An AI-based cooking process optimization and energy efficiency management method

Abstract

The application discloses a cooking process optimization and energy efficiency management method based on AI, which comprises the following steps: collecting multi-dimensional data of a cooking process and constructing a cooking feature vector; inputting the cooking feature vector into a deep learning model based on LSTM and an attention mechanism to predict an energy consumption curve of the cooking process; modeling a cooking parameter optimization as a Markov decision process and solving an optimal control strategy by using a reinforcement learning algorithm; solving a Pareto optimal solution set under a preset constraint condition based on a multi-objective evolutionary algorithm with the optimization objectives of minimizing total energy consumption, minimizing temperature deviation and minimizing completion time, obtaining optimized control parameters and issuing the same to kitchen equipment for execution. The application also discloses a corresponding energy efficiency management system. The application realizes significant energy saving and consumption reduction under the premise of ensuring cooking quality through the joint optimization of deep learning and reinforcement learning, and can be widely applied to commercial kitchens, central kitchens and food processing plants and the like.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and smart energy-saving technology, and in particular to a method for optimizing cooking processes and managing kitchen equipment energy efficiency based on machine learning. Background Technology

[0002] Commercial kitchens are a major area of ​​energy consumption; statistics show that kitchen energy consumption accounts for 30% to 40% of the total energy consumption in the catering industry. Traditional kitchen equipment management has the following problems: First, energy consumption monitoring is rudimentary. The lack of sophisticated methods for collecting and analyzing energy consumption data makes it impossible to accurately assess the energy consumption of various devices, resulting in energy waste being difficult to detect and correct in a timely manner.

[0003] Secondly, cooking parameters are based on experience. Chefs rely on personal experience to control the heat and time, and the differences in operation among different chefs lead to inconsistent energy utilization efficiency, making it difficult to establish standardized energy-saving operating procedures.

[0004] Third, equipment operation is disorderly. During peak periods, multiple devices start up simultaneously, causing load shocks to the power grid, increasing electricity costs, and the energy competition between devices also reduces overall energy efficiency.

[0005] Fourth, there is a lack of predictive maintenance. Repairing equipment only after it fails not only affects operational efficiency but also increases maintenance costs, and energy efficiency decline caused by equipment aging is not addressed in a timely manner.

[0006] To address the aforementioned issues, existing energy-saving technologies primarily employ simple timed control or power limiting strategies. However, these solutions have significant shortcomings: they adopt a one-size-fits-all approach, failing to consider the differences in characteristics between various dishes and equipment; they employ a passive response method, adjusting only after energy consumption exceeds limits, lacking foresight; a large amount of accumulated cooking data is not effectively mined and utilized; and excessive energy-saving measures often affect cooking efficiency and food quality, resulting in a poor user experience. Summary of the Invention

[0007] This invention provides an AI-based method and system for optimizing cooking processes and managing energy efficiency. By analyzing historical cooking data through machine learning technology, it establishes an energy consumption prediction model and optimization strategy, achieving energy saving and consumption reduction while ensuring cooking quality. This overcomes the problems of extensive energy consumption monitoring, reliance on experience for cooking parameters, disordered equipment operation, and lack of predictive maintenance in existing technologies.

[0008] The technical solution of this invention is as follows: To achieve the above-mentioned objectives, this invention provides an AI-based method for optimizing cooking processes and managing energy efficiency, comprising the following steps: Data acquisition steps: Collect multi-dimensional data of the cooking process. The multi-dimensional data includes ingredient characteristic data (weight, type, initial temperature, and moisture content of ingredients), equipment characteristic data (rated power, thermal efficiency, and aging coefficient of equipment), environmental characteristic data (ambient temperature, ambient humidity, and altitude), and energy consumption characteristic data (historical power curves and peak-valley electricity price data).

[0009] Feature construction steps: The collected multi-dimensional data is cleaned, features are extracted, normalized, and time-series aligned to construct cooking feature vectors. Data cleaning is used to remove noise and outliers; feature extraction is used to extract valuable feature indicators from the raw data; normalization is used to unify features of different dimensions to the same scale; and time-series alignment is used to synchronize data with different sampling frequencies to a unified time axis.

[0010] Energy consumption prediction steps: The cooking feature vector is input into a deep learning model based on LSTM and an attention mechanism to predict the energy consumption curve and confidence interval of the cooking process. This deep learning model specifically includes: an input layer for receiving the cooking feature vector; an LSTM encoder for extracting temporal dependencies during the cooking process; an attention mechanism layer for automatically learning the weights of each time step and identifying key periods affecting energy consumption; a fully connected layer for mapping the output of the attention mechanism layer to the predicted energy consumption value; and an output layer for outputting the energy consumption curve and the corresponding confidence interval.

[0011] Parameter optimization steps: The cooking parameters are modeled as a Markov decision process, defining a state space, action space, and reward function. A reinforcement learning algorithm is used to solve for the optimal control strategy. The state space includes the current temperature, target temperature, remaining time, and equipment load; the action space is the power adjustment level, ranging from 0 to 100%; the reward function is: R = α × quality score + β × energy saving rate − γ × time penalty, where α, β, and γ are weight coefficients. Preferably, the reinforcement learning algorithm is the Proximal Policy Optimization (PPO) algorithm.

[0012] Multi-objective optimization steps: Based on a multi-objective evolutionary algorithm, with the optimization objectives of minimizing total energy consumption, minimizing temperature deviation, and minimizing completion time, the Pareto optimal solution set is solved under preset constraints to obtain the optimized control parameters. The preset constraints include temperature uniformity constraints, food safety constraints, equipment capacity constraints, and grid load constraints. Preferably, the multi-objective evolutionary algorithm is the NSGA-III algorithm.

[0013] Execution steps: Send the optimized control parameters to the kitchen equipment for execution.

[0014] As a further improvement of the present invention, the method also includes an anomaly detection step: using the isolated forest algorithm to detect anomalies in energy consumption data and temperature data during the cooking process, and issuing an early warning signal when an anomaly is detected, so as to intervene in a timely manner and avoid energy waste or safety accidents.

[0015] As a further improvement of the present invention, the method further includes a feedback learning step: collecting actual energy consumption data, calculating energy saving rate and quality retention degree, updating the parameters of the energy consumption prediction model and the reinforcement learning strategy online, and realizing continuous optimization and adaptation of the model.

[0016] This invention also provides an AI-based cooking process optimization and energy efficiency management system, which includes a data acquisition module, a feature engineering module, an energy consumption prediction module, a parameter optimization module, a multi-objective optimization module, and a decision execution module. The data acquisition module collects multi-dimensional data about the cooking process; the feature engineering module preprocesses the multi-dimensional data to construct cooking feature vectors; the energy consumption prediction module includes a deep learning model based on LSTM and attention mechanisms to predict the energy consumption curve of the cooking process; the parameter optimization module includes a policy network based on reinforcement learning to model the optimization of cooking parameters as a Markov decision process and solve for the optimal control strategy; the multi-objective optimization module uses a multi-objective evolutionary algorithm to jointly optimize total energy consumption, temperature deviation, and completion time to solve for the Pareto optimal solution set; and the decision execution module distributes the optimized control parameters to the kitchen equipment for execution.

[0017] Preferably, the system further includes a feedback learning module for collecting actual energy consumption data, calculating energy saving rate and quality retention, and updating the parameters of the energy consumption prediction model and the reinforcement learning policy network online.

[0018] The advantages of this invention are: (1) Significant energy saving effect. Through the accurate prediction of energy consumption by deep learning model and the dynamic optimization of reinforcement learning strategy, the comprehensive energy saving rate can reach 20% to 25% while ensuring the quality of dishes. Actual tests show that the average energy consumption is reduced by 22% and the peak power is reduced by 18%.

[0019] (2) Cooking quality improved, temperature control accuracy improved from ±8°C to ±3°C, an increase of 62.5%; cooking consistency improved from 75% to 92%, an increase of 17 percentage points.

[0020] (3) Equipment utilization rate improved. Through multi-objective optimization algorithm, the equipment load and running time were reasonably allocated, and the equipment utilization rate increased from 65% to 84%, an increase of 19 percentage points.

[0021] (4) It has strong adaptability. Through the online feedback learning mechanism, the model can continuously learn new data and adapt to different chefs' operating habits and seasonal changes without the need for manual adjustment of parameters.

[0022] (5) It has good interpretability. Based on the attention mechanism, the model can output the analysis of key influencing factors, which makes it easier for chefs to understand the optimization logic and enhance their trust and acceptance of AI decision-making. Attached Figure Description

[0023] Figure 1 This is a diagram illustrating the overall architecture of the AI ​​energy efficiency management system of this invention. Figure 2 This is a structural diagram of the LSTM-Attention energy consumption prediction model of this invention; Figure 3 This is a flowchart of the reinforcement learning parameter optimization process of the present invention; Figure 4 This is a schematic diagram of the Pareto front for multi-objective optimization in this invention; Figure 5 This is a comparison chart showing the actual application effects of the present invention. Detailed Implementation

[0024] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0025] like Figure 1-5 As shown: Example 1: Energy-saving control of intelligent gas stove This embodiment uses the optimization of heat control in the process of stir-frying Chinese dishes as an example to illustrate the specific implementation of the present invention.

[0026] Data Acquisition: Temperature sensors and gas flow sensors are installed on the smart gas stove to collect cookware temperature and gas flow data once per second; ambient temperature data is collected through environmental sensors; and information on dish type and portion size is obtained through user input or image recognition.

[0027] Feature construction: After cleaning the collected raw data, extract the characteristics of ingredients (weight, type, initial temperature), equipment (rated power, thermal efficiency), environmental characteristics (room temperature), and process characteristics (target temperature, heating rate requirements) to construct a unified cooking feature vector.

[0028] Energy consumption prediction: The cooking feature vector is input into a pre-trained LSTM-Attention model to predict the baseline energy consumption curve for this cooking process. This model is trained on a large amount of historical cooking data and can accurately predict the energy consumption distribution for different dishes and portion sizes.

[0029] Parameter optimization: The heat control is modeled as a Markov decision process. The state space includes the current pot temperature, target temperature, remaining cooking time, and gas stove load; the action space is the heat level (continuously adjustable from 0 to 100%); the reward function comprehensively considers the food quality score, energy saving rate, and time efficiency. The PPO algorithm is used to train the optimal control strategy.

[0030] Multi-objective optimization: Based on the basic scheme of reinforcement learning policy generation, the NSGA-III algorithm is further used to jointly optimize energy consumption, temperature deviation and completion time, and solve the Pareto optimal solution set under the constraints of temperature uniformity, food safety and equipment capacity.

[0031] Execution and Feedback: The optimized firepower control curve is sent to the gas stove for execution. After cooking is completed, actual energy consumption data is collected to calculate energy saving rate and quality retention, which are used to update model parameters online.

[0032] Actual results: In actual commercial kitchen tests, after adopting the method of this embodiment, gas consumption was reduced by 18%; the taste score of dishes increased from 4.2 points to 4.6 points (out of 5.0 points); the complexity of chef operations was significantly reduced, with the average number of adjustments reduced from 23 to 8.

[0033] Example 2: Energy Efficiency Optimization of Oven Baking This embodiment uses the optimization of temperature curves for batch bread baking as an example to illustrate another implementation of the present invention.

[0034] Data acquisition: Temperature and power sensors are installed in the industrial oven to collect internal oven temperature and real-time power data once per second; the type, quantity, and initial temperature of the bread are recorded.

[0035] Feature construction and model training: Baking feature vectors are extracted based on historical baking data to train an LSTM-Attention energy consumption prediction model. This model can accurately predict the energy consumption distribution of the entire baking cycle (preheating-baking-cooling process) based on bread type, quantity, and initial temperature.

[0036] Parameter optimization and multi-objective optimization: The temperature control of the baking process is modeled as a Markov decision process, and reinforcement learning is used to optimize key parameters such as preheating rate, baking constant temperature, and holding time. Based on this, the NSGA-III multi-objective evolutionary algorithm is used to further optimize energy consumption and baking time while ensuring bread quality. The algorithm also considers peak and off-peak electricity prices, scheduling high-energy-consuming baking tasks during off-peak hours to minimize electricity costs.

[0037] Implementation and Results: The optimized temperature profile was sent to the oven for execution. In actual commercial kitchen testing, the method described in this embodiment reduced energy consumption per batch by 25%; baking time was shortened by 12%; bread quality consistency was significantly improved, with batch-to-batch variations in appearance, color, and internal structure noticeably reduced; based on the scale of a commercial kitchen, annual electricity savings were approximately 25,000 yuan.

[0038] Example 3: Overall System Implementation This embodiment uses a multi-device collaborative operation scenario in a commercial kitchen as an example to illustrate the overall workflow of the system of the present invention.

[0039] System Deployment Architecture: The system described in this invention adopts a distributed architecture in actual deployment. The data acquisition module is deployed on the device side, collecting real-time data on food characteristics, equipment characteristics, environmental characteristics, and energy consumption characteristics through power sensors, temperature sensors, timers, and environmental sensors installed on each kitchen device. Each device's data acquisition module subunit aggregates and sends the collected raw data to the feature engineering module deployed on a local server or in the cloud via an IoT gateway.

[0040] Collaborative workflow between modules: After receiving the raw data, the feature engineering module cleans, extracts, normalizes, and aligns the multi-dimensional data to construct a unified cooking feature vector, which is then sent to the energy consumption prediction module and the parameter optimization module respectively.

[0041] After receiving the cooking feature vector, the energy consumption prediction module predicts the energy consumption curve of this cooking using its internal LSTM-Attention deep learning model. This model specifically includes an input layer, an LSTM encoder, an attention mechanism layer, a fully connected layer, and an output layer.

[0042] After receiving the cooking feature vector, the parameter optimization module models the cooking parameters as a Markov decision process (the state space includes the current temperature, target temperature, remaining time and equipment load, the action space is the power adjustment level 0-100%, and the reward function is R = α×quality score + β×energy saving rate − γ×time penalty). It generates a preliminary optimal control strategy through its internal reinforcement learning policy network (using the PPO algorithm).

[0043] The multi-objective optimization module receives the outputs from the energy consumption prediction module and the parameter optimization module, and uses the NSGA-III algorithm to jointly optimize the total energy consumption, temperature deviation, and completion time. Under the constraints of temperature uniformity, food safety, equipment capacity, and power grid load, it solves the Pareto optimal solution set to obtain the final optimized control parameters.

[0044] The decision execution module sends the optimized control parameters to the actuators of the kitchen equipment for execution.

[0045] The feedback learning module collects actual energy consumption data after cooking is completed, calculates energy saving rate and quality retention, and updates the parameters of the LSTM-Attention network in the energy consumption prediction model and the parameters of the reinforcement learning policy network in the parameter optimization module online.

[0046] Actual results: In actual commercial kitchen tests, after adopting the system of this embodiment, the peak power was reduced by 18% and the total energy consumption was reduced by 22%, while the serving time and quality of all dishes met the requirements.

[0047] This invention can be widely applied to various commercial kitchens, central kitchens, food processing plants, and other scenarios, and is suitable for a variety of cooking equipment such as gas stoves, induction cookers, ovens, steamers, and fryers. The method of this invention can be implemented through software programs, which can be stored in computer-readable storage media and executed by a processor. The system of this invention can adopt a distributed architecture, with the data acquisition module deployed on the device side and the AI ​​model deployed on a local server or in the cloud, achieving data communication and control command issuance between the device and the server through the Internet of Things (IoT).

[0048] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A cooking process optimization and energy efficiency management method based on AI, characterized in that, Includes the following steps: (1) Data collection steps: Collect multi-dimensional data of the cooking process; (2) Feature construction steps: preprocess the multi-dimensional data to construct cooking feature vectors; (3) Energy consumption prediction step: Input the cooking feature vector into a deep learning model based on LSTM and attention mechanism to predict the energy consumption curve of the cooking process; (4) Parameter optimization steps: The optimization of cooking parameters is modeled as a Markov decision process, defining the state space, action space and reward function, and using reinforcement learning algorithm to solve the optimal control strategy; (5) Multi-objective optimization steps: Based on the multi-objective evolutionary algorithm, with the optimization objectives of minimizing total energy consumption, minimizing temperature deviation and minimizing completion time, the Pareto optimal solution set is solved under the preset constraints to obtain the optimized control parameters; (6) Execution steps: Send the optimized control parameters to the kitchen equipment for execution.

2. The AI-based cooking process optimization and energy efficiency management method according to claim 1, characterized in that: The multi-dimensional data includes food characteristic data, equipment characteristic data, environmental characteristic data, and energy consumption characteristic data; the state space of the Markov decision process includes the current temperature, target temperature, remaining time, and equipment load, the action space is the power adjustment level, and the reward function is: R=α×quality score+β×energy saving rate−γ×time penalty, where α, β, and γ are weighting coefficients.

3. The AI-based cooking process optimization and energy efficiency management method according to claim 1, characterized in that: The reinforcement learning algorithm is a near-end policy optimization algorithm, and the multi-objective evolutionary algorithm is the NSGA-III algorithm; the preset constraints include temperature uniformity constraints, food safety constraints, equipment capacity constraints, and power grid load constraints.

4. The AI-based cooking process optimization and energy efficiency management method according to claim 1, characterized in that, It also includes a feedback learning step: collecting actual energy consumption data, calculating energy saving rate and quality retention, and updating the parameters of the energy consumption prediction model and the reinforcement learning strategy online.

5. An AI-based cooking process optimization and energy efficiency management system, characterized in that, include: The data acquisition module is used to collect multi-dimensional data during the cooking process; The feature engineering module is used to preprocess the multi-dimensional data and construct cooking feature vectors; The energy consumption prediction module includes a deep learning model based on LSTM and attention mechanisms, used to predict the energy consumption curve of the cooking process; The parameter optimization module includes a reinforcement learning-based policy network, which is used to model the optimization of cooking parameters as a Markov decision process and solve for the optimal control policy. The multi-objective optimization module is used to jointly optimize total energy consumption, temperature deviation, and completion time based on a multi-objective evolutionary algorithm, and solve for the Pareto optimal solution set. The decision execution module is used to send the optimized control parameters to the kitchen equipment for execution.

6. The AI-based cooking process optimization and energy efficiency management system according to claim 5, characterized in that, The data acquisition module collects multi-dimensional data including food characteristic data, equipment characteristic data, environmental characteristic data, and energy consumption characteristic data; in the parameter optimization module, the state space of the Markov decision process includes the current temperature, target temperature, remaining time, and equipment load, the action space is the power adjustment level, and the reward function is: R=α×quality score+β×energy saving rate−γ×time penalty.

7. The AI-based cooking process optimization and energy efficiency management system according to claim 5, characterized in that: The parameter optimization module uses a near-end strategy optimization algorithm as a reinforcement learning algorithm; the multi-objective optimization module uses the NSGA-III algorithm as a multi-objective evolutionary algorithm, and the preset constraints include temperature uniformity constraints, food safety constraints, equipment capacity constraints, and power grid load constraints.

8. The AI-based cooking process optimization and energy efficiency management system according to claim 5, characterized in that: It also includes a feedback learning module, which is used to collect actual energy consumption data, calculate energy saving rate and quality retention, and update the parameters of the energy consumption prediction model and the reinforcement learning policy network online.

Images