Remote control and posture adjustment system for photovoltaic hanging type sweeper

Abstract

The invention discloses a remote control and posture adjustment system for a photovoltaic hanging type cleaning machine, and relates to the technical field of photovoltaic cleaning machine control. Wind load, posture and tension data are fused through state perception alignment, array constraint modeling, swing risk prediction and remote decoupling control; the system is used for evaluating the swing risk and dynamically adjusting the running speed and posture, adaptively setting the brush pressure, the reciprocating frequency and the segmentation path in combination with a component surface pollution image recognition result, and optimizing control parameters in cooperation with multi-machine collaborative scheduling and anomaly analysis. The operation safety, the cleaning effect and the overall operation efficiency of the hanging type cleaning machine can be remarkably improved in a complex multi-dip-angle photovoltaic power station.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic sweeper control technology, and in particular to a remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper. Background Technology

[0002] In recent years, large-scale photovoltaic power plants have gradually been deployed in complex environments such as mountain slopes, Gobi deserts, aquatic surfaces where fisheries and solar power complement each other, and large industrial and commercial rooftops. In these power plants, the modules are arranged in long rows, and there are significant differences in the tilt angle, azimuth angle, row spacing, and frame height of each row of modules. To reduce manual operation and maintenance costs, photovoltaic wall-mounted cleaning machines, which are hung on the ends of the modules or on guide rails, are increasingly used for the daily cleaning of contaminants such as dust, sand, salt spray crystals, and bird droppings on the module surfaces. However, in actual operation, there are problems such as high wind load and sway risks, complex array structures and inconsistent tilt angles, coarse granularity of remote control and insufficient multi-machine coordination, lack of linkage between cleaning intensity and risk, and lack of evidence chain and closed-loop optimization capabilities.

[0003] Currently, Chinese invention patent application number 202311537710.3 discloses a method and system for controlling the operation of a photovoltaic cleaning machine, as well as a storage medium. The photovoltaic cleaning machine has detection components installed at its upper and lower ends. These detection components are used to detect the reference position on the photovoltaic array currently being cleaned by the photovoltaic cleaning machine. The line connecting the upper and lower detection components is perpendicular to the upper and lower edges of the photovoltaic cleaning machine. The motion control method includes: determining whether the photovoltaic cleaning machine is in a tilted or stuck state based on the time it takes for the detection components to detect the reference position; and adjusting the operating state of the photovoltaic cleaning machine when it is in the tilted or stuck state to control its normal operation.

[0004] The aforementioned technologies are insufficient to achieve remote multi-machine collaborative control of photovoltaic wall-mounted sweepers based on multi-source state perception and array geometric constraints, as well as adaptive attitude adjustment and sweeping strategy linkage for swing risk in large-scale photovoltaic power plants with complex wind loads and inconsistent component tilt angles. Summary of the Invention

[0005] The technical problem solved by this invention is that existing technologies are difficult to implement in large-scale photovoltaic power plants with complex wind loads and inconsistent component tilt angles, enabling remote multi-machine collaborative control of photovoltaic wall-mounted sweepers based on multi-source state perception and array geometric constraints, as well as adaptive attitude adjustment for sway risk and linkage with sweeping strategies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper includes a state perception alignment module, an array constraint modeling module, a swing risk prediction module, a remote decoupling control module, a sweeping strategy programming module, and an evidence chain feedback module.

[0008] The state-aware alignment module collects attitude data, tension data, displacement data, and micro-meteorological data and synchronizes them in time to obtain an alignment operation state dataset;

[0009] The array constraint modeling module establishes an array geometric constraint set based on the aligned running state dataset and component arrangement data;

[0010] The swing risk prediction module calculates the swing risk index and generates a set of attitude safety domain parameters by combining the array geometric constraint set.

[0011] The remote decoupling control module outputs remote path commands and attitude compensation commands based on the attitude safety domain parameter set.

[0012] The cleaning strategy orchestration module outputs an adaptive cleaning parameter set based on remote path instructions and attitude compensation instructions;

[0013] The evidence chain feedback module generates analysis results and outputs adjustment signals based on remote path instructions, attitude compensation instructions, adaptive cleaning parameter sets, and alignment operation status datasets.

[0014] Preferably, the state perception alignment module includes an attitude acquisition unit, a suspension tension acquisition unit, a walking displacement acquisition unit, a micro-meteorological acquisition unit, and an alignment processing unit;

[0015] The attitude acquisition unit is used to acquire attitude angle data and angular velocity data based on the IMU and output them as attitude data;

[0016] The suspension tension acquisition unit is used to acquire real-time tension data of the hanging traction structure and output it as tension data.

[0017] The walking displacement acquisition unit is used to acquire displacement data and velocity data along the direction of component travel and output them as displacement data;

[0018] The micro-meteorological acquisition unit is used to collect wind speed data, wind direction data, and gust factor data and output them as micro-meteorological data.

[0019] The alignment processing unit is used to clean and synchronize the attitude data, tension data, displacement data and micro-meteorological data in time, form the original equipment sensing data and output the alignment operation status dataset.

[0020] Preferably, the processing logic of the alignment processing unit is as follows:

[0021] Add a unified timestamp to attitude angle data, real-time tension data, displacement data and wind speed data, remove outliers that exceed the preset physical boundaries, interpolate missing segments to complete them, resample according to the preset sampling period and generate an aligned running status dataset.

[0022] Preferably, the array constraint modeling module includes a component layout acquisition unit, a tilt angle partitioning unit, and a reachable forbidden region generation unit;

[0023] The component layout acquisition unit is used to acquire component row and column boundary data, end buffer distance data and component tilt angle data, and merge the component row and column boundary data, end buffer distance data and component tilt angle data to output component layout data;

[0024] The tilt angle partitioning unit is used to generate tilt angle partitioning identifiers based on component tilt angle data;

[0025] The reachable forbidden region generation unit is used to generate an array geometric constraint set by combining the tilt partition identifier and the alignment running status dataset.

[0026] Preferably, the sway risk prediction module includes a wind load estimation unit, a sway characteristic calculation unit, and a safety domain generation unit;

[0027] The wind load estimation unit is used to calculate wind load parameters based on wind speed data, wind direction data, and gust factor data.

[0028] The swing characteristic calculation unit is used to calculate the swing amplitude index and swing trend index based on attitude angle data, angular velocity data and real-time tension data, and output the swing risk index.

[0029] The safety domain generation unit is used to generate an attitude safety domain parameter set by combining the swing risk index, wind load parameters and array geometric constraint set.

[0030] Preferably, the generation logic of the attitude safety domain parameter set is as follows:

[0031] When the gust factor data is higher than the preset gust threshold, the allowable roll limit and the allowable walking speed limit are reduced to the preset first and second percentages, respectively.

[0032] When the fluctuation amplitude of the real-time tension data exceeds the preset tension fluctuation threshold, the allowable pitch adjustment range is reduced to the preset third percentage.

[0033] When the attitude angle data meets the preset stability conditions, it is restored to the default safe domain parameters of the corresponding tilt angle partition.

[0034] Preferably, the remote decoupling control module includes an outer loop path planning unit, an inner loop attitude compensation unit, and a collaborative scheduling unit;

[0035] The outer loop path planning unit is used to generate remote path instructions based on the array geometric constraint set and the attitude safety domain parameter set.

[0036] The inner ring attitude compensation unit is used to perform anti-sway correction on the remote path command based on the aligned running state dataset and output attitude compensation command.

[0037] The collaborative scheduling unit is used to schedule the remote path instructions and attitude compensation instructions of multiple photovoltaic wall-mounted sweepers in a specific time sequence and output a collaborative control queue.

[0038] Preferably, the cleaning strategy programming module includes a pollution sensing unit, a risk linkage unit, and a parameter generation unit;

[0039] The pollution sensing unit is used to collect image data of the component surface and extract pollution density features and pollution type features to form a pollution assessment feature set;

[0040] The risk linkage unit is used to correlate and calculate the pollution assessment feature set with the swing risk index to generate the cleaning intensity requirement index.

[0041] The parameter generation unit is used to output an adaptive cleaning parameter set based on the cleaning intensity requirement index and the attitude safety domain parameter set. The adaptive cleaning parameter set includes brush pressure level, reciprocating frequency level and segmented path mode.

[0042] Preferably, the evidence chain feedback module includes a log recording unit, an anomaly attribution unit, and a policy update unit;

[0043] The log recording unit is used to associate and store evidence of remote path instructions, attitude compensation instructions, adaptive cleaning parameter sets and alignment running status datasets to form running evidence data.

[0044] The anomaly attribution unit is used to identify attitude over-limit events, tension change events, or cleaning interruption events based on operational evidence data, and output anomaly attribution results.

[0045] The strategy update unit is used to feed back the abnormal attribution results to the swing risk prediction module and the cleaning strategy orchestration module, and output adjustment signals.

[0046] Preferably, the logic for generating the adjustment signal is as follows:

[0047] If the swing risk indicator remains within the preset high risk threshold, a deceleration signal or a pause signal will be output.

[0048] If the attitude angle data exceeds the upper limit of the allowable attitude corresponding to the attitude safety domain parameter set, a recovery signal will be output.

[0049] If the pollution assessment feature set shows persistent pollution and the swing risk index is within the preset low risk threshold, then an enhanced cleaning signal will be output.

[0050] The adjustment signal is sent to the notification terminal.

[0051] The beneficial effects of this invention are as follows: This invention integrates wind load, attitude, and tension data through state-aware alignment, array constraint modeling, sway risk prediction, and remote decoupling control. This data is used to assess sway risk and dynamically adjust operating speed and attitude. Combined with the results of component surface contamination image recognition, the invention adaptively sets brush pressure, reciprocating frequency, and segmented path. Furthermore, it optimizes control parameters through multi-machine collaborative scheduling and anomaly analysis. This can significantly improve the operational safety, cleaning effect, and overall operating efficiency of the wall-mounted sweeper in complex multi-tilt photovoltaic power plants. Attached Figure Description

[0052] Figure 1 This is a basic flowchart of a remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper provided in one embodiment of the present invention. Detailed Implementation

[0053] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0054] Example, refer to Figure 1 This paper presents a remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper, including a state perception alignment module, an array constraint modeling module, a swing risk prediction module, a remote decoupling control module, a sweeping strategy arrangement module, and an evidence chain feedback module.

[0055] The state-aware alignment module collects attitude data, tension data, displacement data, and micro-meteorological data and synchronizes them in time to obtain an alignment operation state dataset.

[0056] The array constraint modeling module establishes an array geometric constraint set based on the aligned running state dataset and component layout data.

[0057] The swing risk prediction module calculates the swing risk index and generates the attitude safety domain parameter set by combining the array geometric constraint set.

[0058] The remote decoupling control module outputs remote path commands and attitude compensation commands based on the attitude safety domain parameter set.

[0059] The cleaning strategy orchestration module outputs an adaptive cleaning parameter set based on remote path instructions and attitude compensation instructions.

[0060] This invention integrates wind load, attitude, and tension data through state-aware alignment, array constraint modeling, sway risk prediction, and remote decoupling control. This data is used to assess sway risk and dynamically adjust operating speed and attitude. Combined with the results of component surface contamination image recognition, the brush pressure, reciprocating frequency, and segmented path are adaptively set. Furthermore, with multi-machine collaborative scheduling and anomaly analysis to optimize control parameters, this invention can significantly improve the operational safety, cleaning effect, and overall work efficiency of wall-mounted sweepers in complex multi-tilt photovoltaic power plants.

[0061] In this embodiment, the remote control and attitude adjustment system for the photovoltaic wall-mounted sweeper is applied to a mountain photovoltaic power station. The power station modules are arranged at various tilt angles, and there are crosswinds and gusts on site year-round. Multiple wall-mounted sweepers run back and forth along the guide rails between the modules.

[0062] The evidence chain feedback module generates analysis results and outputs adjustment signals based on remote path instructions, attitude compensation instructions, adaptive cleaning parameter sets, and alignment operation status datasets.

[0063] The state-aware alignment module includes an attitude acquisition unit, a suspension tension acquisition unit, a walking displacement acquisition unit, a micro-meteorological acquisition unit, and an alignment processing unit.

[0064] The attitude acquisition unit is used to acquire attitude angle data and angular velocity data based on the IMU and output them as attitude data.

[0065] The suspension tension acquisition unit is used to acquire real-time tension data of the suspended traction structure and output it as tension data.

[0066] The walking displacement acquisition unit is used to collect displacement and velocity data along the direction of component travel and output them as displacement data.

[0067] The micro-meteorological acquisition unit is used to collect wind speed data, wind direction data, and gust factor data and output them as micro-meteorological data.

[0068] The alignment processing unit is used to clean and synchronize attitude data, tension data, displacement data and micro-meteorological data in time, forming raw equipment sensing data and outputting an aligned operation status dataset.

[0069] The processing logic of the alignment processing unit is as follows:

[0070] Add a unified timestamp to attitude angle data, real-time tension data, displacement data and wind speed data, remove outliers that exceed the preset physical boundaries, interpolate missing segments to complete them, resample according to the preset sampling period and generate an aligned running status dataset.

[0071] The array constraint modeling module includes a component layout acquisition unit, a tilt angle partitioning unit, and a reachable forbidden region generation unit.

[0072] The component layout acquisition unit is used to acquire component row and column boundary data, end buffer distance data and component tilt angle data, and merge the component row and column boundary data, end buffer distance data and component tilt angle data to output component layout data.

[0073] The tilt angle partitioning unit is used to generate tilt angle partitioning identifiers based on component tilt angle data. The specific process is as follows:

[0074] During the power plant design phase, the design tilt angle data for each row of components is exported from the as-built drawings, denoted as... ;

[0075] During the system's online debugging phase, IMU sensors installed on the wall-mounted sweeper collected multiple frames of attitude angle data as the sweeper moved slowly and uniformly along each row of components, and calculated the estimated actual tilt angle for the corresponding row. ;

[0076] The component tilt angle data are obtained by taking a weighted average of the design value and the measured value. The mathematical expression is:

[0077] ;

[0078] in, To design the weight of tilt angle data in the overall results, The weight of the measured tilt angle data in the overall result, Preferred , .

[0079] To reduce the impact of installation errors and local torsion, the tilt angle data of all components were sorted to obtain an ascending sequence. .

[0080] A one-dimensional clustering algorithm based on a distance threshold is used, with a threshold value set for the tilt angle difference. .

[0081] from Starting as the initial center of the first partition, the tilt angle data of subsequent components are scanned sequentially. :

[0082] like If so, it will be assigned to the current partition;

[0083] like Then the current partition will be terminated, a new partition will be created, and... As the initial center of the new partition.

[0084] in, This represents the average tilt angle data of each component within the current tilt angle partition. It serves as the representative tilt angle for this tilt angle partition and as a comparison benchmark when subsequent component tilt angle data are assigned to the corresponding tilt angle partition.

[0085] This one-dimensional clustering process is equivalent to a simplified implementation of the traditional k-means clustering algorithm in one-dimensional space with a fixed distance threshold, and is a specific application of existing data grouping algorithms.

[0086] The mean of the tilt angle samples within each partition is calculated to obtain the first... Representative value of each partition The mathematical expression is:

[0087] ;

[0088] Among them, is the first The set of indexes contained in each partition. This represents the number of samples.

[0089] according to Sort them from smallest to largest and assign them tilt zone identifiers ID in sequence. The smallest one is "tilt zone identifier 1", the second smallest one is "tilt zone identifier 2", and so on to get "tilt zone identifier K".

[0090] For each component row i, find its corresponding partition. :

[0091] ;

[0092] The guide rail segment corresponding to this component is marked as "tilt zone identifier". This forms the "tilt partition identifier field" in the component layout data. This "tilt partition identifier" will be used as an index in the subsequent generation of attitude safety domain parameter sets and path planning to find the corresponding default attitude reference angle and velocity limit.

[0093] Through the above process, the tilt zoning unit transforms continuous component tilt angle data into a finite number of discrete tilt angle zoning identifiers, thus solving the angle disturbance problem caused by slight installation errors on site.

[0094] The reachable forbidden region generation unit is used to generate an array geometric constraint set by combining the tilt partition identifier and the alignment running status dataset.

[0095] The sway risk prediction module includes a wind load estimation unit, a sway characteristic calculation unit, and a safety domain generation unit.

[0096] The wind load estimation unit is used to calculate wind load parameters based on wind speed data, wind direction data, and gust factor data.

[0097] In this embodiment, the wind load estimation unit calculates wind load parameters based on wind speed data, wind direction data, and gust factor data, using an existing wind load calculation algorithm based on the dynamic pressure formula, specifically:

[0098] For those in the first The component rows of each tilt zone, where the representative value of the tilt angle for that zone is known. and the azimuth angle of the component arrangement direction ;

[0099] Collect on-site wind direction data , expressed as the angle projected onto the horizontal plane;

[0100] Calculate the angle between the wind direction and the component normal direction. The mathematical expression is:

[0101] ;

[0102] Collect wind speed data Calculate the effective wind speed acting in the normal direction of the component. The mathematical expression is:

[0103] ;

[0104] Using air density Combined with the component's shape and wind pressure coefficient (Determined by design or experimental calibration) and gust factor data Calculate normal wind pressure The mathematical expression is:

[0105] ;

[0106] In the sliding time window Inside, for multiple frames Calculate the mean With root mean square value The mathematical expression is:

[0107] , ;

[0108] The normal wind pressure, mean, root mean square (RMS) value, and gust factor data are combined and output as wind load parameters. The normal wind pressure reflects the instantaneous wind pressure, and the RMS value reflects the intensity of wind load fluctuations.

[0109] The sway characteristic calculation unit is used to calculate the sway amplitude index and sway trend index based on attitude angle data, angular velocity data and real-time tension data, and output the sway risk index.

[0110] The swing characteristic calculation unit uses sliding window statistics and exponentially weighted moving average, combined with attitude angle data, angular velocity data, and real-time tension data. The specific logic is as follows:

[0111] In each length of Within the sliding event window, for the roll angle and pitch angle ,calculate:

[0112] Roll angle amplitude The mathematical expression is:

[0113] ;

[0114] Pitch angle amplitude The mathematical expression is:

[0115] ;

[0116] Synthetic swing amplitude The mathematical expression is:

[0117] ;

[0118] Real-time tension data within the same window Calculate the standard deviation :

[0119] ;

[0120] in, The average tension within the window. It reflects the degree of tension fluctuation caused by the oscillation.

[0121] Pre-calibrate the allowable swing amplitude threshold during the debugging phase. and tension fluctuation threshold ;

[0122] Normalize the above quantities:

[0123] , ;

[0124] Swing amplitude index Defined as:

[0125] ;

[0126] in, and The weighting coefficient can be set to 0.5 and 0.5 in this embodiment.

[0127] The rate of change is calculated by using the difference in the swing amplitude index between adjacent sliding windows. The mathematical expression is:

[0128] ;

[0129] High-frequency noise is filtered out using an exponentially weighted moving average (EWMA):

[0130] ;

[0131] in, It is a smoothing coefficient, with a value range of (0,1), preferably 0.3.

[0132] when When, it indicates that the oscillation has a tendency to intensify. When the time is right, it indicates that the oscillation is showing a weakening trend.

[0133] Swing risk indicator Taking into account amplitude, tension fluctuation, and trend, it is defined as:

[0134] ;

[0135] in, , and The weighting coefficients can be calibrated through on-site experiments; in this embodiment, they can be taken as... , , .

[0136] Ultimately By linearly scaling and limiting it to the [0.1] range, it serves as an indicator of swing risk.

[0137] The safety domain generation unit is used to generate an attitude safety domain parameter set by combining the swing risk index, wind load parameters and array geometric constraint set.

[0138] The logic for generating the attitude safety domain parameter set is as follows:

[0139] When the gust factor data is higher than the preset gust threshold, the allowable roll limit and the allowable walking speed limit are reduced to the preset first percentage and second percentage, respectively.

[0140] When the fluctuation amplitude of the real-time tension data exceeds the preset tension fluctuation threshold, the allowable pitch adjustment range is reduced to the preset third percentage.

[0141] When the attitude angle data meets the preset stability conditions, it is restored to the default safe domain parameters of the corresponding tilt angle partition.

[0142] The remote decoupling control module includes an outer loop path planning unit, an inner loop attitude compensation unit, and a collaborative scheduling unit.

[0143] The outer loop path planning unit is used to generate remote path instructions based on the array geometric constraint set and the attitude safety domain parameter set.

[0144] In this embodiment, the outer loop path planning unit generates remote path instructions based on the array geometric constraint set and the attitude safety domain parameter set. It employs a reciprocating coverage path and safety weight adjustment path planning logic, and can be combined with the existing A* path search algorithm. Specifically:

[0145] Each row of component guide rails is arranged in fixed step lengths along the length direction. Discretize them to form several path nodes, denoted as ,in For row index, This is a discrete index on that row;

[0146] For the ends of two adjacent rows, the connecting nodes that can span rows are marked in the array geometric constraint set according to the component row spacing and guide rail structure.

[0147] Based on the tilt zone identifier and attitude safety domain parameter set of the corresponding node, an upper limit for the allowable walking speed is added to each node. And attitude constraint level labels ("wide safety domain", "medium safety domain", "narrow safety domain").

[0148] For nodes located in high-risk areas of gusts or areas of structural interference, they are marked as "forbidden nodes" based on the array geometric constraint set.

[0149] Adopting rules similar to existing reciprocating coverage path planning algorithms:

[0150] Walk from one end to the other in the first row;

[0151] Turn within the end buffer and cross over to the second line;

[0152] Then proceed in reverse on the second line, and so on, covering all lines.

[0153] During the generation process, skip areas marked as "forbidden nodes" and use detours or shortened coverage for such areas.

[0154] Calculate the security cost for each node:

[0155] ;

[0156] in, This can be obtained from the historical statistical value of the most recent swing risk indicator at that position. These are the weighting coefficients.

[0157] Based on the generated reciprocating path, using nodes as the graph structure, the existing A* path search algorithm is used to adjust the local path, prioritizing the node sequence with the lower overall cost, to obtain the optimized node sequence.

[0158] The optimized node sequence is converted into a remote path instruction sequence, with each instruction including: target node index, target displacement, target velocity, and execution timestamp.

[0159] ;

[0160] The generated remote path command is sent to the inner loop attitude compensation unit and the cooperative scheduling unit.

[0161] The inner loop attitude compensation unit is used to perform anti-sway correction on remote path commands based on the aligned running state dataset and output attitude compensation commands.

[0162] In this embodiment, the inner loop attitude compensation unit uses existing proportional-derivative (PD) control algorithms and velocity feedforward algorithms to perform anti-sway correction on remote path commands. The specific process is as follows:

[0163] Obtain the current displacement based on the alignment running state dataset. and current speed According to the target displacement given in the corresponding remote path command and target speed Calculate displacement error With speed error The mathematical expression is:

[0164] , ;

[0165] The component reference attitude angle at this location is obtained from the tilt zone identifier and array geometry constraint set. and Obtain the current roll angle from the aligned running state dataset. and current pitch angle Calculate attitude error The mathematical expression is:

[0166] , ;

[0167] Speed ​​and attitude are adjusted using a PD control law:

[0168] ;

[0169] ;

[0170] ;

[0171] in, , , , , , These are the control parameters that have been calibrated on-site.

[0172] Will Used to correct the traction motor speed, and This is converted into differential control quantities for the left and right wheels or left and right ropes, forming anti-sway posture compensation.

[0173] Based on the current swing risk indicators Further limit the corrected target velocity:

[0174] ;

[0175] ;

[0176] in, This is the intermediate target speed, which, without considering risks, is the target speed that the controller believes should be executed at the current time. The final safe execution speed, that is, the safe speed limit actually issued to the executing agency after considering the current swing risk, if the risk is very low ( (close to 0), then It hardly slows down, especially if the risk is high. It will be significantly smaller The more dangerous the situation, the slower you should walk. As a risk factor, the operating speed is automatically reduced when the swing risk index increases.

[0177] Generate attitude compensation commands :

[0178] ;

[0179] The signal is sent to the actuator to achieve attitude decoupling control.

[0180] The collaborative scheduling unit is used to schedule the remote path instructions and attitude compensation instructions of multiple photovoltaic wall-mounted sweepers in a specific time sequence and output a collaborative control queue.

[0181] In this embodiment, the collaborative scheduling unit uses a time-slot-based greedy scheduling algorithm to schedule the remote path instructions and attitude compensation instructions of multiple sweepers in a specific time sequence. The process is as follows:

[0182] The time axis is divided into discrete time slots of fixed length;

[0183] The expected spatial position of each sweeper in each time slot is predicted. If the predicted minimum distance between two sweepers in the same time slot is less than the preset safety distance, it is determined that there is a potential conflict.

[0184] Multiple sweepers are sorted according to preset priorities (e.g., based on battery level, current altitude, or task urgency).

[0185] A greedy strategy is adopted, starting with the highest priority sweeper and locking its remote path command and attitude compensation command;

[0186] For low-priority sweepers that are in conflict, perform one of the following adjustments to their path instructions during the conflict time slot:

[0187] Reduce the target speed within the corresponding time slot proportionally;

[0188] The instructions in the current time slot are shifted backward by one or more time slots.

[0189] After conflict resolution, the commands from multiple sweepers are merged in chronological order to form a collaborative control queue. :

[0190] ;

[0191] in, Number the sweeper The instruction sequence number. For execution time.

[0192] The collaborative control queue is distributed to each hanging sweeper to achieve multi-level collaborative operation.

[0193] The cleaning strategy programming module includes a pollution sensing unit, a risk linkage unit, and a parameter generation unit.

[0194] The pollution sensing unit is used to collect image data of the component surface and extract pollution density and pollution type features to form a pollution assessment feature set.

[0195] In this embodiment, the contamination sensing unit uses a convolutional neural network (CNN) semantic segmentation algorithm, an existing image processing algorithm, to identify contamination areas on the component surface image. The specific process is as follows:

[0196] An industrial camera is installed at the front end of the hanging sweeper to acquire images of the component surface along the running direction, forming component surface image data. Exposure control and fixed lighting methods are used to reduce the impact of light changes on the recognition results.

[0197] The acquired images are subjected to distortion correction and brightness normalization, and then scaled to a uniform resolution, such as 512×512 pixels.

[0198] A convolutional neural network model based on the U-Net structure is adopted. The preprocessed image is used as input and the class probability of each pixel is output. The pixel classes include: background, normal component surface, dust deposition, bird droppings and salt spray crystals. By thresholding the maximum class probability, a binary or multi-class pollution area mask map is obtained.

[0199] The contamination density feature is obtained by calculating the proportion of pixels in the contaminated area to pixels in the component area. The area proportion of different types of pollution areas is statistically analyzed, and the type with the largest area proportion is taken as the dominant pollution type to obtain pollution type characteristics, such as dust, bird droppings or salt spray crystals. If more refined type classification is needed, texture features (such as gray-level co-occurrence matrix GLCM) and color features can be further extracted, and the existing classification algorithm of support vector machine (SVM) can be used for type discrimination.

[0200] The pollution density characteristics, pollution type characteristics, and pollution distribution information in the component coordinate system are combined to form a pollution assessment feature set.

[0201] The risk linkage unit is used to correlate the pollution assessment feature set with the swing risk index to generate the cleaning intensity requirement index.

[0202] In this embodiment, the risk linkage unit correlates the pollution assessment feature set with the swing risk index to generate a cleaning intensity requirement index. The specific process is as follows:

[0203] Normalize the pollution density characteristics:

[0204] ;

[0205] in, As a reference density for severe pollution, a type factor is set according to the pollution type. For example: dust 0.5, salt spray crystals 0.8, bird droppings stains 1.0.

[0206] Swing risk indicators Normalization (If it is already in [0,1], then use it directly).

[0207] Define cleaning intensity requirements :

[0208] ;

[0209] in, , , The weighting coefficients are non-negative and satisfy the following conditions: ,For example , , ;

[0210] Will Limited to the range [0,1], if the result is less than the preset basic cleaning intensity Then it will be upgraded to This ensures that basic cleaning is still carried out even under low-risk conditions.

[0211] If pollution is concentrated in localized areas, the risk linkage unit will be high. The region is mapped as a localized enhanced cleaning region so that a segmented path mode can be used for multiple iterations in the subsequent parameter generation unit.

[0212] The parameter generation unit is used to output an adaptive cleaning parameter set based on the cleaning intensity requirement index and the attitude safety domain parameter set. The adaptive cleaning parameter set includes brush pressure level, reciprocating frequency level and segmented path mode.

[0213] In this embodiment, the parameter generation unit outputs an adaptive cleaning parameter set based on the cleaning intensity requirement index and the attitude safety domain parameter set, including brush pressure level, reciprocating frequency level, and segmented path mode. The specific process is as follows:

[0214] Will Mapped to several discrete gears:

[0215] when Low strength;

[0216] when Medium intensity;

[0217] when High strength.

[0218] Read the maximum allowable roll, maximum allowable walking speed, and maximum allowable brush pressure of the current node from the attitude safety domain parameter set, and calculate the safety factor. :

[0219] ;

[0220] Will The range is limited to (0,1) to reduce the intensity of cleaning in high-risk areas.

[0221] Set the maximum brush pressure level to The maximum reciprocating frequency level is ;

[0222] Calculate the original target gear:

[0223] , ;

[0224] Then adjust based on the safety factor:

[0225] , ;

[0226] in, This is for floor function.

[0227] If the pollution assessment feature set shows that the pollution is evenly distributed and For low or medium intensity, select the long-stroke reciprocating mode as the segmented path mode;

[0228] If the pollution is locally concentrated and For high intensity, the polluted area is divided into several short segments, and a short-journey, multiple-round-trip mode is selected;

[0229] If the sway risk index is high, the short-stroke mode should be selected in any gear to reduce the cumulative sway during a single pass.

[0230] The final output is an adaptive cleaning parameter set, which includes the brush pressure level corresponding to the gear. Reciprocating frequency level corresponding to gear In the segmented path mode, the adaptive cleaning parameter set is sent to the cleaning actuator and the outer loop path planning unit to generate remote path instructions with enhanced or weakened cleaning actions in the next cycle.

[0231] The evidence chain feedback module includes a log recording unit, an anomaly attribution unit, and a policy update unit.

[0232] The log recording unit is used to associate and store remote path instructions, attitude compensation instructions, adaptive cleaning parameter sets and alignment running status datasets to form running evidence data.

[0233] The anomaly attribution unit is used to identify attitude over-limit events, tension change events, or cleaning interruption events based on operational evidence data, and outputs the anomaly attribution results.

[0234] The strategy update unit is used to feed back the abnormal attribution results to the swing risk prediction module and the sweeping strategy orchestration module, and output adjustment signals.

[0235] The logic for generating the adjustment signal is as follows:

[0236] If the swing risk indicator remains within the preset high-risk threshold, a deceleration signal or a pause signal will be output.

[0237] If the attitude angle data exceeds the upper limit of the allowable attitude corresponding to the attitude safety domain parameter set, a recovery signal will be output.

[0238] If the pollution assessment feature set shows persistent pollution and the swing risk index is within the preset low risk threshold, then an enhanced cleaning signal will be output.

[0239] The adjustment signal will be sent to the notification end.

[0240] This invention collects and synchronizes attitude angle data, angular velocity data, suspension tension data, displacement data, wind speed data, wind direction data, and gust factor data through a state-aware alignment module to form an aligned operating state dataset. The sway risk prediction module utilizes a wind load estimation algorithm based on dynamic pressure formulas and existing time series analysis algorithms based on sliding window statistics and exponentially weighted moving averages to calculate wind load parameters, sway amplitude indicators, tension fluctuation characteristics, and sway trend indicators, and outputs sway risk indicators. Compared to traditional methods that rely solely on a single wind speed or empirical threshold for protection, this system achieves a joint assessment of wind load parameters and sway characteristics. It can identify high-risk operating conditions in advance under strong winds and gusts, providing a reliable basis for speed reduction and attitude recovery, and significantly reducing the risks of excessive sweeper sway, rope impact, and equipment falls.

[0241] The array constraint modeling module, based on component layout and tilt angle data, clusters actual tilt angles into several tilt angle partition identifiers through tilt angle partitioning units. Combining row and column boundaries, end buffer distances, and cross-row connection relationships, it generates an array geometric constraint set. This constraint set is used to restrict the reachability and prohibition regions of nodes in the outer loop path planning unit, enabling structure-aware path planning for mountainous terrain, multi-tilt angle scenarios, and local installation deviations. This effectively avoids the scraping and jamming problems that occur at the ends and cross-rows in traditional control methods based on fixed travel or simple limits.

[0242] The remote decoupled control module generates remote path commands based on the array geometric constraint set and attitude safety domain parameter set through the outer loop path planning unit. Then, the inner loop attitude compensation unit performs anti-sway correction on the remote path commands by combining the aligned running state dataset and outputs attitude compensation commands. The inner loop adopts existing control algorithms such as PD control and speed feedforward, and combines the sway risk index to safely reduce the intermediate target speed to obtain the final safe execution speed. Through the hierarchical decoupling of position control and attitude control, the system can not only cover the entire array according to the planned path, but also adjust the speed and attitude in real time when the local wind load increases or the sway intensifies, suppressing roll and pitch sway, and improving the fitting stability of the hook-mounted sweeper in different tilt angle zones.

[0243] The collaborative scheduling unit divides the remote path commands and attitude compensation commands of multiple photovoltaic wall-mounted sweepers into time slots and calculates predicted positions. A greedy scheduling algorithm based on time slots is used to detect and resolve potential conflicts. For sweepers with lower priority, the target speed is automatically reduced or the translation execution time is shifted within the conflict time slot. Multi-machine collaborative operation is achieved by generating a collaborative control queue. Compared with existing solutions that rely solely on manual staggered start / stop and lack fine-grained scheduling logic, this system can achieve parallel sweeping and dynamic staggered peak control of multiple wall-mounted sweepers while ensuring safe distances and avoiding collisions, thus improving the overall operating efficiency of large-scale arrays.

[0244] The cleaning strategy is adaptively orchestrated based on the pollution assessment feature set and the swing risk index, taking into account both cleaning effect and component life.

[0245] In the cleaning strategy orchestration module, the pollution perception unit uses existing convolutional neural network semantic segmentation and texture analysis algorithms to extract pollution density and pollution type features from component surface image data, forming a pollution assessment feature set. The risk linkage unit correlates the pollution assessment feature set with the sway risk index to obtain the cleaning intensity requirement index. The parameter generation unit then combines the attitude safety domain parameter set to generate an adaptive cleaning parameter set, including brush pressure level, reciprocating frequency level, and segmented path mode, to achieve an intensity risk linkage strategy that strengthens cleaning when pollution is heavy and the risk is low, and weakens cleaning when pollution is light or the risk is high. Compared with traditional systems with fixed cleaning intensity, this invention effectively avoids excessive friction on the component surface while ensuring the removal of stubborn pollution, thus balancing power generation performance and component lifespan.

[0246] The evidence chain feedback module associates and stores remote path commands, attitude compensation commands, adaptive cleaning parameter sets, and aligned operating status datasets to form operational evidence data. It then identifies attitude over-limit events, tension mutation events, and cleaning interruption events through an anomaly attribution unit, feeding the anomaly attribution results back to the sway risk prediction module and the cleaning strategy orchestration module to update threshold and level mapping rules. This design enables the system to continuously learn and optimize strategies, facilitating ongoing adjustments to parameter configurations based on field data and improving long-term reliability and cost-effectiveness.

[0247] In summary, this invention constructs a system framework that integrates remote control, adaptive attitude adjustment, multi-machine collaboration, and cleaning strategy arrangement by collaboratively processing multi-source state perception data, array geometric constraint information, wind load and sway characteristics, and image contamination information. This significantly improves the safety, operating efficiency, and cleaning effect of the undermount sweeper in complex photovoltaic scenarios.

[0248] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0249] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper, characterized in that, It includes a state-aware alignment module, an array constraint modeling module, a swing risk prediction module, a remote decoupling control module, a cleaning strategy orchestration module, and an evidence chain feedback module; The state-aware alignment module collects attitude data, tension data, displacement data, and micro-meteorological data and synchronizes them in time to obtain an alignment operation state dataset; The array constraint modeling module establishes an array geometric constraint set based on the aligned running state dataset and component arrangement data; The swing risk prediction module calculates the swing risk index and generates a set of attitude safety domain parameters by combining the array geometric constraint set. The remote decoupling control module outputs remote path commands and attitude compensation commands based on the attitude safety domain parameter set. The cleaning strategy orchestration module outputs an adaptive cleaning parameter set based on remote path instructions and attitude compensation instructions; The evidence chain feedback module generates analysis results and outputs adjustment signals based on remote path instructions, attitude compensation instructions, adaptive cleaning parameter sets, and alignment operation status datasets.

2. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 1, characterized in that, The state perception alignment module includes an attitude acquisition unit, a suspension tension acquisition unit, a walking displacement acquisition unit, a micro-meteorological acquisition unit, and an alignment processing unit; The attitude acquisition unit is used to acquire attitude angle data and angular velocity data based on the IMU and output them as attitude data; The suspension tension acquisition unit is used to acquire real-time tension data of the hanging traction structure and output it as tension data. The walking displacement acquisition unit is used to acquire displacement data and velocity data along the direction of component travel and output them as displacement data; The micro-meteorological acquisition unit is used to collect wind speed data, wind direction data, and gust factor data and output them as micro-meteorological data. The alignment processing unit is used to clean and synchronize the attitude data, tension data, displacement data and micro-meteorological data in time, form the original equipment sensing data and output the alignment operation status dataset.

3. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 2, characterized in that, The processing logic of the alignment processing unit is as follows: Add a unified timestamp to attitude angle data, real-time tension data, displacement data and wind speed data, remove outliers that exceed the preset physical boundaries, interpolate missing segments to complete them, resample according to the preset sampling period and generate an aligned running status dataset.

4. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 1, characterized in that, The array constraint modeling module includes a component layout acquisition unit, a tilt angle partitioning unit, and a reachability forbidden region generation unit. The component layout acquisition unit is used to acquire component row and column boundary data, end buffer distance data and component tilt angle data, and merge the component row and column boundary data, end buffer distance data and component tilt angle data to output component layout data; The tilt angle partitioning unit is used to generate tilt angle partitioning identifiers based on component tilt angle data; The reachable forbidden region generation unit is used to generate an array geometric constraint set by combining the tilt partition identifier and the alignment running status dataset.

5. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 1, characterized in that, The swing risk prediction module includes a wind load estimation unit, a swing characteristic calculation unit, and a safety domain generation unit. The wind load estimation unit is used to calculate wind load parameters based on wind speed data, wind direction data, and gust factor data. The swing characteristic calculation unit is used to calculate the swing amplitude index and swing trend index based on attitude angle data, angular velocity data and real-time tension data, and output the swing risk index. The safety domain generation unit is used to generate an attitude safety domain parameter set by combining the swing risk index, wind load parameters and array geometric constraint set.

6. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 5, characterized in that, The generation logic of the attitude safety domain parameter set is as follows: When the gust factor data is higher than the preset gust threshold, the allowable roll limit and the allowable walking speed limit are reduced to the preset first and second percentages, respectively. When the fluctuation amplitude of the real-time tension data exceeds the preset tension fluctuation threshold, the allowable pitch adjustment range is reduced to the preset third percentage. When the attitude angle data meets the preset stability conditions, it is restored to the default safe domain parameters of the corresponding tilt angle partition.

7. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 1, characterized in that, The remote decoupling control module includes an outer loop path planning unit, an inner loop attitude compensation unit, and a collaborative scheduling unit. The outer loop path planning unit is used to generate remote path instructions based on the array geometric constraint set and the attitude safety domain parameter set. The inner ring attitude compensation unit is used to perform anti-sway correction on the remote path command based on the aligned running state dataset and output attitude compensation command. The collaborative scheduling unit is used to schedule the remote path instructions and attitude compensation instructions of multiple photovoltaic wall-mounted sweepers in a specific time sequence and output a collaborative control queue.

8. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 1, characterized in that, The cleaning strategy arrangement module includes a pollution sensing unit, a risk linkage unit, and a parameter generation unit; The pollution sensing unit is used to collect image data of the component surface and extract pollution density features and pollution type features to form a pollution assessment feature set; The risk linkage unit is used to correlate and calculate the pollution assessment feature set with the swing risk index to generate the cleaning intensity requirement index. The parameter generation unit is used to output an adaptive cleaning parameter set based on the cleaning intensity requirement index and the attitude safety domain parameter set. The adaptive cleaning parameter set includes brush pressure level, reciprocating frequency level and segmented path mode.

9. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 1, characterized in that, The evidence chain feedback module includes a log recording unit, an anomaly attribution unit, and a policy update unit. The log recording unit is used to associate and store evidence of remote path instructions, attitude compensation instructions, adaptive cleaning parameter sets and alignment running status datasets to form running evidence data. The anomaly attribution unit is used to identify attitude over-limit events, tension change events, or cleaning interruption events based on operational evidence data, and output anomaly attribution results. The strategy update unit is used to feed back the abnormal attribution results to the swing risk prediction module and the cleaning strategy orchestration module, and output adjustment signals.

10. The remote control and attitude adjustment system for a photovoltaic wall-mounted sweeper as described in claim 9, characterized in that, The logic for generating the adjustment signal is as follows: If the swing risk indicator remains within the preset high risk threshold, a deceleration signal or a pause signal will be output. If the attitude angle data exceeds the upper limit of the allowable attitude corresponding to the attitude safety domain parameter set, a recovery signal will be output. If the pollution assessment feature set shows persistent pollution and the swing risk index is within the preset low risk threshold, then an enhanced cleaning signal will be output. The adjustment signal is sent to the notification terminal.

Images