A knowledge graph and multi-algorithm fusion large kitchen energy efficiency optimization method

Abstract

The knowledge graph and large kitchen energy efficiency optimization method based on multi-algorithm fusion provided by the application, S1: constructing a knowledge graph mode layer based on LLM; S2: obtaining target kitchen site information and instantiating, generating individual attribute assertions and relationship assertions; generating individual hierarchical reasoning rules, and constructing a knowledge graph instance layer; S3: collecting and preprocessing multi-source time sequence data of the target kitchen site, storing the data into a time sequence database, and establishing a semantic mapping relationship; S4: performing energy efficiency analysis and prediction through multi-algorithm fusion, and generating a device collaborative optimization strategy; S5: issuing the device collaborative optimization strategy to energy-consuming devices for execution, and monitoring the execution effect in real time, forming a closed-loop feedback optimization mechanism. In the application, the kitchen energy efficiency knowledge graph is constructed through LLM, and the closed-loop feedback of multi-source data processing, semantic mapping and energy efficiency optimization strategy is combined, so that the energy consumption minimization and energy efficiency improvement in the multi-device collaborative scene of the large kitchen are realized, and the environmental comfort and device low loss are taken into account, and the intelligent closed-loop control target is achieved.

Description

Technical Field

[0001] This invention relates to the fields of knowledge engineering, artificial intelligence and energy management, and in particular to a method for optimizing energy efficiency in large kitchens by integrating knowledge graphs and multiple algorithms. Background Technology

[0002] Large kitchens are important service areas in shopping malls, hotels, hospitals, offices, and industrial plants, with a significant proportion of energy consumption. Currently, a patent (application number: CN202510683082.2) discloses a kitchen energy data management system, which achieves spatiotemporal alignment of electrical, thermal, and mechanical data through edge layer protocol adaptive conversion and millisecond-level time synchronization technology, eliminating data silos. Another patent (application number: CN202411575077.1) discloses a commercial kitchen energy consumption monitoring and energy-saving management system, which uploads sensor data to a cloud platform and uses big data analysis technology and intelligent control modules to achieve energy consumption data analysis and energy saving. The limitations of existing patents mainly include: only achieving spatiotemporal alignment of energy data, without involving multi-device collaborative energy efficiency optimization; relying on single big data analysis for energy saving, lacking deep integration and intelligent optimization capabilities, all remaining at the level of "data processing" or "isolated optimization," and failing to break through the core need for multi-device collaboration in large kitchens. Summary of the Invention

[0003] The purpose of this invention is to provide a method for optimizing energy efficiency in large kitchens by integrating knowledge graphs and multiple algorithms. It proposes a multi-device collaborative energy efficiency optimization method, which integrates equipment association information through knowledge graphs and combines multiple algorithms to achieve dynamic collaboration among equipment, personnel, environment, and between equipment. This solves the pain point of traditional solutions being "independent" and significantly improves the accuracy and systematic nature of energy efficiency optimization, filling the technical gap in system-level collaborative optimization of large kitchens.

[0004] To achieve the above objectives, this invention proposes a method for optimizing energy efficiency in large kitchens by fusing knowledge graphs with multiple algorithms, comprising the following steps: S1: Construct an energy efficiency knowledge ontology and entity-level reasoning rules based on LLM to form the pattern layer of the knowledge graph; S2: Obtain specific information about the target kitchen site and instantiate it as an individual of the corresponding entity in the energy efficiency knowledge ontology, generate attribute assertions and relation assertions of the individual; and combine the predefined rules in the knowledge graph to dynamically generate individual-level reasoning rules and construct the instance layer of the knowledge graph. S3: Collect multi-source time-series data of the target kitchen site, store it in the time-series database after preprocessing, and establish a semantic mapping relationship between the energy efficiency knowledge ontology and the time-series data; S4: Based on the knowledge graph, energy efficiency analysis and prediction are performed through the fusion of multiple algorithms to generate energy efficiency optimization strategies for equipment collaboration; S5: The energy efficiency optimization strategy is sent to the corresponding energy-consuming devices for execution through the device collaboration controller, and the execution effect of the energy efficiency optimization strategy is monitored in real time to form a closed-loop feedback optimization mechanism.

[0005] Furthermore, S1 includes the following steps: S101: Extracting Domain Terms: Input domain-related documents into the LLM and, through prompting engineering, allow the LLM to extract terms from the domain-related documents, including entities, attributes, and relationships; S102: Construct entity hierarchy: Let the LLM classify the extracted terms to form hierarchical relationships; S103: Define attributes and relationships: Allow the LLM to establish attributes and relationships for each of the entities described; S104: Generate entity-level rules: Input text or cases describing energy efficiency optimization strategies into the LLM and convert them into rule expressions composed of the entities, attributes, and relationships; S105: Ontology Evaluation and Improvement: The energy efficiency knowledge ontology and entity-level reasoning rules initially constructed are reviewed and improved by LLM and domain experts, and an ontology library and rule library are generated.

[0006] Furthermore, S2 includes the following: S201: Constructing instantiated individuals in the ontology: Instantiate the core elements in the target kitchen site, that is, create instantiated individuals corresponding to each core element in the energy efficiency knowledge ontology and store them in the ontology library; S202: Extraction of individual static attributes and relationships: Based on the predefined attributes and relationships in the energy efficiency knowledge ontology, extract the attribute assertions and relationship assertions corresponding to the individual from the collected information data of the target kitchen site, and store them in the ontology library; S203: Individual-level rule establishment: The individual automatically inherits the entity-level reasoning rules to which it belongs, and based on the inheritance, constructs detailed individual-level reasoning rules with parameter constraints or triggering conditions based on the individual's attribute values, relationships and historical data, and stores them in the rule base; S204: Knowledge Graph Update: The completed individuals, attributes, relationships, and rules are incrementally updated to the knowledge graph, and the constraints and rules defined in the knowledge graph are used for verification to dynamically update the knowledge graph.

[0007] Furthermore, in S201, the core elements include, but are not limited to, equipment entities, spatial areas, operating states, and operational activities.

[0008] Furthermore, S3 includes the following: S301: The time-series data is collected and preprocessed by the sensing unit deployed at the target kitchen site, and stored in the time-series database; S302: Construct a semantic mapping table between the energy efficiency knowledge ontology and the time-series data.

[0009] Furthermore, in S301, the sensing unit includes, but is not limited to, an environmental sensor, an equipment status sensor, and a visual perception module. The sensing unit collects data on equipment operation, environment, time and space, and personnel, and stores the data in the time-series database in the form of identifiers, indicator types, values, and timestamps after preprocessing.

[0010] Furthermore, in S302, based on the identifiers in the time-series database and combined with the energy efficiency knowledge ontology, the semantic mapping table is constructed. The semantic mapping table includes the identifiers and fields in the time-series database and the mapping relationship between them and the corresponding entities and attributes of the unified resource identifiers in the energy efficiency knowledge ontology.

[0011] Furthermore, S4 includes the following: S401: Construct an energy consumption prediction and analysis model: Identify the equipment, areas, environment and operational activities related to the target energy consumption, obtain the semantic relationships of the ontology in the knowledge graph related to the equipment, areas, environment and operational activities and the business logic of the rule base, and obtain the corresponding historical data and real-time data in the time series database, and input them into the prediction algorithm. The generated prediction data is the energy consumption prediction result. S402: Construct a collaborative optimization strategy model: Based on the real-time and predicted data of the equipment, region, environment, and operational activities related to the target energy consumption, and combined with the corresponding constraints and rules in the knowledge graph, construct an optimization problem that includes decision variables, multi-objective functions, and multi-dimensional constraints; by solving the optimization problem, generate the globally optimal energy efficiency optimization strategy.

[0012] Compared with existing technologies, the beneficial effects of this invention are mainly reflected in the following aspects: This application constructs a large-scale kitchen energy efficiency knowledge graph. By defining multi-dimensional semantic relationships such as membership, causality, and space between entities such as equipment, environment, area, and personnel, discrete perceptual data is integrated into structured knowledge with contextual association. On this basis, by utilizing the knowledge graph relationships and combining energy consumption prediction algorithms, optimization algorithms, and multi-objective algorithms, feature engineering and constraint generation are performed, significantly improving the accuracy of prediction and the rationality of energy consumption optimization decisions, thereby enhancing the efficiency of energy consumption prediction and decision-making. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the large-scale kitchen energy efficiency optimization method that integrates knowledge graphs and multiple algorithms in this invention. Detailed Implementation

[0014] The following will describe in more detail the energy efficiency optimization method for large kitchens based on the fusion of knowledge graphs and multiple algorithms of the present invention, with reference to the schematic diagrams. Preferred embodiments of the invention are illustrated. It should be understood that those skilled in the art can modify the invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the invention.

[0015] The invention is described more specifically by way of example in the following paragraphs with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0016] like Figure 1 As shown, this invention proposes a method for optimizing the energy efficiency of a large-scale kitchen by fusing knowledge graphs with multiple algorithms, comprising the following steps: S1: Construct an energy efficiency knowledge ontology and entity-level reasoning rules based on LLM (Large Language Model) to form the pattern layer of the knowledge graph; S101: Extract Domain Terms: Input domain-related documents into the LLM, including equipment manuals, operating procedures, energy management standards, etc., and through prompting engineering, allow the LLM to extract terms from the domain-related documents. The terms include entities, properties, and relationships. S102: Constructing an entity hierarchy: Allowing the LLM to classify entities in the extracted terms and form hierarchical relationships; S103: Define attributes and relationships: Allow the LLM to establish attributes and relationships for each entity; S104: Generate entity-level rules: Input text or cases describing energy efficiency optimization strategies into the LLM and convert them into rule expressions composed of entities, attributes and relationships in the energy efficiency knowledge ontology; S105: Ontology Evaluation and Improvement: The initially constructed energy efficiency knowledge ontology and entity-level reasoning rules are reviewed and improved by LLM and domain experts, and an ontology library and rule library are generated.

[0017] Example 1: LLM is used to extract terms such as entities and attributes from relevant energy efficiency standards.

[0018] When inputting a domain-related document into an LLM program, namely the national standard GB 21456-2024 "Minimum Allowable Values ​​of Energy Efficiency and Energy Efficiency Grades for Household and Similar Kitchen Appliances", the program prompts the following: "You are a knowledge graph construction expert. Please extract the core entities and attributes from the following standard text and present them in the Protégé ontology language, and record them in the CamelCase English standard format." The expected output of LLM is as follows: Device entity classes may include: Electric Rice Cooker, Induction Cooker, Microwave Oven, etc.; Attribute classes may include: hasPowerRating, hasCapacity, usesEnergy, hasMaintainace, etc.

[0019] An example of hierarchical organization of the extracted entities using LLM can be: Thing │ ├── Equipment │ │ ├── Cooking Equipment │ │ │ ├── RiceCooker (electric rice cooker) │ │ │ ├── PressureCooker │ │ │ ├── GasStove (Gas Stove) │ │ │ ├── InductionCooker (Induction Cooker) │ │ │ ├── Oven │ │ │ └── Grill (BBQ Equipment) │ │ │ │ │ ├── Storage Appliance │ │ │ ├── Refrigerator │ │ │ └── WineCooler │ │ │ │ │ └── Cleaning Appliance │ │ ├── Dishwasher │ │ └── FoodWasteDisposer │ │ LLM can also extract basic rules from GB 21456-2024, such as adding a rule for CookingAppliance: any individual cooking appliance must be associated with at least one individual EnergyType entity through the usesEnergy attribute.

[0020] The rule expression is: CookingAppliance⊑uesEnergy min 1 EnergyType.

[0021] Therefore, through step S1, entities, attributes, relationships, and rules, including equipment entities, spatial regions, operating states, and operational activities, can be established.

[0022] S2: Obtain specific information such as equipment, functional areas, operating status and operation methods of the target kitchen site, and instantiate them into individuals of the corresponding entities in the energy efficiency knowledge ontology, generating attribute assertions and relation assertions of the individuals; and combine them with predefined rules in the knowledge graph to dynamically generate hierarchical reasoning rules for individuals and construct the instance layer of the knowledge graph. S201: Constructing instantiated individuals in the ontology: Instantiate the core elements in the target kitchen site. The core elements include, but are not limited to, equipment entities, spatial areas, operating status and operational activities. That is, create the corresponding instantiated individuals of each core element in the energy efficiency knowledge ontology and store them in the ontology library. S202: Extraction of static attributes and relationships of individuals: Based on the predefined attributes and relationships in the energy efficiency knowledge ontology, extract the attribute assertions and relationship assertions of the corresponding individuals from the information data collected from the target kitchen site, and store them in the ontology library; S203: Individual-level rule establishment: Individuals automatically inherit the reasoning rules of their respective entity levels, and based on the inheritance, construct more detailed individual-level reasoning rules with parameter constraints or triggering conditions based on the individual's attribute values, relationships and historical data, and store them in the rule base; S204: Knowledge Graph Update: Incrementally update the knowledge graph with the completed individuals, attributes, relationships and rules, and verify them using the constraints and rules defined in the knowledge graph, and dynamically update the knowledge graph.

[0023] Example 2: Collect specific information on equipment, functional areas, operating status, and operational activities in a specific kitchen. For example, if the kitchen is divided into a cooking area, a preparation area, a washing and disinfection area, and a storage area, and the cooking area has 3 rice cookers, 1 induction cooker, 2 steam ovens, and 2 gas stoves, then create the following individuals for RiceCooker: RC_001, RC_002, and RC_003. For individual RC_001, it has the following attributes and attribute values: manufacturedBy(Midea) hasCapacity(4.0) and hasPowerRating(1200).

[0024] One rule for rice cookers is: if the temperature sensor detects overheating (>150°C), then THEN immediately cut off the power and issue an alarm. In the SWRL language of protégé, this is expressed as: / / If the rice cooker is cooking and the temperature is >150 and no one is in the cooking area, then cut off the power immediately.

[0025] RiceCooker(?rc)^hasStatus(?rc,"cooking"^^xsd:string)^TempSensor(?ts)^hasSensor(?rc,?ts)^hasTemperatureValue(?ts,?temp)^sw rlb:greaterThan(?temp,150.0^^xsd:double)^Location(?loc)^hasLocation(?rc,?loc)^hasOccupancy(?loc,"Unoccupied"^^xsd:string) ->triggerEmergencyPowerOff(?rc)^triggerAlarm(?rc) S3: Collect multi-source time-series data from the target kitchen site, preprocess it, store it in the time-series database, and establish a semantic mapping relationship between the energy efficiency knowledge ontology and the time-series data; S301: Through sensing units deployed at specific target kitchen sites, including but not limited to environmental sensors, equipment status sensors and visual perception modules, the sensing units collect data on equipment operation, environment, time and space and personnel, and store the data in the time series database in the form of identifiers, index types, values ​​and timestamps after preprocessing. S302: Based on specific identifiers in the time-series database and combined with the energy efficiency knowledge ontology, construct and maintain a semantic mapping table between the energy efficiency knowledge ontology and the time-series data. The semantic mapping table includes the identifiers and fields in the time-series database and the mapping relationship between them and the corresponding entities and attributes of the unified resource identifiers in the energy efficiency knowledge ontology.

[0026] Example 3: The following sensing units are deployed in the kitchen to systematically collect comprehensive data: smart meters, gas meters, and smart controllers are installed on the equipment; temperature and humidity sensors are deployed in various areas to continuously monitor ambient temperature and humidity; CO, CO2, VOCs, and PM2.5 concentration sensors are installed above the stove to monitor air quality; and cameras are deployed in key passageways and areas of the kitchen to count people and identify activities using computer vision algorithms.

[0027] Optionally, the deployment scheme for the sensing unit of the kitchen ventilation system includes: installing a three-phase smart meter in the main exhaust fan motor circuit in the power distribution cabinet to monitor the main exhaust fan motor data, such as voltage, current, active power, reactive power, power factor, and cumulative power consumption; installing vibration sensors on the fan bearing housing to monitor vibration acceleration and velocity for predictive maintenance; installing differential pressure sensors at the fan inlet and outlet to monitor the pressure difference between the fan inlet and outlet and calculate the actual operating power of the fan; installing status / wind speed sensors and electrically adjustable dampers on the air outlets and smoke hood regulating valves to collect wind speed, damper opening, open / close status, fault signals, etc., to monitor the total air volume of the system and the air volume of key areas, verify the damper regulation effect, and provide data support for air volume balancing and optimization; and installing a high-temperature resistant temperature sensor near the smoke hood filter to monitor the smoke exhaust temperature, triggering an alarm and increasing exhaust ventilation when the temperature is too high to prevent fire.

[0028] The raw data collected by the sensing unit undergoes preprocessing such as outlier filtering, smoothing, and unit standardization. The processed data is then written into a time-series database such as InfluxDB. For example, taking the monitoring of the air valve in a gas stove-fan system as an example, the InfluxDB representation of the air valve opening, airflow, on / off status, and fault status of air outlet 1 (Outlet001) above gas stove 1 (GasStove001) in the cooking area at 12:00 noon on December 30, 2025, can be: damper_monitor,device_type=Fan,device_id=Fan001,outlet_id=Outlet001,zone=kitchen_a opening_percentage=65.2,flow_rate=850.7,status= "running",fault_level=0 1767081600000000000 In this embodiment, a semantic mapping table is established between specific identifiers and fields in the time-series database and entities and attributes in the ontology library. For example, the identifiers and fields of the semantic mapping table for monitoring the air valve of the gas stove-fan system are as follows: Table 1. Examples of semantic mapping representations of entities in time-series databases and ontology libraries:

[0029] Table 2. Examples of semantic mapping representations of attributes in time-series databases and ontology libraries:

[0030] S4: Based on the knowledge graph, energy efficiency analysis and prediction are performed through the fusion of multiple algorithms to generate energy efficiency optimization strategies for equipment collaboration; S401: Construct an energy consumption prediction and analysis model: Identify equipment, areas, environments, and operational activities related to the target energy consumption; obtain the semantic relationships of the ontology and the business logic of the rule base in the knowledge graph related to the equipment, areas, environments, and operational activities; and obtain the corresponding historical and real-time data in the time series database. These are then input into the prediction algorithm, and the generated prediction data is the energy consumption prediction result.

[0031] Example 4: At 12:00 noon on December 30, 2025, a chef was cooking at gas stove 1 (GasStove001) in the cooking area. Gas stove 1 (GasStove001) corresponds to air outlet 1 (Outlet001) of fan 1 (Fan001). First, real-time video data of the chef's hand movements was collected. Hand movement features were extracted using STGNN. For all hand nodes... All time steps Spatiotemporal feature aggregation action feature vector The simplified formula is: ; in: Indicates the weights of the fully connected layer; Indicates the first Layer, First Time step, number Feature vectors of each node; This indicates the bias of the fully connected layer.

[0032] Using ontology rules from a knowledge graph, the cooking method corresponding to an action can be determined. For example, a cooking method identification rule based on action and heat level could include the stirring frequency. 28 times / minute, duration 100 seconds, firepower Level 7, Flame Height If the thickness is 13 cm, it is considered stir-frying and marked as high cooking intensity. The SWRL rule is as follows: CookingActivity(?activity)^ hasStirFrequency(?activity,?freq)^ swrlb:greaterThanOrEqual(?freq,28)^ hasActivityDuration(?activity,?duration)^ swrlb:greaterThanOrEqual(?duration,100)^ hasFirePower(?activity,?power)^ swrlb:greaterThanOrEqual(?power,7)^ hasFlameHeight(?activity,?height)^ swrlb:greaterThanOrEqual(?height,13) -> hasActivityType(?activity,:StirFry)^ hasCookingIntensity(?activity,:High) Secondly, the real-time collected equipment and environmental parameters such as power, fumes, air volume, and damper opening are preprocessed and then fused with cooking method recognition and semantic rules as input. LSTM / Transformer is then used to predict energy consumption.

[0033] S402: Construct a collaborative optimization strategy model: Based on real-time and predicted data of equipment, regions, environment and operational activities related to the target energy consumption, and combined with the corresponding constraints and rules in the knowledge graph, construct an optimization problem that includes decision variables, multi-objective functions and multi-dimensional constraints; by solving the optimization problem, generate a globally optimal energy efficiency optimization strategy.

[0034] Example 5: Construct a coordinated optimization strategy for energy-consuming equipment in the cooking area of ​​a kitchen on December 30, 2025, from 6 PM to 8 PM. The cooking area has two gas stoves (GasStove001, GasStove002) and one induction cooker (InductionCooker001), corresponding to the three air outlets (Outlet001, Outlet002, Outlet003) of a fan (Fan001). Data collected includes gas stove heat output, induction cooker power, fan power, damper opening, ambient temperature and humidity, oil fume concentration, and cooking activity type. The decision variables for this coordinated energy consumption optimization strategy include gas stove heat output, fan power, and air outlet opening. The optimization objectives are: minimizing total energy consumption, maximizing environmental comfort, and minimizing equipment wear and tear. The objective function is: ; in: Assign weights to each objective; ; For each objective function.

[0035] Their corresponding objective functions are, respectively, the total energy consumption minimization function: ; in: Energy consumption coefficient; for The actual operating power of the device at the time step, i.e., in this embodiment, 1. 2. 3 are respectively The operating power of two gas stoves and one induction cooker is measured in time steps. 4 The operating power of one fan unit per time step; The actual energy consumption of the fan in time step t; The unit time step.

[0036] Environmental comfort loss minimization function (with temperature, humidity, and cooking fumes as dimensions): ; in: Weighted by comfort level; , , They are respectively Predicted values ​​for temperature, humidity, and cooking fumes in the cooking zone at each time step; The optimal values ​​for temperature and humidity; This is the compliance threshold for cooking fumes.

[0037] Equipment loss minimization function: ; in, As a loss weight; for The actual operating power of the device at the time step, i.e., in this embodiment, 1. 2. 3 are respectively The operating power of the equipment (two gas stoves and one induction cooker) is measured in time steps. 4 The operating power of one fan unit per time step. This refers to the equipment's rated maximum operating power.

[0038] Therefore, the optimal solution of the objective function in this embodiment is obtained using the NSGA-II algorithm, and the optimization algorithm is as follows: Initialize the population: Input preprocessed real-time data and predicted data, and generate an initial feasible solution based on the cooking method-equipment parameter constraint rules and equipment physical operation thresholds in the knowledge graph. For example, for the scenario where the gas stove GasStove001 is in the stir-fry cooking mode, match the constraints that GasStove001 firepower ≥ 7, Fan001 fan power ≥ 2.2kW, and Outlet001 air outlet opening ≥ 92% during stir-frying. Refer to the historical data average and select the corresponding controllable... The initial solution for the decision variables is set as follows: x_GasStove001=8 (GasStove001 firepower level is 8, normalized: x_GasStove001=0.778), x_Fan001=2.3KW (Fan001 operating power is 2.3kW, normalized: x_Fan001=0.682), x_Outlet001=92% (Outlet001 opening degree is 92%, normalized: x_Fan001=0.920); Iterative optimization process: Genetic operations such as selection, crossover, and mutation are used to generate a new generation of solution population. The fitness value of each solution within the population is calculated. Non-dominated solutions are screened based on the non-dominated sorting method, and population optimization is completed by combining crowding distance. The optimized solutions are then subjected to multi-dimensional constraint verification, including ontology rules and equipment physical thresholds, and invalid solutions that violate the constraints are eliminated. Iterative optimization and constraint verification operations are continuously executed in a loop. When the change in the optimal fitness value of the population tends to stabilize and meets the preset convergence threshold, the algorithm is considered to have converged, the iteration is terminated, and the Pareto optimal solution set is output. Generate a globally optimal collaborative strategy: Based on operational priorities, select the globally optimal solution from the Pareto optimal solution set, and after inverse normalization to restore it to the actual operating parameters of the equipment, transform it into an executable collaborative control strategy. For example, in this embodiment, the specific control values ​​of gas stove firepower, induction cooker power, fan power and corresponding air outlet opening are specified for each time period to complete the collaborative control of cooking equipment and smoke exhaust equipment.

[0039] S5: The energy efficiency optimization strategy is sent to the corresponding energy-consuming devices for execution through the device collaboration controller, and the execution effect of the energy efficiency optimization strategy is monitored in real time to form a closed-loop feedback optimization mechanism.

[0040] Example 6: The minute-by-minute timing control instructions included in the global optimal collaborative strategy are precisely sent to the corresponding kitchen energy consumption equipment. Specifically, fire level control instructions are sent to GasStove001 and GasStove002 (e.g., GasStove001 fire level 8 during stir-frying and fire level 4 during simmering), overall machine operating power control instructions are sent to Fan001 (e.g., 2.3kW during stir-frying and 1.2kW during heat preservation), and air valve opening control instructions are sent to Outlet001, Outlet002, and Outlet003 respectively (e.g., Outlet001 opening is 92% during stir-frying), so that each device can execute synchronously according to the strategy. Meanwhile, temperature sensors, humidity sensors, oil fume concentration sensors, and equipment operation status sensors deployed in the cooking area continuously collect and upload actual operating parameters, environmental status parameters, and operational parameters of each device, enabling dynamic monitoring of the strategy's execution effect. During monitoring, if equipment operating parameters deviate from the control command threshold (e.g., the actual power of the fan deviates from the strategy value by >0.1kW), or environmental parameters fail to meet comfort targets (e.g., oil fume concentration >5mg / m³), the system will detect and detect such deviations. 3 If the temperature deviation is greater than 5℃, the monitoring data will be fed back to the optimization module in real time, triggering a new round of iterative optimization. A new collaborative control strategy adapted to the current working conditions will be generated and issued for execution, forming a closed-loop feedback optimization to ensure that the optimization goals of optimal energy consumption, environmental comfort, and stable equipment operation are continuously achieved.

[0041] In summary, this embodiment proposes a large-scale kitchen energy efficiency optimization method that integrates knowledge graphs and multiple algorithms. This application constructs a large-scale kitchen energy efficiency knowledge graph, defining multi-dimensional semantic relationships such as membership, causality, and space between entities like equipment, environment, region, and personnel. This integrates discrete perceptual data into structured knowledge with contextual associations. Based on this, the knowledge graph relationships are utilized in conjunction with energy consumption prediction algorithms, optimization algorithms, and multi-objective algorithms to perform feature engineering and constraint generation, significantly improving the accuracy of predictions and the rationality of energy consumption optimization decisions, thereby enhancing the efficiency of energy consumption prediction and decision-making.

[0042] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for optimizing energy efficiency in a large kitchen by fusing knowledge graphs with multiple algorithms, characterized in that, Includes the following steps: S1: Construct an energy efficiency knowledge ontology and entity-level reasoning rules based on LLM to form the pattern layer of the knowledge graph; S2: Obtain specific information about the target kitchen site and instantiate it as an individual of the corresponding entity in the energy efficiency knowledge ontology, generate attribute assertions and relation assertions of the individual; and combine the predefined rules in the knowledge graph to dynamically generate individual-level reasoning rules and construct the instance layer of the knowledge graph. S3: Collect multi-source time-series data of the target kitchen site, store it in the time-series database after preprocessing, and establish a semantic mapping relationship between the energy efficiency knowledge ontology and the time-series data; S4: Based on the knowledge graph, energy efficiency analysis and prediction are performed through the fusion of multiple algorithms to generate energy efficiency optimization strategies for equipment collaboration; S5: The energy efficiency optimization strategy is sent to the corresponding energy-consuming devices for execution through the device collaboration controller, and the execution effect of the energy efficiency optimization strategy is monitored in real time to form a closed-loop feedback optimization mechanism.

2. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 1, characterized in that, S1 includes the following steps: S101: Extracting Domain Terms: Input domain-related documents into the LLM and, through prompting engineering, allow the LLM to extract terms from the domain-related documents, including entities, attributes, and relationships; S102: Construct entity hierarchy: Let the LLM classify the extracted terms to form hierarchical relationships; S103: Define attributes and relationships: Allow the LLM to establish attributes and relationships for each of the entities described; S104: Generate entity-level rules: Input text or cases describing energy efficiency optimization strategies into the LLM and convert them into rule expressions composed of the entities, attributes, and relationships; S105: Ontology Evaluation and Improvement: The energy efficiency knowledge ontology and entity-level reasoning rules initially constructed are reviewed and improved by LLM and domain experts, and an ontology library and rule library are generated.

3. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 2, characterized in that, S2 contains the following: S201: Constructing instantiated individuals in the ontology: Instantiate the core elements in the target kitchen site, that is, create instantiated individuals corresponding to each core element in the energy efficiency knowledge ontology and store them in the ontology library; S202: Extraction of individual static attributes and relationships: Based on the predefined attributes and relationships in the energy efficiency knowledge ontology, extract the attribute assertions and relationship assertions corresponding to the individual from the collected information data of the target kitchen site, and store them in the ontology library; S203: Individual-level rule establishment: The individual automatically inherits the entity-level reasoning rules to which it belongs, and based on the inheritance, constructs detailed individual-level reasoning rules with parameter constraints or triggering conditions based on the individual's attribute values, relationships and historical data, and stores them in the rule base; S204: Knowledge Graph Update: The completed individuals, attributes, relationships, and rules are incrementally updated to the knowledge graph, and the constraints and rules defined in the knowledge graph are used for verification to dynamically update the knowledge graph.

4. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 3, characterized in that, In S201, the core elements include, but are not limited to, equipment entities, spatial areas, operating status, and operational activities.

5. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 3, characterized in that, S3 includes the following: S301: The time-series data is collected and preprocessed by the sensing unit deployed at the target kitchen site, and stored in the time-series database; S302: Construct a semantic mapping table between the energy efficiency knowledge ontology and the time-series data.

6. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 5, characterized in that, In S301, the sensing unit includes, but is not limited to, environmental sensors, equipment status sensors and visual perception modules. The sensing unit collects data on equipment operation, environment, time and space and personnel, and stores it in the time series database in the form of identifier, indicator type, value and timestamp after preprocessing.

7. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 5, characterized in that, In S302, based on the identifiers in the time-series database and combined with the energy efficiency knowledge ontology, the semantic mapping table is constructed. The semantic mapping table includes the identifiers and fields in the time-series database and the mapping relationship between them and the corresponding entities and attributes of the unified resource identifiers in the energy efficiency knowledge ontology.

8. The method for optimizing energy efficiency in large kitchens by fusing knowledge graphs and multiple algorithms according to claim 5, characterized in that, S4 includes the following: S401: Construct an energy consumption prediction and analysis model: Identify the equipment, areas, environment and operational activities related to the target energy consumption, obtain the semantic relationships of the ontology in the knowledge graph related to the equipment, areas, environment and operational activities and the business logic of the rule base, and obtain the corresponding historical data and real-time data in the time series database, and input them into the prediction algorithm. The generated prediction data is the energy consumption prediction result. S402: Construct a collaborative optimization strategy model: Based on the real-time and predicted data of the equipment, region, environment, and operational activities related to the target energy consumption, and combined with the corresponding constraints and rules in the knowledge graph, construct an optimization problem that includes decision variables, multi-objective functions, and multi-dimensional constraints; by solving the optimization problem, generate the globally optimal energy efficiency optimization strategy.

Images