Robot energy-saving system and method for intelligent management and control of cold end of indirect cooling auxiliary unit

Abstract

The invention relates to the technical field of fin cleaning, in particular to a robot energy-saving system and method for intelligent control over the cold end of an indirect cooling auxiliary unit, and the system comprises the steps that a fin array surface temperature abnormal area and a stain coverage area are extracted, and a self-adaptive grid is divided according to the stain density; defining a local cleaning strategy and a global cleaning strategy based on the adaptive grid; generating a spiral traversal path based on an adaptive grid and a global cleaning strategy, and optimizing the shortest advancing path of the robot; and the robot moves in the working space according to the shortest advancing path, the local cleaning strategy is executed in sequence, and intelligent energy-saving cleaning of the fins of the indirect cooling auxiliary unit is achieved.

Description

Technical Field

[0001] The present application relates to the field of fin cleaning technology, and in particular to a robot energy-saving system and method for intelligent control of the cold end of an indirect cooling auxiliary unit. Background Art

[0002] With the rapid development of industrial automation and intelligent manufacturing, industrial cleaning robots are widely used in various industrial scenarios, including cleaning the fin arrays of condenser fins in indirect cooling units. These condensers are common equipment in industrial production, and their cooling efficiency directly impacts the energy efficiency of the entire system. However, after long-term operation, dirt accumulates on the fin surfaces, resulting in reduced heat transfer efficiency and increased energy consumption. Timely and effective cleaning can restore the heat transfer performance of the condenser fins and reduce energy consumption.

[0003] Traditional manual cleaning methods are inefficient, labor-intensive, and difficult to ensure consistent cleaning results. Industrial cleaning robots can achieve automated, intelligent cleaning, improving cleaning efficiency and quality. However, simply adopting fixed cleaning paths and strategies cannot adapt to the uneven and complex distribution of dirt on the fin surface. This can lead to over-cleaning or under-cleaning, resulting in poor cleaning results and increased energy consumption.

[0004] Therefore, how to adaptively generate the optimal cleaning path and strategy based on the dirt distribution characteristics of the fin surface to minimize cleaning energy consumption while ensuring the cleaning effect has become a key technical problem that needs to be solved urgently. Summary of the Invention

[0005] The present application aims to solve, at least to some extent, one of the technical problems in the related art. To this end, one purpose of the present application is to propose a robotic energy-saving system and method for intelligent control of the cold end of an indirect cooling auxiliary unit, thereby realizing intelligent energy-saving cleaning of the fins of the indirect cooling auxiliary unit.

[0006] One aspect of the present application provides a robotic energy-saving method for intelligently controlling the cold end of an indirect cooling auxiliary unit, comprising:

[0007] Step S100: extracting the abnormal temperature area and the stain coverage area on the surface of the fin array, and dividing the adaptive grid according to the stain density;

[0008] Step S200: defining a local cleaning strategy and a global cleaning strategy based on an adaptive grid;

[0009] Step S300: Generate a spiral traversal path based on the adaptive grid and the global cleaning strategy to optimize the shortest travel path of the robot;

[0010] Step S400: The robot moves in the workspace according to the shortest travel path and executes the local cleaning strategy in sequence.

[0011] The specific method of extracting the abnormal temperature area and the stain coverage area on the fin array surface and dividing the adaptive grid according to the stain density is as follows:

[0012] Step S110: mapping the fin array to the robot workspace and dividing the robot workspace into an initial grid;

[0013] Step S120: obtaining the temperature distribution matrix of the fin array and extracting the temperature abnormality area using a threshold segmentation algorithm;

[0014] Step S130: Acquire a fin surface image and perform stain recognition on it to form a stain distribution matrix;

[0015] Step S140: Calculate the stain density ρ of each initial grid cell (i, j) based on the stain distribution of pixels in each initial grid cell. ij ;

[0016] Step S150: Preset upper and lower limits of the stain density threshold, divide the initial grid cells into stain density levels according to the stain density, adaptively adjust the grid size, connect and merge isolated grid cells, and obtain adaptive grid cells.

[0017] The specific method of presetting the upper and lower limits of the stain density threshold, dividing the initial grid cells into stain density levels according to the stain density, and adaptively adjusting the grid size is as follows:

[0018] Preset upper and lower limits of the stain density threshold, determine the initial grid cells with a stain density greater than or equal to the upper limit of the stain density threshold as a dense stain area, determine the initial grid cells with a stain density less than the lower limit of the stain density threshold as a sparse stain area, and determine the initial grid cells with a stain density greater than or equal to the lower limit of the stain density threshold and less than the upper limit of the stain density threshold as a normal area;

[0019] Reduce the mesh size of the stain-dense area to the initial mesh size. times, enlarge the grid size of the sparse stain area to the initial grid size times, the grid size of the regular area remains unchanged; the stain-intensive area, the stain-sparse area and the regular area are collectively referred to as stain density levels; the grid unit is obtained by adaptively adjusting the grid size.

[0020] The specific method for connecting and merging isolated grid cells to obtain adaptive grid cells is as follows: if two adjacent grid cells have the same stain density level, the two are determined to be connected in terms of stain distribution, and whether there are connected grid cells with the same stain density level in their four neighborhoods. If so, they are merged into one area, and the process is repeated until no isolated grid cells to be merged exist, thereby obtaining K adaptive grid cells.

[0021] The specific method of defining the local cleaning strategy and the global cleaning strategy based on the adaptive grid is:

[0022] Step S210: For each adaptive grid cell, extract its grid center coordinates, average temperature and stain coverage as the attribute vector of the adaptive grid cell

[0023] Step S220: Obtain the real-time pressure change Δp(t) and the robot displacement deviation ε(t), and construct a state vector S(t) in combination with the attribute vector of the adaptive grid unit; t represents the time;

[0024] Step S230: generating a local cleaning strategy for each adaptive grid unit according to its attribute vector, and generating a global cleaning strategy based on the local cleaning strategy.

[0025] The method for extracting the average temperature and stain coverage rate is as follows: identifying a temperature anomaly area in each adaptive grid unit, and calculating the average temperature of the adaptive grid unit based on the pixel temperature of the temperature anomaly area; identifying the number of pixels in the stain coverage area in each adaptive grid unit, and calculating the stain coverage rate of the adaptive grid unit based on the stain distribution in the stain coverage area.

[0026] The specific method of generating a local cleaning strategy for each adaptive grid unit according to its attribute vector and generating a global cleaning strategy based on the local cleaning strategy is as follows:

[0027] Step S231: For the local cleaning strategy, define its state space as the attribute vector of each adaptive grid Action space a k are the cleaning intensity and cleaning time, and the reward function r is the cleaning energy consumption;

[0028] Step S232: For the global cleaning strategy, define its state space as the total vector S formed by concatenating the state vectors of all adaptive grid cells, the action space A as the cleaning order of all adaptive grid cells, and the reward function R as the overall cleaning effect and total energy consumption; the total energy consumption includes the total cleaning energy consumption and the inter-grid transfer cost;

[0029] Step S233: using a reinforcement learning algorithm to learn an optimal cleaning strategy, where the optimal cleaning strategy includes an optimal local cleaning strategy and an optimal global cleaning strategy.

[0030] The specific method of using the reinforcement learning algorithm to learn the optimal cleaning strategy includes: mapping the continuous average temperature and stain coverage rate into discrete state codes; establishing a mapping table between the state codes, cleaning intensity, and cleaning time based on the discrete state codes; and implementing the reinforcement learning algorithm as a lookup table operation based on the mapping table to find the optimal cleaning strategy;

[0031] The specific method of mapping the continuous average temperature and stain coverage into discrete state codes is: dividing each component equally and arranging the quantized results of all components into state codes in sequence; the components are the average temperature and the stain coverage.

[0032] The specific method of establishing a mapping table between state codes and cleaning intensities and cleaning times based on discrete state codes is as follows: placing the robot in a simulation environment, trying different cleaning intensities and cleaning times for different state codes, evaluating the cleaning effects of the different cleaning intensities and cleaning times, and forming a mapping table;

[0033] According to the mapping table, the reinforcement learning algorithm is implemented as a lookup table operation. Finding the optimal cleaning strategy refers to: for the local cleaning strategy, for the state space of each time step, quantizing each attribute vector component in the state space to obtain a state code, decoding the state code into the row number of the mapping table, and finding the column number with the element value of 1 in the row corresponding to the row number of the mapping table. The cleaning intensity and cleaning time of the corresponding row in the column are the optimal local cleaning strategy corresponding to the attribute vector. On the basis of the local cleaning strategy, for the entire robot workspace, considering the transfer cost between adjacent grids and the total energy consumption of cleaning, the optimal cleaning sequence is learned to obtain the optimal global cleaning strategy.

[0034] The specific method for generating a spiral traversal path based on the adaptive grid and the global cleaning strategy and optimizing the shortest travel path of the robot is as follows:

[0035] Step S310: generating a spiral traversal path covering the entire robot working area according to the cleaning sequence obtained by the global cleaning strategy and the boundary of the robot working space;

[0036] Step S320: For each adaptive grid cell, calculate the intersection point set between its boundary and the spiral traversal path; and extract the local path segment inside the adaptive grid cell based on the intersection point set;

[0037] Step S330: For each adaptive grid unit, based on its local path segments, maximizing the cleaning efficiency index and maximizing the path smoothness index are used as sub-optimization objectives, and an overall optimization objective for the local path segments is defined. The overall optimization objective is solved using a dynamic programming algorithm to obtain the optimized local path segments. The optimized local path segments are smoothly spliced ​​according to the cleaning order of the adaptive grid units to obtain the shortest travel path for the robot;

[0038] Specifically, the method for obtaining the optimized local path segment is as follows: for each adaptive grid unit, firstly, the local path segment Discretize into a series of points Among them, q k,dRepresents a local path segment The dth discrete point on m k Represents a local path segment The total number of discrete points on the path; define two indicators, cleaning efficiency index and path smoothness index, and define the total optimization target of the local path segment based on the two indicators; use the dynamic programming algorithm to solve the total optimization target, and define the state f(d) as the starting point q k,1 To point q k,d The objective function value of the optimal path is obtained, and the state transfer equation is obtained by recursively solving the state transfer equation. The objective function value of the optimal path is obtained by f(m k ), and the optimal predecessor point of each point, and restore the optimized local path segment based on the optimal predecessor point

[0039] Specifically, the splicing method of the shortest travel path is: according to the cleaning order in the global cleaning strategy, the optimized local path segments of each adaptive grid unit are taken out in turn; for two adjacent local path segments, the shortest connection path between the endpoints of the two adjacent local path segments is calculated as a transition segment; the local path segments and the transition segments are alternately spliced ​​to obtain the complete shortest travel path of the robot.

[0040] One aspect of the present application provides a robotic energy-saving system for intelligent cold-end control of an indirect cooling auxiliary unit, comprising:

[0041] Adaptive grid generation module, used to extract the temperature anomaly area and stain coverage area on the fin array surface, and divide the adaptive grid according to the stain density;

[0042] A multi-level strategy generation module for defining local and global cleaning strategies based on adaptive grids;

[0043] The shortest path optimization module is used to generate a spiral traversal path based on an adaptive grid and a global cleaning strategy to optimize the robot's shortest travel path;

[0044] The cleaning strategy execution module is used for the robot to move in the workspace according to the shortest travel path and execute local cleaning strategies in sequence.

[0045] The robot energy-saving system and method for intelligent cold-end control of indirect cooling auxiliary units proposed in this application have the following advantages over the existing technology:

[0046] This application adaptively divides the grid according to the stain distribution density on the fin surface, using fine-grained grids in stain-dense areas and coarse-grained grids in stain-sparse areas, reducing redundant traversal in stain-sparse areas and avoiding unnecessary energy consumption; at the same time, it ensures the cleaning effect in stain-dense areas and avoids secondary cleaning energy consumption caused by insufficient cleaning.

[0047] This application uses a reinforcement learning algorithm to learn energy-saving local cleaning strategies and global cleaning strategies by designing a reasonable reward function. The learned strategies minimize cleaning energy consumption while meeting the cleaning effect. Compared with fixed strategies, they are more flexible and efficient.

[0048] This application discretizes the continuous state space, constructs a mapping table between states and actions, and converts policy learning into a table lookup operation, which reduces the computational complexity of policy learning, improves training efficiency and policy execution efficiency, and indirectly reduces energy consumption.

[0049] This application generates a spiral traversal path based on a global cleaning strategy, extracts local path segments, and uses dynamic programming to smoothly optimize and splice them together to obtain the shortest travel path. While ensuring global coverage, the spiral traversal path reduces unnecessary repeated traversals and saves travel energy. The smooth optimization of local paths reduces sharp turns, improves movement efficiency, and further reduces energy consumption.

[0050] This application uses cleaning performance and energy consumption as optimization targets, while also considering multiple sources of information, such as temperature distribution and stain distribution, to achieve optimal decision-making under multiple objectives and constraints. By comprehensively balancing cleaning performance and energy consumption, this approach minimizes energy consumption while still meeting cleaning requirements, achieving energy savings and efficiency gains. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flow chart of the method for the robotic energy-saving method for intelligent cold-end control of the indirect cooling auxiliary unit provided in this application;

[0052] Figure 2 Flowchart of the method for generating the global cleaning strategy and local cleaning strategy provided in this application;

[0053] Figure 3 A flow chart of the method for generating the shortest travel path of a robot provided in this application;

[0054] Figure 4 Schematic diagram of the robot workspace provided for this application;

[0055] Figure 5 Schematic diagram of the initial grid division provided for this application;

[0056] Figure 6 Schematic diagram of adaptively adjusting grid size provided for this application;

[0057] Figure 7 Functional module diagram of the robotic energy-saving system for intelligent cold-end control of the indirect cooling auxiliary unit provided in this application. DETAILED DESCRIPTION

[0058] To better understand the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely descriptions of exemplary embodiments of the present application and are not intended to limit the scope of the present application in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0059] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The accompanying drawings are for illustration only and are not drawn strictly to scale. As used herein, the terms "substantially," "approximately," and similar terms are used to indicate approximate values, not degrees, and are intended to illustrate inherent deviations in measurements or calculations that would be recognized by a person of ordinary skill in the art. In addition, in this application, the order in which the steps are described does not necessarily represent the order in which these steps would occur in actual operation, unless otherwise specified or inferred from the context.

[0060] It should also be understood that expressions such as "including", "comprising", "having", "containing" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present application, "may" is used to mean "one or more embodiments of the present application". And, the term "exemplary" is intended to refer to an example or illustration.

[0061] Unless otherwise defined, all words used herein (including engineering terms and scientific and technological terms) have the same meaning as commonly understood by those skilled in the art to which this application belongs. It should also be understood that, unless otherwise specified in this application, words defined in commonly used dictionaries should be interpreted as having the same meaning as they do in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0062] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0063] Example 1

[0064] like Figure 1 As shown, the robot energy-saving method for intelligent control of the cold end of the indirect cooling auxiliary unit provided in this application includes:

[0065] Step S100: extracting the abnormal temperature area and the stain coverage area on the surface of the fin array, and dividing the adaptive grid according to the stain density;

[0066] The specific method of extracting the abnormal temperature area and the stain coverage area on the fin array surface and dividing the adaptive grid according to the stain density is as follows:

[0067] Step S110: mapping the fin array to the robot workspace and dividing the robot workspace into an initial grid;

[0068] Figure 4 This is a schematic diagram of the robot workspace provided in this application. The outer rectangular frame represents the robot work area, which is composed of multiple rectangles. Each rectangle represents a fin, and all rectangles form a fin array.

[0069] Figure 5 The initial grid division diagram provided in this application, the initial grid refers to the grid unit before the adaptive grid division, the initial grid unit is a grid unit with uniform size, the robot workspace is an L×W rectangular area, divided into n r row n c The initial grid of columns.

[0070] Step S120: Obtain the temperature distribution matrix T of the fin array and extract the temperature abnormal area A through the threshold segmentation algorithm temp ;

[0071] The temperature distribution matrix is ​​obtained by scanning the fin array in all directions with an infrared thermal imager. The temperature distribution matrix can be expressed as T = [T ij ], where T ij represents the temperature measured on the initial grid cell (i, j);

[0072] The threshold segmentation algorithm refers to the preset temperature segmentation threshold T thres The temperature segmentation threshold is dynamically set according to the material properties and the ambient temperature. The area with a temperature greater than the temperature segmentation threshold is marked as a temperature abnormal area. The temperature abnormal area can be expressed as A temp ={(x,y)|T(x,y)>T thres}; (x, y) represents the pixel coordinates, T(x, y) represents the temperature at the pixel coordinates (x, y);

[0073] Step S130: obtaining a fin surface image and performing stain recognition on it to form a stain distribution matrix M;

[0074] The fin surface image is acquired by a high-definition CCD camera, and an optimized surface image is obtained by image enhancement and denoising. The stain recognition model is trained based on a random forest classifier, and the grayscale mean, variance and texture features of the image block are used as input features x. input , the optimized surface image is divided into N1 image blocks, the input features of each image block are extracted and input into the stain recognition model for stain recognition, and the stain distribution of the fin surface image is output to form the stain distribution matrix M = [m ij ], where m ij =1 means the initial grid cell (i, j) is covered by stains, m ij is the stain distribution value of the initial grid cell (i, j). The stain distribution value is 1 or 0, indicating stain coverage and no stain coverage, respectively.

[0075] Step S140: Calculate the stain density ρ of each initial grid cell (i, j) based on the stain distribution m(x, y) of the pixels in each initial grid cell. ij ;

[0076] The stain density ρ of each initial grid cell (i, j) ij The calculation formula is: Among them, G ij is the pixel coordinate set of the initial grid unit (i, j), (x, y) represents the pixel coordinate, m(x, y) represents the stain distribution value of the pixel coordinate (x, y) in the initial grid unit, N ij is the total number of pixels of the initial grid cell (i, j);

[0077] Step S150: Preset upper and lower limits of the stain density threshold, divide the initial grid cells into stain density levels according to the stain density, adaptively adjust the grid size, connect and merge isolated grid cells, and obtain adaptive grid cells;

[0078] The specific method of presetting the upper and lower limits of the stain density threshold, dividing the initial grid cells into stain density levels according to the stain density, and adaptively adjusting the grid size is as follows:

[0079] Preset upper and lower limits of the stain density threshold, determine the initial grid cells with a stain density greater than or equal to the upper limit of the stain density threshold as a dense stain area, determine the initial grid cells with a stain density less than the lower limit of the stain density threshold as a sparse stain area, and determine the initial grid cells with a stain density greater than or equal to the lower limit of the stain density threshold and less than the upper limit of the stain density threshold as a normal area;

[0080] Reduce the mesh size of the stain-dense area to the initial mesh size. times, enlarge the grid size of the sparse stain area to the initial grid size times, the grid size of the regular area remains unchanged; the stain-intensive area, the stain-sparse area and the regular area are collectively referred to as stain density levels; the grid unit is obtained by adaptively adjusting the grid size.

[0081] Figure 6 This is a schematic diagram of the adaptive adjustment of the grid size provided in this application, where the uneven black-gray area represents the stain, and the local grid size is adaptively adjusted according to the density of the stain.

[0082] The upper and lower limits of the stain density threshold are set by those skilled in the art based on experience. Preferably, the upper limit of the stain density threshold is 1.5, and the lower limit of the stain density threshold is 0.6;

[0083] The values ​​of the upper and lower limits of the stain density threshold are determined with reference to the "SB / T 10738-2012 Specification for Professional Cleaning Services for Dirt on the Surface of Heat Exchanger Fins of Air Conditioning Equipment", which classifies contamination into mild (stain density threshold <15%), moderate (stain density threshold between 15% and 60%), and severe (stain density threshold >60%).

[0084] The specific method for connecting and merging isolated grid cells to obtain adaptive grid cells is as follows: if two adjacent grid cells have the same stain density level, the two are determined to be connected in terms of stain distribution, and whether there are connected grid cells with the same stain density level in their four neighborhoods. If so, they are merged into one area, and the process continues until no isolated grid cells are left to be merged, thereby obtaining K adaptive grid cells.

[0085] The adaptive grid units include regular rectangular adaptive grid units and irregular adaptive grid units.

[0086] The above steps use a fine-grained grid in the area with dense stains and a coarse-grained grid in the area with sparse stains, which not only ensures the cleaning effect but also reduces unnecessary traversal, saving cleaning time and energy consumption.

[0087] Step S200: defining a local cleaning strategy and a global cleaning strategy based on an adaptive grid;

[0088] The specific method of defining the local cleaning strategy and the global cleaning strategy based on the adaptive grid is:

[0089] like Figure 2 As shown, step S210: for each adaptive grid unit, extract its grid center coordinates, average temperature and stain coverage as the attribute vector of the adaptive grid unit

[0090] The grid center coordinates are extracted as follows: for a regular rectangular adaptive grid unit, the grid center coordinates are the coordinates of the intersection of the rectangular diagonals; for an irregular adaptive grid unit, the centroid coordinates are used as the grid center coordinates;

[0091] The calculation formula of the centroid coordinates is: Among them, |G k | is the kth adaptive grid unit G k The area of ​​​​the pixel, (x, y) is the pixel coordinate, x k and y k are the horizontal and vertical coordinates of the centroid coordinates respectively;

[0092] The method for extracting the average temperature and stain coverage ratio is as follows: identifying a temperature anomaly area in each adaptive grid unit, and calculating the average temperature of the adaptive grid unit based on the pixel temperatures of the temperature anomaly area; identifying the number of pixels in the stain coverage area in each adaptive grid unit, and calculating the stain coverage ratio of the adaptive grid unit based on the stain distribution in the stain coverage area;

[0093] The average temperature The calculation formula is: in, represents the temperature of the nth pixel in the temperature anomaly area of ​​the kth adaptive grid unit, n k Represents the total number of pixels in the temperature anomaly area in the adaptive grid cell;

[0094] The stain coverage The calculation formula is: Where (x,y)∈G k represents the kth adaptive grid unit G k Pixel coordinates in N k represents the total number of pixels in the kth adaptive grid unit;

[0095] The attribute vector of the adaptive grid cell can be expressed as

[0096] Step S220: Obtain the real-time pressure change Δp(t) and the robot displacement deviation ε(t), and construct a state vector S(t) in combination with the attribute vector of the adaptive grid unit; t represents the time;

[0097] The real-time pressure change refers to the change in pipeline pressure relative to the initial pressure value at each moment;

[0098] The robot displacement deviation refers to the deviation vector of the robot's position and posture relative to the expected value at each moment, which is used to measure the accuracy and stability of the robot's motion control and can be expressed as ε(t) = [Δx(t), Δy(t), Δθ(t)] T , [·] T represents the transpose of the matrix, Δx(t), Δy(t), and Δθ(t) are the posture deviations of the robot in the three degrees of freedom of X, Y, and θ at time t respectively;

[0099] The state vector is expressed as: in, represents the attribute vector of the first adaptive grid cell at time t, represents the attribute vector of the Kth adaptive grid cell at time t, Δp(t) represents the real-time pressure change at time t, ||ε(t)||2 represents the L2 norm of the robot displacement deviation at time t,

[0100] The state vector integrates four types of measured quantities: temperature, contamination, pressure, and posture, comprehensively reflecting the real-time operating status of the robot system. Different physical quantities have different numerical ranges and variation characteristics, so different calculation indicators are used to extract the state information contained in them. Finally, the four components are arranged in order to provide a basis for subsequent strategy judgment.

[0101] Step S230: generating a local cleaning strategy for each adaptive grid unit according to its attribute vector, and generating a global cleaning strategy based on the local cleaning strategy;

[0102] The specific method of generating a local cleaning strategy for each adaptive grid unit according to its attribute vector and generating a global cleaning strategy based on the local cleaning strategy is as follows:

[0103] Step S231: For the local cleaning strategy, define its state space as the attribute vector of each adaptive grid Action space a k are the cleaning intensity and cleaning time, and the reward function r is the cleaning energy consumption;

[0104] Step S232: For the global cleaning strategy, define its state space as the total vector S formed by concatenating the state vectors of all adaptive grid cells, the action space A as the cleaning order of all adaptive grid cells, and the reward function R as the overall cleaning effect and total energy consumption; the total energy consumption includes the total cleaning energy consumption and the inter-grid transfer cost;

[0105] Step S233: using a reinforcement learning algorithm to learn an optimal cleaning strategy, wherein the optimal cleaning strategy includes an optimal local cleaning strategy and an optimal global cleaning strategy;

[0106] Preferably, the reinforcement learning algorithm adopts the Q-learning algorithm. For the local cleaning strategy, its value function is expressed as The update formula of the value function is: Among them, α is the learning rate, γ is the discount factor, and a k 'For the next action to be performed, To perform action a k The next state after Indicates the status is Next, perform action a k The reward function, Indicates that the status The value function of executing the next action, Indicates that the value function Maximize the next move, represents the current value function, Represents the value function of the next step;

[0107] The learning rate is used to control the proportion of current experience in the update, and its value range is 0 to 1. The learning rate for laboratory and industrial reinforcement learning tasks is usually set to α≈0.1.

[0108] The discount factor is used to weigh the impact of immediate rewards and future rewards, and its value range is 0 to 1. The value of the discount factor for energy-saving tasks without termination discount is usually γ∈[0.95,0.99];

[0109] For the global cleaning strategy, its value function is expressed as Q(S,A), and the update formula of the value function is: Among them, R t (S, A) is the reward function for executing action A when the current state is the total vector S, S' is the next state after executing action A, A' is the next action to be executed, Q t (S', A') represents the value function of executing the next action A' in state S', Indicates that the value function Q t (S', A') maximizes the next action, Q t+1 (S, A) represents the value function for the next step of the global cleaning strategy, Q t (S, A) represents the current value function for the global cleaning strategy;

[0110] Specifically, the reinforcement learning algorithm is used to learn the optimal cleaning strategy. During the learning process, the ε-greedy strategy is used to balance exploration and utilization. ε is a probability value, that is, the action is randomly selected with a probability of ε, and the action with the largest current value function value is selected with a probability of 1-ε. The initial probability ε0 is 0.2. As the learning progresses, ε gradually decreases, and ε is calculated according to ε. t =max(0.05,0.2×0.995 t ) in a decreasing manner, so that the reinforcement learning model converges to the optimal cleaning strategy. The initial value of ε is generally 0.1-0.2, which is a common initial value for robot reinforcement learning.

[0111] Optionally, the specific method of using the reinforcement learning algorithm to learn the optimal cleaning strategy includes: mapping the continuous average temperature and stain coverage rate into discrete state codes; establishing a mapping table between the state codes, cleaning intensity, and cleaning time based on the discrete state codes; and implementing the reinforcement learning algorithm as a lookup table operation according to the mapping table to find the optimal cleaning strategy;

[0112] Specifically, the specific method of mapping the continuous average temperature and stain coverage into discrete state codes is: dividing each component equally and arranging the quantized results of all components into state codes in sequence; the components are average temperature and stain coverage.

[0113] For example, the average temperature It can be divided into three intervals: "low temperature", "normal temperature" and "high temperature", corresponding to codes 0, 1 and 2 respectively; stain coverage It can be divided into three intervals: "clean", "lightly soiled" and "heavily soiled", with corresponding codes of 0, 1 and 2; a possible status code is [1,2] T It means "normal temperature, heavy stains".

[0114] The specific method of establishing a mapping table between state codes and cleaning intensities and cleaning times based on discrete state codes is as follows: placing the robot in a simulation environment, trying different cleaning intensities and cleaning times for different state codes, evaluating the cleaning effects of the different cleaning intensities and cleaning times, and forming a mapping table;

[0115] Specifically, according to the mapping table, the reinforcement learning algorithm is implemented as a lookup table operation. Finding the optimal cleaning strategy means: for the local cleaning strategy, for the state space of each time step, quantize each attribute vector component in the state space to obtain a state code, decode the state code into the row number of the mapping table, and find the column number with the element value 1 in the row corresponding to the row number of the mapping table. The cleaning intensity and cleaning time of the corresponding row in the column are the optimal local cleaning strategy corresponding to the attribute vector. On the basis of the local cleaning strategy, for the entire robot workspace, considering the transfer cost between adjacent grids and the total energy consumption of cleaning, learn the optimal cleaning order and obtain the optimal global cleaning strategy.

[0116] Step S300: Generate a spiral traversal path based on the adaptive grid and the global cleaning strategy to optimize the shortest travel path of the robot;

[0117] like Figure 3 As shown, the specific method of generating a spiral traversal path based on the adaptive grid and the global cleaning strategy and optimizing the shortest travel path of the robot is as follows:

[0118] Step S310: generating a spiral traversal path covering the entire robot working area according to the cleaning sequence obtained by the global cleaning strategy and the boundary of the robot working space;

[0119] Optionally, the spiral traversal path adopts a model such as an Archimedean spiral or a Fermat spiral, or a customized spiral model is designed according to actual needs;

[0120] Step S320: For each adaptive grid cell, calculate the intersection point set between its boundary and the spiral traversal path; and extract the local path segment inside the adaptive grid cell based on the intersection point set;

[0121] Step S330: For each adaptive grid unit, generate an optimized local path segment based on its local path segment, and smoothly splice the optimized local path segments according to the cleaning order of the adaptive grid units to obtain the shortest travel path of the robot;

[0122] The specific method for generating an optimized local path segment for each adaptive grid unit according to its local path segment is as follows: taking maximizing the cleaning efficiency index and maximizing the path smoothness index as sub-optimization objectives, defining the overall optimization objective of the local path segment, and using a dynamic programming algorithm to solve the overall optimization objective to obtain the optimized local path segment.

[0123] For example, there is a robot workspace consisting of four adaptive grid units. After generating a global cleaning strategy, the cleaning order of the four adaptive grid units is G1→G2→G4→G3, and the actual travel path of the robot is optimized according to this cleaning order.

[0124] First, a spiral traversal path covering the entire area is generated based on the global cleaning strategy and the boundary of the robot workspace; the Archimedean spiral is selected as the basic model. The parametric equation of the Archimedean spiral is: Where a and b are the starting radius and radius growth rate, respectively, and τ is a parameter. By adjusting the value ranges of a, b, and τ, a spiral path can be generated that expands outward from the center of the workspace. x(τ) and y(τ) represent the coordinates on the Archimedean spiral at each moment. According to the complete coverage algorithm, a is often set to 0.05m. When the robot cleaning adopts equal pitch, the value range of b is generally 0.02~0.05m·rad -1 ;The parameter value range is τ∈[0,4π], and the unit is rad;

[0125] Secondly, after obtaining the global spiral traversal path, we intersect it with the boundary of each grid cell and extract the local path fragments inside each grid: for each grid cell, we calculate the parametric equation of its boundary; we combine the boundary parametric equation with the parametric equation of the spiral traversal path to solve the parameter value of the intersection point. The parameter value interval of adjacent intersection points [t k,c ,t k,c+1 ]Substitute the parametric equation of the spiral traversal path to obtain the local path segment inside the grid Similarly, local path segments of other grid cells can also be extracted.

[0126] Then, after obtaining the local path segments of each grid cell, these path segments need to be optimized according to their respective local cleaning strategies. With the optimization goal of maximizing cleaning efficiency and path smoothness, a method based on dynamic programming is used to achieve this. Specifically, the method for obtaining the optimized local path segments is as follows: for each adaptive grid cell, first, its local path segments are Discretize into a series of points Among them, q k,d Represents a local path segment The dth discrete point on m k Represents a local path segment The total number of discrete points on the path; define two indicators, cleaning efficiency index and path smoothness index, and define the total optimization target of the local path segment based on the two indicators; use the dynamic programming algorithm to solve the total optimization target, and define the state f(d) as the starting point q k,1 To point q k,d The objective function value of the optimal path is obtained, and the state transfer equation is obtained by recursively solving the state transfer equation to obtain the objective function value of the optimal path f(m k), and the optimal predecessor point of each point, and restore the optimized local path segment based on the optimal predecessor point

[0127] Finally, the optimized local path segments are spliced ​​together according to the cleaning order of each adaptive grid unit to obtain the complete robot shortest travel path. Specifically, the shortest travel path splicing method is as follows: according to the cleaning order in the global cleaning strategy, the optimized local path segments of each adaptive grid unit are sequentially extracted. For two adjacent local path segments, the shortest connecting path between their endpoints is calculated as a transition segment. The local path segments and transition segments are alternately spliced ​​together to obtain the complete robot shortest travel path.

[0128] The cleaning efficiency index E(q k,d ) is calculated as: Among them, M k (q k,d ) is point q k,d Dirt coverage at d(q k,d ,q k,d+1 ) is the distance between two adjacent points. This indicator represents the cleaning effect per unit path length.

[0129] The path smoothness index S(q k,d ,q k,d+1 ) is calculated as: Among them, θ(q k,d ,q k,d+1 ) is the angle between the tangents of two adjacent points. This metric indicates the smoothness of the path; the smaller the angle, the higher the smoothness.

[0130] The overall optimization goal of the local path segment The calculation formula is: Among them, λ'∈[0,1] is the weight coefficient, which is used to balance the cleaning efficiency and path smoothness.

[0131] The state transfer equation is: Among them, f(c) represents the robot's movement from the starting point to the cth point q k,c The objective function value of the optimal path that can be achieved;

[0132] The value of the weight coefficient is set by those skilled in the art based on experience. Preferably, λ' is equal to 0.5.

[0133] Step S400: The robot moves in the workspace according to the shortest travel path and executes the local cleaning strategy in sequence.

[0134] This application achieves energy savings at multiple levels, including cleaning path planning, cleaning strategy optimization, and motion control, through a series of innovations, including adaptive sensing, intelligent decision-making, and optimized execution. Compared to traditional fixed paths and fixed strategies, this application can dynamically adjust according to actual conditions, minimizing energy consumption while ensuring cleaning effectiveness, providing a systematic solution for energy-saving condenser fin cleaning. It also has a certain degree of versatility and can be extended to other types of equipment to address energy-saving issues in surface cleaning.

[0135] Example 2

[0136] like Figure 7 As shown, the robot energy-saving system for intelligent control of the cold end of the indirect cooling auxiliary unit provided in this application includes:

[0137] Adaptive grid generation module, used to extract the temperature anomaly area and stain coverage area on the fin array surface, and divide the adaptive grid according to the stain density;

[0138] A multi-level strategy generation module for defining local and global cleaning strategies based on adaptive grids;

[0139] The shortest path optimization module is used to generate a spiral traversal path based on an adaptive grid and a global cleaning strategy to optimize the robot's shortest travel path;

[0140] The cleaning strategy execution module is used for the robot to move in the workspace according to the shortest travel path and execute local cleaning strategies in sequence.

[0141] The methods, apparatus, and devices of the present application may be implemented in many ways. For example, the methods, apparatus, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the methods according to the present application. Therefore, the present application also covers recording media that store programs for executing the methods according to the present application.

[0142] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0143] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A robotic energy-saving method for intelligent control of the cold end of an indirect cooling auxiliary unit, characterized in that: include: Extract the abnormal temperature area and stain coverage area on the fin array surface, and divide the adaptive grid according to the stain density; Define local cleaning strategies and global cleaning strategies based on adaptive grids; Generate spiral traversal paths based on adaptive grid and global cleaning strategy to optimize the shortest travel path of the robot; The robot moves in the workspace according to the shortest travel path and executes the local cleaning strategy in sequence.

2. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 1, characterized in that: The specific method of extracting the abnormal temperature area and the stain coverage area on the fin array surface and dividing the adaptive grid according to the stain density is as follows: Map the fin array to the robot workspace and divide the robot workspace into an initial grid; Obtain the temperature distribution matrix of the fin array and extract the temperature anomaly area using the threshold segmentation algorithm; Obtaining a fin surface image and performing stain identification on it to form a stain distribution matrix; According to the stain distribution of pixels in each initial grid cell, the stain density ρ of each initial grid cell (i, j) is calculated ij ; The upper and lower limits of the stain density threshold are preset, and the initial grid cells are divided into stain density levels according to the stain density. The grid size is adaptively adjusted, and the isolated grid cells are connected and merged to obtain adaptive grid cells.

3. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 2, characterized in that: The specific method of presetting the upper and lower limits of the stain density threshold, dividing the initial grid cells into stain density levels according to the stain density, and adaptively adjusting the grid size is as follows: Preset upper and lower limits of the stain density threshold, determine the initial grid cells with a stain density greater than or equal to the upper limit of the stain density threshold as a dense stain area, determine the initial grid cells with a stain density less than the lower limit of the stain density threshold as a sparse stain area, and determine the initial grid cells with a stain density greater than or equal to the lower limit of the stain density threshold and less than the upper limit of the stain density threshold as a normal area; Reduce the mesh size of the stain-dense area to the initial mesh size. times, enlarge the grid size of the sparse stain area to the initial grid size times, the grid size of the regular area remains unchanged; the stain-intensive area, the stain-sparse area and the regular area are collectively referred to as stain density levels; the grid unit is obtained by adaptively adjusting the grid size.

4. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 3 is characterized in that: The specific method for connecting and merging isolated grid cells to obtain adaptive grid cells is as follows: if two adjacent grid cells have the same stain density level, the two are determined to be connected in terms of stain distribution, and whether there are connected grid cells with the same stain density level in their four neighborhoods. If so, they are merged into one area, and the process is repeated until no isolated grid cells to be merged exist, thereby obtaining K adaptive grid cells.

5. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 4, characterized in that: The specific method of defining the local cleaning strategy and the global cleaning strategy based on the adaptive grid is: For each adaptive grid cell, extract its grid center coordinates, average temperature and stain coverage as the attribute vector of the adaptive grid cell Obtain the real-time pressure change Δp(t) and robot displacement deviation ε(t), and construct the state vector S(t) by combining the attribute vector of the adaptive grid unit; t represents the time; For each adaptive grid unit, a local cleaning strategy is generated according to its attribute vector, and a global cleaning strategy is generated based on the local cleaning strategy.

6. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 5, characterized in that: The method for extracting the average temperature and stain coverage rate is as follows: identifying a temperature anomaly area in each adaptive grid unit, and calculating the average temperature of the adaptive grid unit based on the pixel temperature of the temperature anomaly area; identifying the number of pixels in the stain coverage area in each adaptive grid unit, and calculating the stain coverage rate of the adaptive grid unit based on the stain distribution in the stain coverage area.

7. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 6, characterized in that: The specific method of generating a local cleaning strategy for each adaptive grid unit according to its attribute vector and generating a global cleaning strategy based on the local cleaning strategy is as follows: For the local cleaning strategy, the state space is defined as the attribute vector of each adaptive grid. Action space a k are the cleaning intensity and cleaning time, and the reward function r is the cleaning energy consumption; For the global cleaning strategy, its state space is defined as the total vector S formed by concatenating the state vectors of all adaptive grid cells, the action space A is the cleaning order of all adaptive grid cells, and the reward function R is the overall cleaning effect and total energy consumption; the total energy consumption includes the total cleaning energy consumption and the inter-grid transfer cost; A reinforcement learning algorithm is used to learn an optimal cleaning strategy, wherein the optimal cleaning strategy includes an optimal local cleaning strategy and an optimal global cleaning strategy.

8. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 7, characterized in that: The specific method for generating a spiral traversal path based on the adaptive grid and the global cleaning strategy and optimizing the shortest travel path of the robot is as follows: Based on the cleaning sequence obtained by the global cleaning strategy and the boundary of the robot workspace, a spiral traversal path covering the entire robot workspace is generated; For each adaptive grid cell, the intersection point set between its boundary and the spiral traversal path is calculated; based on the intersection point set, the local path segment inside the adaptive grid cell is extracted; For each adaptive grid unit, based on its local path segments, the overall optimization goal of the local path segments is defined, with maximizing the cleaning efficiency index and maximizing the path smoothness index as sub-optimization objectives. The overall optimization goal is solved using a dynamic programming algorithm to obtain the optimized local path segments. The optimized local path segments are smoothly spliced ​​according to the cleaning order of the adaptive grid units to obtain the shortest travel path of the robot. The splicing method of the shortest travel path is as follows: according to the cleaning order in the global cleaning strategy, the optimized local path segments of each adaptive grid unit are taken out in turn. For two adjacent local path segments, the shortest connection path between the endpoints of the two adjacent local path segments is calculated as the transition segment. The local path segments and the transition segments are alternately spliced ​​to obtain the complete shortest travel path of the robot.

9. The robot energy-saving method for intelligent cold-end control of an indirect cooling auxiliary unit according to claim 8, characterized in that: The method for obtaining the optimized local path segment is as follows: for each adaptive grid unit, firstly Discretize into a series of points Among them, q k,d Represents a local path segment The dth discrete point on m k Represents a local path segment The total number of discrete points on the path; define two indicators, cleaning efficiency index and path smoothness index, and define the total optimization target of the local path segment based on the two indicators; use the dynamic programming algorithm to solve the total optimization target, and define the state f(d) as the starting point q k,1 To point q k,d The objective function value of the optimal path is obtained, and the state transfer equation is obtained by recursively solving the state transfer equation. The objective function value of the optimal path is obtained by f(m k ), and the optimal predecessor point of each point, and restore the optimized local path segment based on the optimal predecessor point 10. A robotic energy-saving system for intelligent control of the cold end of an indirect cooling auxiliary unit, which is used to implement the robotic energy-saving method for intelligent control of the cold end of an indirect cooling auxiliary unit according to any one of claims 1 to 9, characterized in that: include: Adaptive grid generation module, used to extract the temperature anomaly area and stain coverage area on the fin array surface, and divide the adaptive grid according to the stain density; A multi-level strategy generation module for defining local and global cleaning strategies based on adaptive grids; The shortest path optimization module is used to generate a spiral traversal path based on an adaptive grid and a global cleaning strategy to optimize the robot's shortest travel path; The cleaning strategy execution module is used for the robot to move in the workspace according to the shortest travel path and execute local cleaning strategies in sequence.

Images