layer (import and parsing layer), a data collection layer, an intelligent processing layer (i.e., data processing layer), and an application layer (i.e., report generation layer). The import layer is used to import process package parameters and parse the data required for cost calculation; the data collection layer is used to collect multi-modal prices for the data required for cost calculation; the intelligent processing layer is used to compensate for price collection and supplement the cost calculation impact weights; and the application layer is used to generate multi-scenario cost calculation reports that include at least a baseline scenario and an extreme volatility scenario.

[0098] The import layer includes intelligent process package parsing: Through format verification and the establishment of a parameter extraction rule base, process package import parameters are semantically parsed to extract cost factors. Furthermore, a knowledge graph is built to embed intelligent inference, mapping the cost factors of equipment, raw materials, and energy consumption with supply channels. For example, a list of qualified suppliers is obtained and the equipment, raw materials, and energy consumption in the cost factors are matched and associated with products supplied by qualified suppliers. Unmatched items are automatically searched and replaced with existing supply products based on the nature of the equipment and raw materials. In other words, the cost factors of equipment, raw materials, and energy consumption are associated with their possible suppliers and price fluctuation patterns.

[0099] Multimodal price collection at the data collection layer includes: equipment, raw material price collection, and energy price collection.

[0100] Equipment and raw material price collection: Deploy an intelligent crawler based on NLP (natural language processing) to support adaptive parsing of price pages on supply channel websites against DOM structure changes, intelligent integration with industry B2B platform APIs, automatic extraction of email / IM inquiry records, and integrated OCR recognition of unstructured quotations to automatically extract key price information. For example, identify the digits and units in the latest quote: ￥5,600 / ton.

[0101] Taking gasification equipment quotes as an example: the intelligent crawler uses BERT and dynamic XPath parsing operations to parse the DOM structure of the supplier's official website (XPath: / / div[@class="price"] / span); the BERT model understands the chaotic price information on the webpage, accurately identifies the real quote, and avoids missing the promotional price with a horizontal line through the original price, and locates the XPath webpage data.

[0102] Resisting DOM structural changes and revisions: For example, before the revision of a supplier's official website, the price was on the right side of the page, but after the revision, it moved to the middle. When the original XPath becomes invalid, it automatically attempts to switch to adjacent text positioning (such as wholesale price: the last number); the industry B2B platform API interface includes open interfaces such as 1688 / HC360, and directly obtains data through official dialogues with platforms such as 1688.

[0103] Email Inquiry: LangChain is used to schedule the email inquiry workflow, generate inquiry emails using a large language model or a pre-trained language model, monitor the sending and receiving functions through the SMTP protocol, automatically mark supplier replies with a BERT-based classifier, and perform semantic analysis of attached quotations.

[0104] Python

[0105] # LangChain schedules and generates inquiry emails

[0106] prompt = "Generate inquiry email: Gasification equipment, pressure 3.2MPa / volume 50m³, calorific value compensation range 12-18MJ / kg, CIF price including tax"

[0107] email = gpt4.generate(prompt) # Output professional inquiry content Output example: "Dear Manager XX: Gasification equipment, pressure 3.2MPa / volume 50m³, calorific value compensation range 12-18MJ / kg, please quote 100 sets, CIF (x port) including 13% VAT. Tiered quotes are also available for orders of 500 sets or more. I look forward to your reply!"

[0108] Inbox intelligent monitoring uses a BERT-based classifier to automatically mark supplier replies (priority sorting): urgent (no reply within 24 hours) > quotation with attachments > regular reply > spam. Attachment parsing: PDF / image quotations are automatically converted to Excel using the Adobe PDFExtract API.

[0109] IM instant intelligent inquiry uses RPA to simulate manual operations. Combined with the official interface provided by the instant messaging open platform API, it allows RPA robots to obtain chat lists, send messages, read files (such as quotation attachments), and perform other operations. It can then add supplier contacts, determine whether it is the first communication, send a standard inquiry template for the first communication, and ask personalized questions based on historical records for multiple communications. It uses AI-based intelligent negotiation and automatically records the promised price in the CRM process. At the same time, it adds anti-crawler strategies, simulates typing speed at random intervals of 50-300ms, and appends emoticons at intervals. Alternative solution when there is no API: UI automation: RPA directly simulates manual operations (recognizing buttons and input boxes on the screen). For example: automatically open corporate WeChat → click "Address Book" → search for contacts → enter and send a message.

[0110] Telephone intelligent inquiry uses ASR speech recognition, combined with GPT-4 voiceprint and text sentiment analysis, speech templates, and TTS text-to-speech to conduct intelligent inquiry and negotiation. Example: Voice interaction pipeline, Python

[0111] def call_negotiation():

[0112] tts.say("Hello, Mr. Wang, I am the purchasing AI assistant of XX Company") # Using Azure Neural TTS

[0113] while True:

[0114] text = asr.listen() # speech to text

[0115] if "lowest price" in text:

[0116] sentiment = analyze_emotion(text) # Detect the other person's emotions

[0117] response = gpt4.generate_negotiation(sentiment)

[0118] tts.say(response) # Text to speech

[0119] ASR converts the supplier's voice message, "Minimum 2.28 million," into text. Real-time voiceprint recognition can also be added to confirm the caller's identity. Sentiment analysis detects hesitation, triggering GPT-4 negotiation techniques. Example: During a supplier price-cutting negotiation, ASR recognizes the key phrase: "This price is a loss." Sentiment analysis detects anger (92% confidence). The system automatically triggers a soothing speech template (TTS generates: "We understand your difficulties. Let's explore a long-term cooperation plan...") and records the voiceprint characteristics to confirm the caller's identity.

[0120] Furthermore, to avoid duplicate inquiries across multiple channels or price changes across different channels at different times, cross-channel conversation tracking can be performed. For example, a unified conversation ID can be established, linking email subjects, IM chat records, and phone recordings. For example, inquiry #2025-METH-15 ←→ WeChat conversation [Gasifier Quote] ←→ 2025 / 3 / 15 14:30 call; follow up on related records to build a supplier profile, such as customary phone quotes, automatic appointment call time slots, always reducing prices in the third quarter, and setting automatic inquiries at 10:00 on Mondays in the first month of the third quarter; all AI communication records are automatically stored on the blockchain, in compliance with electronic contract signing regulations.

[0121] Energy price collection: Build interfaces for multiple energy supply channels, including virtual power plant API integration, digital twin wind and / or photovoltaic power station API integration, and power market API integration.

[0122] AI multi-channel intelligent inquiry, based on AI and supply channels, conducts automated intelligent inquiry, and supports intelligent inquiry via email, IM instant messaging, and telephone.

[0123] Furthermore, each supplier can deploy a lightweight AI agent locally to conduct price inquiries based on the privacy protection of federated learning. The process example is as follows:

[0124] Supplier A's local AI agent receives the quotation request;

[0125] Calculate the best quote locally (data does not leave the supplier's server);

[0126] Only the encrypted price range is returned: "2.1-2.3 million yuan".

[0127] Energy price collection: Build interfaces for multiple energy supply channels, including virtual power plant API docking, digital twin wind and / or photovoltaic power station API docking, and power market API docking; for example, centralized power station SCADA data access, distributed power station lot equipment data access, virtual power plant API docking such as IEEE 1888 standard protocol, wind and solar digital twin power station forecast data, real-time power generation forecast data, LSTM time series model access, power grid dispatch interface PJM / Day-Ahead interface, frequency regulation quotation crawler, etc.; information obtained through data interaction based on the interface includes but is not limited to price information, for example, by docking with a virtual power plant API, you can obtain the power plant's electricity price information, and by docking with a digital twin wind power station API, you can obtain the wind power station's power generation and corresponding electricity price information. Furthermore, you can also dock with the green certificate trading interface to interact with the green power certificate system.

[0128] The intelligent processing layer's compensation for price collection includes: compensation for non-one-time purchase price data. Usually, the raw materials needed for the project are usually large in quantity due to scale or cycle reasons, and the inventory scale or storage conditions are generally difficult to meet the requirements of one-time purchase. Some equipment in the project may also have short-life consumable replacement parts, and energy consumption such as water and electricity is continuous consumption. These non-one-time purchases of raw materials, equipment, and energy consumption are marked, and the corresponding continuous supply quotations are obtained from the supply channels through the data collection layer. The continuous supply quotation is the total quotation for continuous supply within a period of time, such as the project cycle, obtained through the negotiated agreement, which can be used for direct cost calculation; and at the same time, the periodicity of the corresponding price data is characterized by acquisition of corresponding historical price data or continuous recording of the obtained price data to determine whether the price data has periodicity. For example, the system can demonstrate regular cycles on a time scale (such as day and night / season), fluctuations within ±2σ of the historical mean, duration matching the inherent cycle (e.g., PV output is higher during the day and lower at night), and predictability through the extraction of the main frequency through Fourier transform. σ is the standard deviation, a statistical measure of data dispersion, reflecting the average deviation of data points from the mean. Fluctuations within ±2σ are considered normal, covering approximately 95% of the data, while those exceeding ±3σ are considered abnormal, covering approximately 99.7% of the data. Furthermore, σ can be recalculated periodically based on raw material characteristics to adapt to data changes and dynamically adjust σ. For example, for high-value materials such as catalysts, the standard deviation can be tightened to ±2.5σ, while for less sensitive materials such as raw water, it can be relaxed to ±4σ.

[0129] When continuous supply quotation data collection fails, price data with cyclical characteristics is used to start ARIMA-based time series price forecasting calculation, which is suitable for cyclical raw materials such as catalysts. For example, the quarterly historical price series of a molecular sieve catalyst is: [22, 23, 21, 24] yuan / kg, and the predicted price for the next cycle is: 23.5 yuan / kg (95% confidence interval 22.1-24.9); or Holt-Winters forecast: for example, when calculating costs, the circulating water agent prices from January to March are [125,000, 118,000, 132,000] yuan / ton, and the seasonal indexes are [1.05, 0.98, 1.12] respectively. Cost calculation requires subsequent price forecasts, which are adjusted according to the data frequency settings. The horizontal component: filters random fluctuations, such as the price of 11 in February. .8→12.36. Smaller values ​​are taken when data noise is high. Trend component: suppresses short-term sudden changes, such as the sudden increase of 13.2 in March to a slight trend adjustment of 0.49. Smaller values ​​are taken when the trend is stable. Seasonal component: retains regular fluctuations, such as the Q1 peak season index of 1.05-1.12. Larger values ​​are taken when seasonality is strong. For example, the April price will be 134,900 yuan / ton and the subsequent months as needed; or historical prices are indeed supplemented. For example, the monthly purchase price of methanol synthesis catalyst (PDS) is missing. The historical prices (10,000 yuan / ton) are: [180, 182, 179, 185, ?, 183] (May data is missing). The characteristic quarterly purchasing peak season (Q2 index is 1.15) has a standard deviation of σ=2.1 over the past 12 months. The main Fourier analysis period is 3 months (quarterly cycle). The fluctuation range is: 179-185 (all within μ±2σ are periodic). Holt-Winters forecast compensation result: 1.843 million yuan is used as the price in May.

[0130] When the collection of non-periodic prices and one-time purchase prices fails, for example, if a compliant supplier loses contact, is out of stock, or is unable to supply for other reasons, the knowledge graph is used to complete the inference and use the associated alternative supply channel quotes or historical quotes for weighted calculations. For example, if the supply channel price collection for a desulfurization protective agent A fails, and the historical quote is 11,000 yuan / cubic meter, the intellectual property graph is used to associate a desulfurization protective agent B with the same function and collect the quote, resulting in a quote of 12,500 yuan / cubic meter. Furthermore, to avoid excessive price differences affecting the accuracy of cost calculations, a preset threshold is set for the comparison of the replacement price and historical quotes to trigger an alert, such as a price difference threshold of 10%. For example, when the valve quote for supplier A is missing, the quote from the associated external supply channel of the same type or the second-choice supplier B, or a combination of existing historical quotes, is used for weighted calculations, with the weighting coefficient based on the volatility of historical quotes. Furthermore, the knowledge graph is updated based on solution associations and transaction activity / equipment similarity.

[0131] Furthermore, some emergencies, such as extreme weather, can cause large-scale price collection omissions or anomalies. The price collection compensation in the intelligent processing layer also includes price range prediction when 20%-30% of data is missing. Abnormal prices are filtered through the IsolationForest algorithm, the current reasonable price range is calculated using Huber regression, and missing data is supplemented by Monte Carlo simulation. This can not only cope with the confusion of missing data, but also keenly identify price traps.

[0132] For example, consider a scenario where 30% of suppliers haven't quoted a price, or one supplier has quoted an abnormally high price. For example, the market price of 304 stainless steel valves is concentrated between 2,000 and 2,500 yuan. In this case, price collection and compensation are performed using the Isolation Forest algorithm, which detects fraudulent quotes and filters out inflated quotes of 3,000 yuan and inflated quotes of 1,500 yuan. Huber regression then ignores the impact of significant erroneous data and calculates the current reasonable price range (2,150 ± 150 yuan). Finally, Monte Carlo simulation randomly assigns a reasonable range of values ​​and, through a large number of simulations, compensates for the impact of individual missing data. The resulting data is then supplemented, resulting in a final guide price of 2,080-2,380 yuan (90% confidence level). Recommended action: Reject quotes above 2,450 yuan.

[0133] The emergency fluctuation processing module sets the fluctuation threshold of the corresponding characteristics according to the price data characteristics, and triggers the weight calculation of the emergency fluctuation if the fluctuation threshold is exceeded.

[0134] Price data features set fluctuation thresholds for corresponding features. For example, for fluctuations in wind and solar power prices, the threshold setting can be based on the price data, available output data, and wind speed data of the corresponding wind and solar power prices connected to the API. For example, in the scenario of a complete wind power outage, the threshold setting is wind speed < 2m / s for 1 hour. For fluctuations in raw material prices, the threshold setting can be based on whether it is inventory-based, directly based on price fluctuations or inventory levels. For example, in the scenario of catalyst shortages, where there is inventory, the threshold setting is inventory < safety threshold, or the threshold is set based on periodic price fluctuations. Fluctuations with periodic characteristics, such as suddenness on a time scale, no fixed period, fluctuations exceeding ±3σ of the historical mean, abnormal duration, and fluctuations that cannot be explained by a periodic model.

[0135] The weight calculation of emergency fluctuations adopts the following modified algorithm:

[0136] , where Ccorrection is the corrected cost under the emergency fluctuation scenario; Cbenchmark is the benchmark cost; ki is the weight coefficient of the i-th emergency fluctuation scenario, reflecting the sensitivity of the fluctuation scenario to the cost; Si is the intensity coefficient of the i-th emergency fluctuation scenario, reflecting the intensity of the fluctuation.

[0137] For example, consider a sudden typhoon that simultaneously causes: wind power outage (wind speed 0.5 m / s for 3 hours) and catalyst transport delays (exceeding safety stock by 2 days). C is the baseline cost, which is the theoretical cost under normal operating conditions, e.g., equipment cost + raw material cost + energy cost. ki is the weight coefficient for the i-th emergency fluctuation scenario, reflecting the cost sensitivity of that fluctuation scenario. The value of K is determined based on historical data statistics, business rules and risk appetite constraints, mathematical model quantitative calculations, and iterative optimization. It is determined based on different scenarios. For example, if wind power fluctuations result in an average cost increase of 25% over the past five interruptions, k = 0.25. Alternatively, if there is no historical data, weighting can be calculated based on the normal wind power cost of ￥0.3 / kWh, the cost of switching to mains electricity of ￥0.58 / kWh, and the proportion of wind power consumption. In the case of a catalyst material shortage, the production suspension loss is ￥200,000 / day, the emergency procurement premium is +30%, and the baseline cost is ￥500,000. k = (200,000 + 500,000 × 0.3) / 500,000 = 0.7 → After adjustment by business rules: k = 0.4. Business rules can be set by company departments based on the company's risk appetite. For example, the Production Department might decide: "k = 0.4 still underestimates the risk of catalyst stockouts. We recommend ≥ 0.5." The Finance Department might decide: "We need to control k ≤ 0.45 to ensure budget security." → Final compromise value: k = 0.4, or set risk levels, such as 1-5 corresponding to k = 0.1-0.5. Mathematical models for quantitative calculation and iterative optimization, such as machine learning, can use regression models to fit the optimal k value based on historical data.

[0138] Si is the intensity coefficient for the i-th emergency fluctuation scenario, reflecting the intensity of the fluctuation. The S value is quantified by threshold comparison. For example, if the threshold setting is wind speed < 2 m / s for 1 hour, and the actual feedback is wind speed 0.5 m / s for 3 hours, the intensity calculation S = ((2-0.5) / 2) * (3 / 1) = 2.25. For example, if catalyst inventory is out of stock for less than 3 days and the supply is out of stock for 2 days, the intensity calculation S = 2 / 3 = 0.67. Furthermore, when multiple factors trigger the same scenario, a composite intensity calculation is performed based on proportional intensities. Furthermore, to prevent distortion of extreme values, the S value is usually capped, for example, S_max = 3. Further fluctuation thresholds can be linked to business rules. For example, a business constraint sets a raw water price cap of 6.5 yuan per ton, with an emergency triggered if the price exceeds the business limit.

[0139] Assuming that the base cost C benchmark is 10 million yuan, the revised cost C under the typhoon impact scenario is calculated to be 18.305 million yuan.

[0140] The intelligent processing layer also supplements cost calculation weightings, including compensation for raw material quality fluctuations. Weighted compensation is set based on raw material standards. For example, biomass raw materials can fluctuate in ash content, affecting the calorific value per unit of raw material. High ash content fluctuations require increased pretreatment or additional fuel gas budgets. This can be achieved by weighting contingency fluctuations. For example, a percentage threshold for ash content can be set for subsequent corrections. For example, a straw ash content time series (%) is [15, 16, 18, 17, 35, 16], with a contract basis of 15%. Trend analysis shows that the first four points meet normal harvest season fluctuations (±20%), while the fifth point suddenly increases to 35% (+133%). Business rules state that an ash content >25% automatically triggers a quality anomaly. The fifth point indicates extreme fluctuation (activating contingency fluctuation correction), while the remaining points indicate cyclical fluctuations. Compensation can be incorporated into the ongoing supply quote during price collection. Alternatively, separate weights for ash content deviation and calorific value loss can be calculated for cost compensation, or experimental data can be used to fit the calculation. Furthermore, adjustments can be set to trigger supplier inquiry weightings (with ash-sensitive suppliers receiving priority).

[0141] Furthermore, the cost-weighting supplements within the intelligent processing layer include carbon price offsets, which offset the cost of captured CO2 proportionally to the carbon trading price. Green chemical processes include CO2 capture. CO2 raw materials account for approximately 20-30% of methanol production costs. This is determined by the carbon content of the raw materials (e.g., coal, natural gas, biomass) and the carbon content of the product (e.g., methanol). For example, on the power generation side, sustainability certification of the biomass raw materials (e.g., FSC certification hash) is recorded. On the chemical side, the real-time green electricity ratio of the methanol synthesis process (via smart meter data uploaded to the blockchain) is monitored. Blockchain-based evidence storage is designed to calculate CO2 emissions and capture volumes. A certain percentage of carbon price offsets can be set based on carbon market mechanisms. For example, a price data acquisition channel based on the carbon market's carbon price acquisition mechanism is established. The data collection layer obtains the carbon price and offset ratio, e.g., a carbon price of 100-160 / ton and an offset ratio of 40%-60%. Carbon capture calculations and cost offsets are then performed based on raw material consumption.

[0142] The application layer performs multi-scenario cost calculations including at least the baseline scenario and the extreme fluctuation scenario based on the cost elements analyzed by the import layer, the corresponding price data obtained by the data collection layer, and the cost calculation weights supplemented by the intelligent processing layer, and outputs a cost report.

[0143] For example, the cost report includes: a baseline scenario with equipment costs of 18.2 million yuan, raw material and energy costs of 6.7 million yuan, and a baseline cost of 24.9 million yuan; an extreme scenario (wind power outage) with equipment costs of 18.2 million yuan, raw material and energy costs of 6.7 million yuan, and an emergency supplement of 3.15 million yuan, for an extreme scenario cost of 28.05 million yuan; other scenario calculation examples include a green premium scenario with a base cost of 24.9 million yuan minus a carbon offset of 500,000 yuan, for a green premium scenario cost of 24.4 million yuan; and real-time scenario costs (photovoltaic fluctuations, digital twin feedback showing PV output changing from 50MW to 5MW (at 8:05:32) and the historical pattern library matching showing non-periodic fluctuations. The system automatically increased the wind power weight to 85% and added grid procurement at 8:05:33), outputting real-time cost changes: a base cost of 24.9 million yuan to 29.5 million yuan (+18.4%) due to a 90% drop in PV output.

[0144] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0145] Those skilled in the art may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0146] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0147] Finally, it should be noted that, in this document, relationships such as first and second, etc., are used solely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms include, comprise, or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0148] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A report generation system, characterized in that: include: The import and parsing layer is used to import the process package corresponding to the green chemical project and parse the process package to obtain production information; A data collection layer for acquiring evaluation information matching the production information; A data processing layer, configured to compensate the evaluation information and determine the impact weights corresponding to different production scenarios; wherein the impact weights correspond to the intensity of fluctuations in the evaluation information due to different production scenarios; The report generation layer is used to generate an evaluation report of the green chemical project in different production scenarios based on the compensated evaluation information, the impact weight and the production information.

2. The report generation system according to claim 1, characterized in that Import the parsing layer, including: A parsing sublayer, configured to extract equipment information, raw material information, and energy consumption information from the process package based on semantic parsing and in combination with a parameter extraction rule library, and determine the equipment information, raw material information, and energy consumption information as the production information; The derivation sublayer is used to establish a mapping relationship between the production information and the information channel; wherein the information channel is a channel for obtaining the evaluation information.

3. The report generation system according to claim 1, wherein: Data collection layer, including: The information capture layer is used to obtain evaluation information corresponding to equipment and raw materials from web pages; The interface docking layer is used to connect to the energy supply channel interface and obtain the corresponding energy assessment information; The integrated communication layer is used to obtain real-time evaluation information based on artificial intelligence technology and combined with multiple communication technologies.

4. The report generation system according to claim 3, wherein: Comprehensive communication layer, including: The email communication layer is used to determine the email workflow schedule, generate and send target emails, receive and parse reply emails to the target emails, and obtain real-time evaluation information; The online program layer is used to simulate manual operation, send a standard evaluation information acquisition template to the target object, and determine the real-time evaluation information based on the response information fed back by the target object; The telephone communication layer is used to obtain real-time assessment information in a conversational manner by converting text to speech.

5. The report generation system according to claim 3, characterized in that: The data collection layer also includes: The evaluation and negotiation layer is used to verify the evaluation information without the data being out of the domain.

6. The report generation system according to claim 1, wherein: Data processing layer, including: For production factors that are obtained multiple times, obtain the corresponding evaluation information in batches and identify the periodic evaluation information; production factors include raw materials, equipment, and energy consumption; After data collection fails, use periodic evaluation information to predict current evaluation information; When data collection fails and information prediction fails, substitute evaluation information is obtained through knowledge graph completion inference.

7. The report generation system according to claim 6, characterized in that: The data processing layer also includes: When the missing assessment information reaches the specified proportion, the acquired assessment information will be anomaly checked, and the missing assessment information will be simulated and supplemented within the calculated reasonable range.

8. The report generation system according to claim 1, wherein: Data processing layer, including: According to the fluctuation threshold corresponding to the assessment information, the emergency fluctuation weight is determined for the assessment information that exceeds the fluctuation threshold.

9. The report generation system according to claim 8, characterized in that: The fluctuation threshold corresponds to the raw material.

10. The report generating system according to claim 8, wherein: The data processing layer also includes: Obtain deduction information, and use the deduction information to deduct and compensate the evaluation information.

Images