A robot intelligent manufacturing workshop energy consumption optimization method based on internet of things

Abstract

The application discloses a kind of robot intelligent manufacturing workshop energy consumption optimization methods based on Internet of Things, belong to the field of energy consumption optimization, including: based on genetic algorithm optimization path, generate the candidate path set containing diversified path;For time sequence scheme, the energy consumption uncertainty caused by task change and equipment aging is predicted using Monte Carlo simulation algorithm, generate the prediction distribution containing energy consumption fluctuation range;If the energy consumption fluctuation range of prediction distribution exceeds 10% of preset threshold, then the path set and time sequence scheme are adjusted by reinforcement learning algorithm, generate optimized path and time sequence combination;Otherwise, keep current path and time sequence scheme, obtain stable energy consumption allocation scheme;Task execution sequence and equipment operating parameters are extracted from the final adaptive energy consumption optimization scheme, transmitted to workshop control system, complete the real-time deployment of energy consumption optimization.

Description

Technical Field

[0001] This invention belongs to the field of energy consumption optimization, and in particular relates to an energy consumption optimization method for a robot-based intelligent manufacturing workshop based on the Internet of Things. Background Technology

[0002] In the field of intelligent manufacturing, energy consumption optimization in robotic intelligent manufacturing workshops is an important direction for promoting sustainable industrial development. Energy efficiency directly affects production costs and environmental impact; therefore, optimizing energy consumption has become a core research topic in intelligent manufacturing. With the widespread application of IoT technology, workshops can collect production data in real time, providing new possibilities for energy consumption optimization. However, most existing methods rely on static energy consumption models or single optimization objectives, making it difficult to adapt to the complex and ever-changing environment and task requirements during production, resulting in low energy utilization efficiency.

[0003] While current research has made some progress in energy consumption optimization, it generally suffers from limitations. Many solutions are designed only for specific equipment or a single production stage, lacking consideration for the multi-task coordination across the entire workshop. This makes it difficult for the system to achieve optimal global energy consumption allocation when faced with changes in production tasks or fluctuations in equipment status. Furthermore, existing methods often ignore the uncertainties in energy consumption, such as the impact of production task adjustments or changes in environmental factors, leading to a disconnect between optimization results and actual needs. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an energy consumption optimization method for a robot-based intelligent manufacturing workshop based on the Internet of Things, comprising:

[0005] By using IoT technology to acquire production task data, robot location data, and equipment operating status data from workshop sensors, a real-time dataset is generated.

[0006] The real-time dataset is classified, and production tasks are divided into high-priority task groups and low-priority task groups based on task priority and energy consumption parameters.

[0007] Based on the task classification results, a path planning algorithm is selected to plan the robot's path and generate a path set;

[0008] Extract path data from the path set, combine it with equipment operating status data, calculate the equipment power-on / off sequence, and generate a timing scheme.

[0009] For the aforementioned timing scheme, predict the energy consumption uncertainty caused by task changes and equipment aging, and generate a predicted distribution of the energy consumption fluctuation range;

[0010] Based on the predicted distribution, the path set and timing scheme are adjusted to generate an optimized path and timing combination.

[0011] Extract global energy consumption data from the optimized path and timing combination, iteratively adjust task allocation and equipment operating parameters, and generate a task scheduling scheme with optimal global energy consumption.

[0012] The globally energy-optimal task scheduling scheme is verified, and the energy consumption parameters are updated based on the latest data to generate the final adaptive energy consumption optimization scheme.

[0013] The task execution sequence and equipment operating parameters in the final adaptive energy consumption optimization scheme are transmitted to the workshop control system to complete the real-time deployment of energy consumption optimization.

[0014] Preferably, the process of generating the real-time dataset includes:

[0015] If the production task data includes a task identifier and a deadline, then the task priority is calculated based on the deadline to obtain a task priority sequence.

[0016] Based on the task priority sequence and robot position data, calculate the optimal path from the current position to the task target point and generate path planning data;

[0017] Analyze equipment status data, calculate the energy consumption parameters of each device in path planning, and obtain a set of energy consumption parameters;

[0018] If the task priority, path planning, or energy consumption parameters in the real-time dataset exceed a preset threshold, sensor data will be reacquired using IoT technology to update the real-time dataset.

[0019] Preferably, the process of dividing production tasks into high-priority task groups and low-priority task groups based on task priority and energy consumption parameters includes:

[0020] Obtain the real-time production task dataset, extract task priority and energy consumption parameters, standardize the data format, and obtain the preprocessed dataset;

[0021] The preprocessed dataset is classified using the K-means clustering algorithm to determine high-priority task groups and low-priority task groups, thus obtaining initial classification results.

[0022] If there is a mismatch between task priority and energy consumption parameters in the initial classification results, the classification will be readjusted according to the preset priority threshold and energy consumption threshold to obtain an optimized classification result.

[0023] Preferably, the process of generating the path set includes:

[0024] If the task classification results show that the proportion of high-priority task groups exceeds the preset threshold, then the high-priority task list is obtained from the task classification module to determine the trigger path planning conditions.

[0025] Based on the high-priority task list, obtain task location data and workshop layout information, and generate a set of spatial coordinates for the task locations.

[0026] The A* search algorithm is used to process the set of spatial coordinates of the task location, and combined with the spatial constraints of the workshop layout, the set of shortest paths is obtained.

[0027] If there are path conflicts in the shortest path set, the path set is analyzed by the conflict detection module to identify conflicting path pairs, and the Dijkstra algorithm is used to replan the conflicting paths to obtain an optimized path set.

[0028] If the task classification results show that the proportion of high-priority task groups does not exceed the preset threshold, the population is initialized through a genetic algorithm to obtain an initial path set containing multiple paths.

[0029] The individual paths in the initial path set are encoded and represented by a sequence of path nodes;

[0030] If a path individual satisfies the preset fitness function, then the path individual is retained; otherwise, a new path individual is generated through crossover and mutation operations.

[0031] Based on the new path individuals generated by the crossover and mutation operations, a selection strategy is used to determine the set of path individuals to be retained;

[0032] If the set of retained path individuals satisfies the convergence condition, the optimized candidate path set is output; otherwise, the crossover and mutation operations are returned to generate new path individuals.

[0033] Preferably, the process of generating the timing scheme includes:

[0034] Path data is obtained from the path set, and the path data set is obtained by parsing the path nodes and connection relationships;

[0035] The path data set is combined with the device operating status data to generate a fused dataset;

[0036] If the fused dataset meets the preset integrity threshold, the device power-on / off timing is optimized using a linear programming algorithm to obtain the optimized timing result.

[0037] Based on the optimized timing results, the device startup time and shutdown time are extracted to generate a timing data table;

[0038] Based on the time series data table, time series analysis methods are used to detect outliers in the time series data and obtain corrected time series data.

[0039] Using the corrected timing data, a timing scheme containing startup and shutdown times is generated.

[0040] Preferably, the process of generating the predicted distribution of energy consumption fluctuation range includes:

[0041] Monte Carlo simulations are used to obtain multiple sets of randomly sampled scenarios from time-series task changes and equipment aging data to generate an initial energy consumption uncertainty distribution;

[0042] If the variance of the initial energy consumption uncertainty distribution exceeds a preset threshold, the sampling parameters are adjusted through iterative Monte Carlo simulation to obtain a converged energy consumption uncertainty distribution.

[0043] Based on the converged energy consumption uncertainty distribution, the probability density function of the energy consumption fluctuation range is generated by kernel density estimation, and the energy consumption fluctuation range is determined.

[0044] By integrating the probability density function, the upper and lower limits of the energy consumption fluctuation range are obtained, and a prediction distribution is generated.

[0045] Preferably, the process of generating the optimized path and timing combination includes:

[0046] If the energy consumption fluctuation range of the predicted distribution exceeds the preset threshold, the distribution characteristics are obtained from historical energy consumption data through the data acquisition module to obtain the energy consumption fluctuation trend.

[0047] Based on the energy consumption fluctuation trend, a reinforcement learning algorithm is used to iteratively adjust the path set and generate candidate path combinations.

[0048] The reinforcement learning algorithm is used to optimize the temporal scheme, obtain temporal combinations that match the candidate path combinations, and determine the optimized path and temporal combination.

[0049] Preferably, the process of generating a globally energy-optimal task scheduling scheme includes:

[0050] Global energy consumption data is obtained from the task allocation and equipment operating parameters, and an initial energy consumption dataset is generated using a data extraction process.

[0051] The global energy consumption data is analyzed using the particle swarm optimization algorithm, and an iterative adjustment process is performed on the task allocation scheme and equipment operating parameters to determine the optimized task allocation scheme and operating parameter adjustment combination.

[0052] If the optimized task allocation scheme meets the preset energy consumption threshold, then the task execution path is adjusted by path optimization combination to obtain path optimization combination;

[0053] If the energy consumption data of the path optimization combination is lower than the initial energy consumption dataset, the task execution timing is adjusted through time-series optimization combination to determine the time-series optimization combination;

[0054] The particle swarm optimization algorithm is used to jointly iterate the path optimization combination and the timing optimization combination, and the optimal combination of operating parameters is generated by adjusting the operating parameters, thus obtaining the operating parameter adjustment;

[0055] By integrating path optimization, timing optimization, and operational parameter adjustment into the task scheduling scheme, a task scheduling scheme with optimal energy consumption is generated.

[0056] On the other hand, the present invention also provides an electronic device including a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the processor implements the method when executing the computing program.

[0057] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method.

[0058] Compared with the prior art, the present invention has the following advantages and technical effects:

[0059] This invention discloses an industrial production energy consumption optimization method based on the Internet of Things (IoT) and artificial intelligence (AI). The method collects real-time workshop data using IoT technology, classifies tasks using K-means clustering, performs path planning based on priority using either the A* algorithm or a genetic algorithm, and generates equipment timing schemes using linear programming. Furthermore, this invention employs Monte Carlo simulation to predict energy consumption uncertainty, adjusting the scheme through reinforcement learning when necessary. Finally, it uses particle swarm optimization to iteratively optimize global energy consumption and generates an adaptive optimization scheme based on the latest data. This invention achieves intelligent and adaptive task scheduling and energy consumption optimization in industrial production processes, effectively improving production efficiency and reducing energy consumption. Attached Figure Description

[0060] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0061] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention. Detailed Implementation

[0062] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0063] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0064] Example 1

[0065] like Figure 1 As shown, this embodiment provides an energy consumption optimization method for a robot-based intelligent manufacturing workshop based on the Internet of Things, including:

[0066] By using IoT technology to acquire production task data, robot location data, and equipment operating status data from workshop sensors, a real-time dataset containing task priorities, path requirements, and energy consumption parameters is generated.

[0067] IoT technology is used to acquire production task, robot position, and equipment status data from workshop sensors, generating a real-time dataset containing task priorities, path planning, and energy consumption parameters. If the production task data includes a task identifier and deadline, the task priority is calculated based on the deadline, resulting in a task priority sequence. Based on the task priority sequence and robot position data, the A* algorithm is used to calculate the optimal path from the current position to the task target point, generating path planning data. Operating power and load are analyzed using equipment status data to calculate the energy consumption parameters of each device in the path planning, resulting in a set of energy consumption parameters. If the task priorities, path planning, or energy consumption parameters in the real-time dataset exceed preset thresholds, sensor data is reacquired using IoT technology to update the real-time dataset. Based on the updated real-time dataset, the K-means algorithm is used to cluster the task priorities and energy consumption parameters to determine the task execution order and equipment allocation scheme. Using the path planning data and equipment allocation scheme, control commands are generated and issued to the robots and equipment to obtain the task execution results.

[0068] The K-means clustering algorithm was used to classify the real-time dataset. Based on task priority and energy consumption parameters, the production tasks were divided into high-priority task groups and low-priority task groups to obtain the task classification results.

[0069] A real-time production task dataset is acquired, and task priority and energy consumption parameters are extracted. The data format is standardized to obtain a preprocessed dataset. K-means clustering is used to classify the preprocessed dataset, determining high-priority and low-priority task groups based on task priority and energy consumption parameters, resulting in initial classification results. If mismatches exist between task priority and energy consumption parameters in the initial classification results, the classification is readjusted based on preset priority and energy consumption thresholds to obtain optimized classification results. By analyzing the optimized classification results, the average priority and average energy consumption of each task group are calculated to obtain task group characteristic parameters. Based on the task group characteristic parameters, if the average energy consumption of a high-priority task group exceeds a preset threshold, the tasks in that group are sorted to determine the priority ranking. A scheduling algorithm is used to allocate tasks according to the priority ranking results, combined with real-time system resource status, to obtain a task execution sequence. By monitoring the running status of the task execution sequence, the dynamic changes in priority and energy consumption are recorded, and the real-time dataset is updated to obtain the updated classification criteria.

[0070] Specifically, in real-time production task classification scenarios, it is first necessary to collect task feature data, such as priority scores (range 1-10, e.g., task A has a priority of 8.5) and energy consumption parameters (e.g., task B has an energy consumption of 3.2 kW·h). Through data standardization, the original data with a priority mean of 6.3 and a standard deviation of 2.1 is converted to a standard normal distribution using the Z-score method. When initializing K-means clustering, k=2 is set, and the k-means++ algorithm is used to select initial centroids. For example, in the first iteration, the centroids of the high-priority group are [0.8, -0.5], and the centroids of the low-priority group are [-0.6, 0.3]. During the iteration process, Euclidean distance is calculated. When the feature vector of task C is [0.7, -0.4], its distance to the high-priority centroid (1.2) is less than its distance to the low-priority centroid (2.8), so it is assigned to the high-priority group. After 5 iterations, the algorithm terminates when the distance the center point moves is less than the threshold of 0.01, yielding the following classification results: the high-priority group contains 12 tasks with a mean feature priority of 8.1±0.9 and energy consumption of 4.5±1.2 kW·h; the low-priority group contains 18 tasks with a mean priority of 4.3±1.5 and energy consumption of 2.1±0.8 kW·h. The good clustering effect is verified by the silhouette coefficient evaluation (overall value 0.62).

[0071] If the task classification results show that the proportion of high-priority task groups exceeds the preset threshold of 0.6, then the robot path is planned based on the A algorithm, and the shortest path set is generated by combining the workshop layout and task location.

[0072] If the task classification results show that the proportion of high-priority task groups exceeds a preset threshold, a list of high-priority tasks is retrieved from the task classification module to determine the triggering conditions for path planning. Based on the high-priority task list, task location data and workshop layout information are obtained to generate a set of spatial coordinates for the task locations. The A* algorithm is used to process the set of spatial coordinates for the task locations, combined with the spatial constraints of the workshop layout, to obtain a set of shortest paths. If path conflicts exist in the shortest path set, the conflict detection module analyzes the path set to identify conflicting path pairs. Based on the conflicting path pairs, Dijkstra's algorithm is used to replan the conflicting paths, resulting in an optimized path set. By optimizing the path set, a robot path allocation scheme is generated to determine the execution path for each robot. Based on the robot path allocation scheme, path execution time data is obtained to assess the execution efficiency of the path set.

[0073] Specifically, in the task classification results, if the proportion of high-priority task groups exceeds a preset threshold of 0.6, the system will automatically trigger Algorithm A for robot path planning. Assume the current workshop layout is a 10m x 10m rectangular area, with task positions at coordinates (2,3), (5,7), and (8,2). Algorithm A first calculates the Manhattan distances to each task point from the robot's current position (0,0), which are 5, 12, and 10 respectively. Through heuristic search using Algorithm A, the system generates the shortest path from (0,0) to (2,3) as [(0,0)→(1,0)→(2,0)→(2,1)→(2,2)→(2,3)], with a path length of 5. Next, the system plans the path from (2,3) to (5,7), generating [(2,3)→(3,3)→(4,3)→(5,3)→(5,4)→(5,5)→(5,6)→(5,7)], with a path length of 7. Finally, the system plans the path from (5,7) to (8,2), generating [(5,7)→(6,7)→(7,7)→(8,7)→(8,6)→(8,5)→(8,4)→(8,3)→(8,2)], with a path length of 9. Through iterative optimization using Algorithm A, the system finally generates the shortest path set.

[0074] Otherwise, a set of candidate paths containing diverse paths is generated by optimizing the path based on a genetic algorithm.

[0075] A genetic algorithm is used to initialize the population, obtaining an initial path set containing multiple paths. Individual encoding is represented by a sequence of path nodes, and encoded path individuals are obtained from the initial path set. If a path individual satisfies a preset fitness function, it is retained; otherwise, new path individuals are generated through crossover and mutation operations. Based on the new path individuals generated by crossover and mutation operations, a selection strategy is used to determine the set of path individuals to be retained. If the set of retained path individuals meets the convergence condition, an optimized candidate path set is output; otherwise, crossover and mutation operations are returned to generate new path individuals. From the optimized candidate path set, paths that satisfy the diversified path conditions are selected to generate the final path set. For the final path set, a path optimization algorithm is used to fine-tune the paths, obtaining the final optimized diversified path set.

[0076] Specifically, when optimizing paths using a genetic algorithm, the population first needs to be initialized. For example, 50 paths are randomly generated, each containing 10 nodes with coordinates ranging from 0 to 100 in a two-dimensional space. The fitness function is designed as the sum of the path length and the obstacle collision penalty. The path length is calculated using Euclidean distance, and the collision penalty is set to increase the cost by 100 units per collision. A roulette wheel selection strategy is used, retaining the top 30% of individuals in fitness for the next generation. The crossover operation uses a two-point crossover method, randomly selecting the middle four nodes of two parent paths for swapping, with a crossover probability of 0.8. The mutation operation randomly replaces the coordinates of one node in the path with a probability of 0.05, with a mutation range of ±5 units. After 100 generations of iteration, 10 optimal paths are selected to form a candidate set, where the shortest path length is 245.7 units, the longest path length is 278.3 units, and the average difference between paths is 15.2 units. By calculating the cosine similarity between paths, it is ensured that the similarity within the set is all below 0.3, meeting the diversity requirements. Path smoothing was achieved using cubic B-spline interpolation with a control point spacing of 8 units, resulting in a maximum curvature of no more than 0.15 for the final executable path. Energy consumption modeling showed that the optimized path saved an average of 12.3% more energy than the initial path.

[0077] Extract path data from the shortest path set or candidate path set, combine it with equipment operating status data, calculate the equipment start-up and shutdown timing using a linear programming algorithm, and generate a timing scheme that includes start-up and shutdown times.

[0078] Path data is obtained from the candidate path set. By parsing path nodes and connection relationships, a path data set is obtained. Data fusion processing is used to combine the path data set with device operating status data to generate a fused dataset. If the fused dataset meets a preset integrity threshold, the device power-on / off timing is optimized using a linear programming algorithm to obtain an optimized timing result; otherwise, the process returns to the path data extraction step to obtain data again. Based on the optimized timing result, device startup and shutdown times are extracted to generate a timing data table. According to the timing data table, time series analysis methods are used to detect outliers in the timing data, resulting in corrected timing data. Using the corrected timing data, a timing scheme containing startup and shutdown times is generated. Key time nodes are extracted from the timing scheme, stored in a database, and the final timing scheme file is generated.

[0079] Specifically, when extracting path data from the shortest path set, assume the candidate path set contains three paths with lengths of 120 meters, 150 meters, and 180 meters respectively. Dijkstra's algorithm calculates the shortest path to be 120 meters, passing through nodes A, B, and C, corresponding to device numbers DEV001, DEV002, and DEV003. Device operating status data includes: DEV001's startup power is 5.2kW, with a shutdown delay of 30 seconds; DEV002's startup power is 3.8kW, with a shutdown delay of 45 seconds; and DEV003's startup power is 4.5kW, with a shutdown delay of 60 seconds. This data is input into a linear programming model, with the objective function set as minimizing total energy consumption. Constraints include a maximum parallel power limit of 10kW for the devices and a task completion time not exceeding 300 seconds. Solving using the simplex method, the startup time of DEV001 is 0 seconds and the shutdown time is 120 seconds; the startup time of DEV002 is 30 seconds and the shutdown time is 165 seconds; and the startup time of DEV003 is 75 seconds and the shutdown time is 225 seconds. Timing scheme verification shows that the total energy consumption is (5.2×120+3.8×135+4.5×150) / 3600=0.52kWh, satisfying all constraints.

[0080] For time-series schemes, Monte Carlo simulation algorithm is used to predict energy consumption uncertainties caused by task changes and equipment aging, and a predicted distribution including the range of energy consumption fluctuations is generated.

[0081] Multiple sets of randomly sampled scenarios are obtained from time-series task changes and equipment aging data using Monte Carlo simulation to generate an initial energy consumption uncertainty distribution. If the variance of the initial energy consumption uncertainty distribution exceeds a preset threshold, the sampling parameters are adjusted through iterative Monte Carlo simulation to obtain a converged energy consumption uncertainty distribution. Based on the converged energy consumption uncertainty distribution, kernel density estimation is used to generate a probability density function for the energy consumption fluctuation range, thus determining the energy consumption fluctuation range. By integrating the probability density function, the upper and lower limits of the energy consumption fluctuation range are obtained, generating a predicted distribution. If the confidence interval width of the predicted distribution exceeds a preset threshold, performance degradation features are extracted from the equipment aging data to update the input parameters of the Monte Carlo simulation, resulting in an optimized predicted distribution. Based on the optimized predicted distribution, quantile analysis is used to extract key quantiles of the energy consumption fluctuation range, determining the final energy consumption prediction interval. By visualizing the boundary values ​​of the final energy consumption prediction interval, a distribution map of the energy consumption fluctuation range is generated, and the quantification results are output.

[0082] Specifically, in time-series energy consumption prediction, the first step is to construct an equipment aging model. For example, a Weibull distribution can be used to describe the equipment failure rate, with a shape parameter β = 2.5 and a scale parameter η = 10000 hours, to simulate the equipment performance degradation curve. By collecting historical operating data (e.g., the monthly average temperature of a wind turbine bearing gradually increases from 40℃ to 65℃), a linear regression model of temperature and energy consumption is established: y = 0.8x + 120, where x is temperature, y is energy consumption (kWh), and R0 is the energy consumption. 2 =0.92. Next, a Monte Carlo simulation framework was designed, with 10,000 iterations. In each iteration, the remaining device lifetime (based on a Weibull distribution) and task load fluctuation (normal distribution N(μ = 150kW, σ)) were randomly sampled. 2 =25). In each simulation, the sampled equipment state parameters are input into an LSTM neural network (128 hidden layers, dropout = 0.2), which outputs the energy consumption sequence for the next 24 hours, while simultaneously adding a task load disturbance term ΔP ~ N(0, 3.2). Finally, all simulation results are aggregated to generate a prediction band with an 80% confidence interval. For example, the baseline prediction value of a data center cooling system at t+12h is 215kWh, with a fluctuation range of [208, 223]kWh. To quantify the sources of uncertainty, Sobol sensitivity analysis is used, which reveals that task load variation contributes 62%, equipment aging contributes 31%, and model error contributes 7%.

[0083] If the energy consumption fluctuation range of the predicted distribution exceeds the preset threshold of 10%, the path set and time series scheme are adjusted through reinforcement learning algorithm to generate an optimized path and time series combination.

[0084] If the predicted energy consumption fluctuation range exceeds a preset threshold, the data acquisition module obtains distribution characteristics from historical energy consumption data to determine the energy consumption fluctuation trend. Based on this trend, a reinforcement learning algorithm iteratively adjusts the path set to generate candidate path combinations. The reinforcement learning algorithm then optimizes the time-series scheme to obtain time-series combinations that match the candidate path combinations, thus determining the optimized path and time-series combination. If the energy consumption fluctuation range of the optimized path and time-series combination still exceeds the preset threshold, a genetic algorithm is used to perform a secondary optimization of the path set and time-series scheme, resulting in an updated path and time-series combination. Based on the updated path and time-series combination, a simulation module predicts energy consumption fluctuations to obtain the predicted energy consumption distribution. If the predicted energy consumption distribution fluctuation range does not exceed the preset threshold, the output module generates the final path and time-series combination, determining the optimization result. The storage module associates the optimization result with historical energy consumption data, updates the distribution characteristics of the data acquisition module, and obtains a new energy consumption fluctuation trend.

[0085] Otherwise, retain the current path and timing scheme to obtain a stable energy consumption allocation scheme.

[0086] Initial energy consumption allocation data is obtained through a preset path and timing scheme to determine the energy consumption allocation scheme. If the initial energy consumption allocation data exceeds a preset energy balance threshold, a linear regression algorithm is used to optimize the path scheme, resulting in an optimized path scheme. Based on the optimized path scheme, the timing scheme is adjusted, new timing data is obtained, and an updated timing scheme is determined. Using the updated timing scheme and the optimized path scheme, resource scheduling is performed to obtain the scheduled energy consumption allocation data and determine the energy balance status. If the scheduled energy consumption allocation data meets a preset stability analysis threshold, real-time energy consumption data is obtained through the energy consumption monitoring module to determine the stability of the energy consumption allocation. Based on the real-time energy consumption data, a K-means clustering algorithm is used to classify the energy consumption allocation schemes, resulting in classified energy consumption allocation schemes. Using the classified energy consumption allocation schemes, a scheme evaluation is performed to obtain the final stable energy consumption allocation scheme and determine the stable scheme.

[0087] Specifically, by introducing an energy consumption optimization algorithm while retaining the current path and timing scheme, a stable energy consumption allocation scheme can be obtained. First, a dynamic programming-based energy consumption allocation algorithm is used to limit the total system energy consumption to within 1000W, while ensuring that the energy consumption requirements of each subsystem are met.

[0088] Global energy consumption data is extracted from the optimized path and timing combination. The particle swarm optimization algorithm is used to iteratively adjust the task allocation and equipment operating parameters to generate a task scheduling scheme with optimal global energy consumption.

[0089] Global energy consumption data is obtained from task allocation and equipment operating parameters. An initial energy consumption dataset is generated through a data extraction process, resulting in global energy consumption data. The global energy consumption data is analyzed using a particle swarm optimization (PSO) algorithm. An iterative adjustment process is performed on the task allocation scheme and equipment operating parameters to determine the optimized task allocation scheme and operating parameter adjustment combination. If the optimized task allocation scheme meets a preset energy consumption threshold, the task execution path is adjusted using path optimization combination, resulting in a path optimization combination. If the energy consumption data of the path optimization combination is lower than the initial energy consumption dataset, the task execution timing is adjusted using time-series optimization combination, resulting in a time-series optimization combination. The PSO algorithm is used to jointly iterate on the path optimization combination and the time-series optimization combination, generating the energy-optimal operating parameter combination based on the operating parameter adjustments, resulting in the operating parameter adjustment. The path optimization combination, time-series optimization combination, and operating parameter adjustment are integrated through a task scheduling scheme to generate a task scheduling scheme with optimal energy consumption. If the global energy consumption data of the task scheduling scheme does not reach the preset energy consumption threshold, the PSO algorithm is returned to re-execute the iterative adjustment process to determine the final energy-optimal task scheduling scheme.

[0090] Specifically, when extracting global energy consumption data by optimizing path and time series combinations, the system first collects equipment operating parameters through sensors. For example, a motor on a production line has a power of 5.2 kW and a speed of 1500 rpm, while a temperature sensor records an ambient temperature of 25 degrees Celsius. This data is uploaded to the energy consumption analysis system in real time via an IoT platform. The system uses a sliding window algorithm to smooth the time series data, with a window size of 10 seconds and a step size of 1 second, to eliminate noise interference and obtain a stable energy consumption curve. Next, the Dynamic Time Warping (DTW) algorithm is used to align the energy consumption time series data of different devices, calculate a similarity matrix, and identify energy consumption anomalies. For example, the instantaneous energy consumption of a device suddenly increases to 7.8 kW at 2 PM, deviating from the baseline value by 35%. Based on this data, a multiple linear regression model is used to construct the energy consumption model, with equipment power, operating time, and ambient temperature as independent variables. The energy consumption function is fitted as E = 2.3P + 0.8T + 0.5θ, where P is power, T is operating time, θ is the temperature coefficient, and R is the model determination coefficient. 2The efficiency reached 0.92. During the particle swarm optimization phase, 50 particles were initialized, each representing a task allocation scheme. For example, the particle position vector [0.6, 0.3, 0.1] indicates that 60% of the tasks are allocated to device A, 30% to device B, and 10% to device C. The fitness function is defined as total energy consumption E = Σ(E_i × t_i), where E_i is the energy consumption per unit time of the i-th device, and t_i is the running time. During iteration, the global optimal solution is updated by comparing the historical best of the particles with the swarm optimal solution. For example, in the 20th iteration, the optimal solution was found to reduce the total energy consumption to 42.7 kWh, a reduction of 18.6% compared to the initial scheme. The final generated scheduling scheme prioritizes high-load tasks for device C (energy efficiency ratio 3.2 kWh / ton), which has the best energy efficiency ratio, and adjusts its operating parameters to 85% of its rated speed. At this point, the overall system energy consumption reaches the global minimum of 39.4 kWh, while also satisfying the production cycle time constraint.

[0091] The task scheduling scheme with the optimal global energy consumption is verified by the real-time data processing module. The energy consumption parameters are updated by combining the latest data collected by the Internet of Things, and the final adaptive energy consumption optimization scheme is generated.

[0092] Real-time data streams are acquired through IoT devices, and stream processing technology is used to parse the data to obtain energy consumption parameters. If the parsed energy consumption parameters do not match a preset threshold, the parameters are adjusted based on the deviation to determine the updated energy consumption parameters. The updated energy consumption parameters are loaded using a task scheduling module, and task allocation is optimized using a genetic algorithm to obtain a preliminary task scheduling scheme. For the preliminary task scheduling scheme, global energy consumption data is acquired, and simulation technology is used to verify the energy consumption effect and determine the degree of optimization. If the degree of optimization is lower than a preset threshold, the task allocation strategy is adjusted, and a new scheduling scheme is generated using a genetic algorithm to obtain an improved task scheduling scheme. Based on the improved task scheduling scheme, a dynamic programming algorithm is used to generate an adaptive energy consumption optimization scheme to determine the final scheme. The running status of the optimization scheme is continuously monitored through real-time data streams, feedback data is obtained, and energy consumption parameters are updated to obtain cyclically optimized energy consumption parameters.

[0093] The task execution sequence and equipment operating parameters are extracted from the final adaptive energy consumption optimization scheme and transmitted to the workshop control system to complete the real-time deployment of energy consumption optimization.

[0094] The system obtains task execution sequences and equipment operating parameters from the energy consumption optimization scheme. It then uses a preset optimization algorithm to generate optimized task sequences and parameter configurations, resulting in the task execution sequences and equipment operating parameters. These are transmitted to the workshop control system via a data transmission interface. Data integration is completed using a system integration protocol, confirming data transmission completion. If the workshop control system receives the task execution sequences and equipment operating parameters, it determines whether the parameters meet the current production scheduling requirements based on equipment status data and issues parameter adjustment instructions. The equipment operating parameters are updated according to the parameter adjustment instructions, and the energy consumption monitoring module collects equipment status data in real time, obtaining energy consumption monitoring data. If the energy consumption monitoring data shows that energy consumption exceeds a preset threshold, the task execution sequences and equipment operating parameters are regenerated using the optimization algorithm to determine a new optimization scheme. This new optimization scheme is transmitted to the workshop control system via the system integration interface, and the updated task execution sequences and equipment operating parameters are deployed in real time, resulting in a real-time deployment status. Based on the real-time deployment status, the workshop control system's operating logs are collected, and data analysis tools are used to extract anomaly information and assess system stability.

[0095] Specifically, in the adaptive energy consumption optimization scheme, the task execution sequence in the workshop is first optimized using a genetic algorithm. The population size is set to 100, the number of iterations to 50, the crossover probability to 0.8, and the mutation probability to 0.1, resulting in the optimal task sequence [Task A, Task B, Task C, Task D]. Next, the equipment operating parameters are adjusted using a particle swarm optimization algorithm. The number of particles is set to 30, the number of iterations to 40, the inertia weight to 0.7, and the learning factors c1 and c2 to 1.5. After optimization, the operating speed of equipment A is 1200 rpm, the temperature of equipment B is set to 85℃, and the power of equipment C is 2.5 kW. Then, the energy consumption before and after optimization is compared using an energy consumption model. Before optimization, the total energy consumption was 1500 kWh, and after optimization, the total energy consumption was 1200 kWh, a reduction of 20%. Finally, the optimized task execution sequence and equipment operating parameters are transmitted to the workshop control system via the OPCUA protocol, enabling real-time deployment of energy consumption optimization.

[0096] On the other hand, this embodiment also provides an electronic device, including a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the processor implements the method when executing the computing program.

[0097] On the other hand, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method.

[0098] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for optimizing energy consumption in a robot-based intelligent manufacturing workshop based on the Internet of Things, characterized in that, include: By using IoT technology to acquire production task data, robot location data, and equipment operating status data from workshop sensors, a real-time dataset is generated. The real-time dataset is classified, and production tasks are divided into high-priority task groups and low-priority task groups based on task priority and energy consumption parameters. Based on the task classification results, a path planning algorithm is selected to plan the robot's path and generate a path set; Extract path data from the path set, combine it with equipment operating status data, calculate the equipment power-on / off sequence, and generate a timing scheme. For the aforementioned timing scheme, predict the energy consumption uncertainty caused by task changes and equipment aging, and generate a predicted distribution of the energy consumption fluctuation range; Based on the predicted distribution, the path set and timing scheme are adjusted to generate an optimized path and timing combination. Extract global energy consumption data from the optimized path and timing combination, iteratively adjust task allocation and equipment operating parameters, and generate a task scheduling scheme with optimal global energy consumption. The globally energy-optimal task scheduling scheme is verified, and the energy consumption parameters are updated based on the latest data to generate the final adaptive energy consumption optimization scheme. The task execution sequence and equipment operating parameters in the final adaptive energy consumption optimization scheme are transmitted to the workshop control system to complete the real-time deployment of energy consumption optimization.

2. The method according to claim 1, characterized in that, The process of generating the real-time dataset includes: If the production task data includes a task identifier and a deadline, then the task priority is calculated based on the deadline to obtain a task priority sequence. Based on the task priority sequence and robot position data, calculate the optimal path from the current position to the task target point and generate path planning data; Analyze equipment status data, calculate the energy consumption parameters of each device in path planning, and obtain a set of energy consumption parameters; If the task priority, path planning, or energy consumption parameters in the real-time dataset exceed a preset threshold, sensor data will be reacquired using IoT technology to update the real-time dataset.

3. The method according to claim 1, characterized in that, The process of dividing production tasks into high-priority task groups and low-priority task groups based on task priority and energy consumption parameters includes: Obtain the real-time production task dataset, extract task priority and energy consumption parameters, standardize the data format, and obtain the preprocessed dataset; The preprocessed dataset is classified using the K-means clustering algorithm to determine high-priority task groups and low-priority task groups, thus obtaining initial classification results. If there is a mismatch between task priority and energy consumption parameters in the initial classification results, the classification will be readjusted according to the preset priority threshold and energy consumption threshold to obtain an optimized classification result.

4. The method according to claim 1, characterized in that, The process of generating the path set includes: If the task classification results show that the proportion of high-priority task groups exceeds the preset threshold, then the high-priority task list is obtained from the task classification module to determine the trigger path planning conditions. Based on the high-priority task list, obtain task location data and workshop layout information, and generate a set of spatial coordinates for the task locations. The A* search algorithm is used to process the set of spatial coordinates of the task location, and combined with the spatial constraints of the workshop layout, the set of shortest paths is obtained. If there are path conflicts in the shortest path set, the path set is analyzed by the conflict detection module to identify conflicting path pairs, and the Dijkstra algorithm is used to replan the conflicting paths to obtain an optimized path set. If the task classification results show that the proportion of high-priority task groups does not exceed the preset threshold, the population is initialized through a genetic algorithm to obtain an initial path set containing multiple paths. The individual paths in the initial path set are encoded and represented by a sequence of path nodes; If a path individual satisfies the preset fitness function, then the path individual is retained; otherwise, a new path individual is generated through crossover and mutation operations. Based on the new path individuals generated by the crossover and mutation operations, a selection strategy is used to determine the set of path individuals to be retained; If the set of retained path individuals satisfies the convergence condition, the optimized candidate path set is output; otherwise, the crossover and mutation operations are returned to generate new path individuals.

5. The method according to claim 4, characterized in that, The process of generating the timing scheme includes: Path data is obtained from the path set, and the path data set is obtained by parsing the path nodes and connection relationships; The path data set is combined with the device operating status data to generate a fused dataset; If the fused dataset meets the preset integrity threshold, the device power-on / off timing is optimized using a linear programming algorithm to obtain the optimized timing result. Based on the optimized timing results, the device startup time and shutdown time are extracted to generate a timing data table; Based on the time series data table, time series analysis methods are used to detect outliers in the time series data and obtain corrected time series data. Using the corrected timing data, a timing scheme containing startup and shutdown times is generated.

6. The method according to claim 1, characterized in that, The process of generating the predicted distribution of energy consumption fluctuation range includes: Monte Carlo simulations are used to obtain multiple sets of randomly sampled scenarios from time-series task changes and equipment aging data to generate an initial energy consumption uncertainty distribution; If the variance of the initial energy consumption uncertainty distribution exceeds a preset threshold, the sampling parameters are adjusted through iterative Monte Carlo simulation to obtain a converged energy consumption uncertainty distribution. Based on the converged energy consumption uncertainty distribution, the probability density function of the energy consumption fluctuation range is generated by kernel density estimation, and the energy consumption fluctuation range is determined. By integrating the probability density function, the upper and lower limits of the energy consumption fluctuation range are obtained, and a prediction distribution is generated.

7. The method according to claim 1, characterized in that, The process of generating the optimized path and timing combination includes: If the energy consumption fluctuation range of the predicted distribution exceeds the preset threshold, the distribution characteristics are obtained from historical energy consumption data through the data acquisition module to obtain the energy consumption fluctuation trend. Based on the energy consumption fluctuation trend, a reinforcement learning algorithm is used to iteratively adjust the path set and generate candidate path combinations. The reinforcement learning algorithm is used to optimize the temporal scheme, obtain temporal combinations that match the candidate path combinations, and determine the optimized path and temporal combination.

8. The method according to claim 1, characterized in that, The process of generating a globally energy-optimal task scheduling scheme includes: Global energy consumption data is obtained from the task allocation and equipment operating parameters, and an initial energy consumption dataset is generated using a data extraction process. The global energy consumption data is analyzed using the particle swarm optimization algorithm, and an iterative adjustment process is performed on the task allocation scheme and equipment operating parameters to determine the optimized task allocation scheme and operating parameter adjustment combination. If the optimized task allocation scheme meets the preset energy consumption threshold, then the task execution path is adjusted by path optimization combination to obtain path optimization combination; If the energy consumption data of the path optimization combination is lower than the initial energy consumption dataset, the task execution timing is adjusted through time-series optimization combination to determine the time-series optimization combination; The particle swarm optimization algorithm is used to jointly iterate the path optimization combination and the timing optimization combination, and the optimal combination of operating parameters is generated by adjusting the operating parameters, thus obtaining the operating parameter adjustment; By integrating path optimization, timing optimization, and operational parameter adjustment into the task scheduling scheme, a task scheduling scheme with optimal energy consumption is generated.

9. An electronic device comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, characterized in that, When the processor executes the computing program, it implements the method of any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-8.

Images