Energy consumption optimization system for remote operation of tower crane

Abstract

The invention provides an energy consumption optimization system for remote operation of a tower crane, and relates to the technical field of energy consumption management of tower cranes. The system collects tower crane motion characteristics, energy tracks and environment interference factors in real time through a high-precision sensor, and unified data characterization and characteristic dimension reduction are achieved through Lie group algebra and Riemannian manifold. Energy sensitive features are extracted through a hierarchical convolution attention mechanism, a dynamic antiantibody agent cluster is constructed, the energy relationship of each factor is analyzed, and an energy consumption optimization management strategy is generated. The system realizes cross-modal conversion and gradient utilization through adaptive energy routing, records energy conversion efficiency by using a distributed account book, and ensures global energy efficiency balance. According to the system, the energy consumption management efficiency of the tower crane is greatly improved, the comprehensive energy consumption is reduced, the energy recovery efficiency is improved, and a new direction is provided for green and intelligent development of the tower crane.

Description

Technical Field

[0001] The present invention relates to the technical field of energy consumption management of tower cranes, and specifically to an energy consumption optimization system for remote operation of tower cranes. Background Art

[0002] In the field of energy consumption management of tower cranes, local energy consumption data is obtained through motor current monitoring or mechanical vibration sensors, and energy consumption evaluation and optimization are carried out based on preset rules or empirical formulas. Some attempts have introduced multi-sensor fusion technology, but most are limited to simple data superposition or weighted processing, and fail to effectively integrate multi-modal data. In addition, existing energy consumption optimization strategies usually adopt fixed thresholds or linear control algorithms, which are difficult to adapt to the complex and changeable working conditions of tower cranes, resulting in low energy recovery efficiency.

[0003] Although the existing technology has improved the energy consumption monitoring accuracy of tower cranes to a certain extent, there are still significant defects. The lack of a unified representation method for the fusion of multi-source heterogeneous data leads to difficulty in synergistically optimizing the energy flow paths of different modalities, resulting in insufficient real-time performance and adaptability of energy consumption management, which restricts the further promotion of green construction and intelligent development. Summary of the Invention

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: An energy consumption optimization system for remote operation of tower cranes, including: A multi-modal data fusion module, which collects the tower crane movement characteristic quantities, the energy trajectories of auxiliary equipment, and environmental interference factors of the tower crane body through multi-source sensors, and constructs the first feature tensor of the tower crane; A feature processing and analysis module, which uses a hierarchical convolutional attention mechanism to perform feature extraction and dimensionality reduction processing on the first feature tensor, and extracts the second feature tensor of energy-sensitive features; A dynamic opponent module, which takes the tower crane movement characteristic quantities, the energy trajectories of auxiliary equipment, and environmental interference factors in the second feature tensor as multiple participants in the confrontation, constructs an intelligent agent cluster, analyzes the energy relationship and interaction between the participants, and generates an energy consumption optimization management strategy; An energy management module, which establishes an adaptive energy routing topology based on the energy consumption optimization management strategy, and through dynamically adjusting the energy flow path, performs cross-modal conversion and gradient utilization of the mechanical energy, electrical energy, and thermal energy generated by the tower crane under different working conditions; A collaborative control module, which constructs a collaborative control mechanism, and realizes the self-execution of energy scheduling between auxiliary equipment through smart contract and distributed ledger technology, and achieves global energy efficiency balance.

[0005] The multi-modal data fusion module deploys a high-precision three-axis MEMS inertial sensor array at the tower crane slewing mechanism, luffing wire rope, and hoisting motor positions, and uses an AD converter with a sampling frequency of 1 kHz to collect the harmonic components of the joint angular acceleration in real time. At the same time, the ripple characteristics of the three-phase current of the motor are extracted. An infrared thermal imager is installed on the surface of the brake friction plate to capture the temperature field gradient distribution at a rate of 30 frames per second. The lidar scans the boom movement trajectory at a frequency of 10 Hz, and the boom trajectory envelope surface is reconstructed through ICP registration of the point cloud data; the mechanical vibration signal is multi-scale decomposed by a differentiable wavelet transform, and the Morlet wavelet basis function is selected to generate a time-frequency matrix, and the time-domain window length is dynamically adjusted to one-quarter of the vibration main frequency period; the thermal imaging map is detected by a sliding window using a 7×7 asymmetric convolution kernel, and the local overheating area characteristics are extracted by calculating the Laplacian operator response of adjacent pixels, and the threshold is set to the ambient temperature + 15°C; a dynamic coupling relationship is established between the boom movement trajectory and the instantaneous wind speed vector measured by the lidar. In the latent space, the mechanical dynamics characteristics, thermodynamic characteristics, and environmental disturbance characteristics are projected onto the manifold through Lie group algebra, and the weights of each feature channel are dynamically adjusted by a gated recurrent unit, and the weight update period is 100 ms. The finally generated first feature tensor has spatio-temporal continuity, the sampling interval in the time dimension is 50 ms, and the spatial dimension contains 32 feature channels.

[0006] The feature processing and analysis module processes the multi-modal data of the tower crane through a hierarchical convolutional attention mechanism. At the bottom layer, the convolutional layer uses a rotation-sensitive hexagonal topology as the convolution kernel to perform convolution operations on the input feature tensor, enabling it to effectively capture the anisotropic vibration modes generated during the swinging of the jib. For example, during tower crane operation, the differences in vibration amplitude and frequency in different directions of the jib are captured, thereby extracting low-level features related to mechanical vibration. Entering the middle layer processing stage, the dynamic sparse convolution strategy adaptively adjusts the azimuth weight distribution of the convolution kernel according to the real-time changes of the environmental wind speed vector in the first feature tensor. This process dynamically monitors the changes in the direction and magnitude of the wind speed vector and updates the weight distribution of the convolution kernel in real time to adapt to the influence of wind speed on the vibration mode of the tower crane. At the same time, the dual-stream gating mechanism processes the time-domain envelope features and frequency-domain harmonic components of the mechanical vibration signal respectively. The time-domain and frequency-domain features are extracted and combined separately to extract vibration features related to energy consumption. The multi-head attention mechanism at the top layer is used to process the thermal imaging temperature gradient field. The thermal imaging temperature gradient field is decomposed into a radial basis function expansion, and then by calculating the non-linear coupling coefficients between the feature channels, an attention probability map oriented to energy sensitivity is generated. This process is similar to finding key feature points related to energy consumption in complex feature data and assigning them higher weights for more efficient energy consumption optimization. In the dimensionality reduction stage, the high-dimensional feature space is compressed into a Riemannian manifold coordinate system derived from the kinematic equation of the jib. By mapping complex multi-modal features into a low-dimensional Riemannian manifold space, the eigenvalue distribution of the Jacobian matrix related to environmental disturbance factors is retained at the same time. In this way, the feature processing and analysis module extracts key features closely related to energy consumption optimization from multi-modal data and forms the second feature tensor.

[0007] The dynamic pair antibody module constructs a collaborative optimization agent cluster by allocating the tower crane motion feature quantity, auxiliary equipment energy trajectory, and environmental interference factor in the second feature tensor to three agents. The agent for the tower crane motion feature quantity extracts modes from the mechanical vibration harmonic components. For example, when the boom swing angle is ±12°, the vibration component with a main frequency of 2.5 Hz is screened out through a hierarchical convolutional attention mechanism, and the corresponding motion constraints are generated. These constraints are directly related to the energy loss rate of the boom trajectory envelope surface. For example, when the vibration amplitude exceeds the threshold, the agent triggers a damping control command to suppress the high-frequency jitter at the end of the swing arm and reduce the mechanical energy loss. The agent for the energy trajectory updates the priority weight of the energy distribution path based on the spatio-temporal distribution of the thermal imaging temperature gradient field at a period of 0.5 seconds. For example, the area where the temperature gradient exceeds 4 °C / m is marked as a high heat consumption area, and the priority weight is increased to more than 0.8, and tensor fusion is performed with the energy-sensitive features after dimensionality reduction. For example, the weight matrix is multiplied element by element with the 8-dimensional feature channels in the Riemannian manifold coordinate system to generate an optimized energy distribution strategy. The agent for the environmental interference factor generates a compensation coefficient matrix based on the wind speed vector field reconstructed by lidar, with the boom kinematic parameters as the benchmark. For example, when the instantaneous wind speed is 7 m / s and the direction forms an angle of 45° with the boom axis, the wind speed disturbance is mapped to the boom kinematic parameter space through Lie group manifold projection, the pose deviation of each node is calculated, and the compensation coefficient increases non-linearly along the boom length direction, and the end compensation value reaches 1.3 times the benchmark value. The communication link between agents is constructed based on the spatio-temporal continuity feature of the first feature tensor, and the interaction weight is dynamically adjusted through the covariance matrix between energy-sensitive features. The feature channels with a covariance value lower than 0.25 are filtered by dynamic sparse convolution. When the mechanical energy fluctuation amplitude exceeds 15% of the nominal value within 10 seconds, collaborative decision-making is triggered, and the output parameters of each agent are orthogonally fused in the tangent space after Lie group manifold projection. For example, the thermodynamic gradient weight and the mechanical vibration phase difference are superimposed in a ratio of 1:2, and finally a global strategy including energy routing optimization, vibration suppression parameters, and environmental compensation is generated to ensure the improvement of the energy consumption efficiency of the tower crane under complex working conditions.

[0008] During the generation process of the energy consumption optimization management strategy, each agent collaborates closely. By analyzing and processing their respective characteristic quantities, an energy consumption optimization management strategy is generated. The agent for the tower crane motion characteristic quantity monitors the change of the boom joint torque in real time and performs covariance analysis with the energy-sensitive characteristics to generate a mechanical energy loss suppression parameter, which reflects the dynamic relationship between torque change and energy loss. The agent for the energy trajectory generates a thermodynamic efficiency optimization coefficient based on the spatio-temporal distribution of local overheating regions in the thermal imaging temperature gradient field, combined with the gated attention mechanism. This coefficient is used to optimize the energy distribution path to ensure the efficient utilization of energy in the system. The agent for the environmental disturbance factor generates an energy manifold compensation vector according to the dynamic coupling relationship between the boom trajectory envelope and the wind speed vector. This vector is used to compensate for the impact of wind speed changes on the boom movement, thereby reducing energy loss. These parameters are projected and fused through the Riemannian manifold coordinate system to form a dynamic routing table for cross-modal energy conversion. In the dynamic routing table, the conversion priority between mechanical energy and thermal energy is determined by the phase difference between the thermodynamic gradient distribution and the mechanical vibration harmonic component. The system calculates the difference between the phase of the mechanical vibration harmonic component and the phase of the thermodynamic gradient distribution. When the phase difference is small, it indicates that the conversion between mechanical energy and thermal energy is more synchronous, and at this time, the priority is high, and energy conversion is carried out preferentially. On the contrary, when the phase difference is large, the priority is low, and the system delays or adjusts the energy conversion path to ensure the efficiency and stability of energy conversion. The energy management module adjusts the energy flow path in real time according to the dynamic routing table, and the distributed ledger of the collaborative control module records the energy conversion efficiency characteristics under various working conditions, providing data support for the continuous optimization of the system, so as to achieve efficient energy consumption management of the tower crane under various working conditions.

[0009] When the energy management module constructs an adaptive energy routing topology, it maps the energy flow characteristics of mechanical energy, electrical energy, and thermal energy to a three-dimensional Riemannian manifold coordinate system. Mechanical energy corresponds to the rotational component, electrical energy corresponds to the translational component, and thermal energy corresponds to the dilation component. By calculating the vector integral of the energy gradient field, a path with the minimum transmission loss is found as the initial routing topology. The transmission loss is determined by measuring the energy attenuation rate on each path, and paths with an attenuation rate exceeding 15% will be excluded. During the cross-modal conversion process, when the infrared thermal imager detects that the temperature gradient in the brake area exceeds 8 °C / m for 5 consecutive seconds and the mechanical vibration phase difference is less than π / 9, the conversion path from mechanical energy to thermal energy is automatically activated, and the conversion efficiency is linearly adjusted according to the real-time temperature gradient value. For every 1 °C / m increase in the gradient, the efficiency increases by 3%. The main harmonic components of the motor phase current ripple characteristics are extracted through spectral analysis. When the ripple amplitude exceeds 12% of the rated value, the electrical energy-mechanical energy conversion efficiency is dynamically adjusted at a rate of 0.8% reduction for every 1% ripple amplitude. The hierarchical energy caching strategy divides the storage according to the projection length of the energy component on the manifold. Energy with a projection length greater than 0.7 directly drives the execution; energy between 0.4 and 0.7 is stored in the cache; energy less than 0.4 is transferred to the backup energy storage. Energy decomposition uses orthogonalization processing on the manifold to decompose the mechanical energy generated by the boom swing into three non-interfering components, and the storage priority of each component is determined by its modulus length in the manifold coordinate system. The dynamic adjustment of the routing topology is achieved by real-time monitoring of the energy manifold distortion index in the second feature tensor. When the environmental wind speed causes the distortion index to exceed the threshold, the system completes the path reconstruction within 100 ms. The new path ensures that the energy transmission efficiency of each node is not low, and at the same time, the path change record is written into the distributed ledger.

[0010] When the collaborative control module constructs a distributed ledger, it stores the conversion records of mechanical energy, electrical energy, and thermal energy in nodes respectively. Each node receives the second feature tensor data at an interval of 0.1 seconds, and establishes an energy conversion priority rule by analyzing the mechanical vibration harmonic components and the thermal imaging temperature gradient field therein. When generating the rule, when it is detected that the correlation coefficient between the 2.5 Hz vibration component and the temperature gradient field exceeds 0.7, this working condition is automatically marked as a priority processing scenario. The underlying verification node continuously compares the energy flow path with the preset topology and calculates the similarity. When the matching degree is lower than 85%, an alarm is triggered. The upper-layer verification node verifies the energy conservation through Lie group manifold projection. When the input-output energy deviation exceeds 5%, it is determined as abnormal. The energy reallocation process selects the optimal solution from the rule table and completes the adjustment within 200 ms after double-layer verification. When updating the ledger, each node verifies the data consistency through feature hashing to ensure that the records are true and reliable.

[0011] The present invention provides an energy consumption optimization system for remote operation of tower cranes, which has the following beneficial effects: 1. Through the multi-modal data fusion and intelligent collaborative optimization mechanism, the present invention significantly improves the energy consumption management efficiency in the remote operation of tower cranes; through the collaborative decision-making of the hierarchical convolutional attention mechanism and the dynamic antibody cluster, it captures the operating state of the tower crane and the characteristics of energy flow in real time, so as to generate optimization strategies suitable for complex working conditions and effectively reduce the comprehensive energy consumption.

[0012] 2. The present invention realizes the efficient conversion and gradient utilization of cross-modal energy through the adaptive energy routing topology, significantly improving the energy recovery efficiency. The system dynamically triggers the optimal energy distribution path according to the energy characteristics, reduces energy waste, and at the same time reduces the wear risk of key components and prolongs the service life of the equipment.

[0013] 3. The present invention realizes the automation and transparency of energy scheduling among auxiliary devices, which not only improves the intelligent level of the remote operation of tower cranes, but also provides industrial application value and social and economic benefits for green construction and sustainable energy management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic flow chart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] Next, the technical solutions in the embodiments of the present invention will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0016] An energy consumption optimization system for the remote operation of tower cranes realizes the refined management of energy consumption of tower cranes under complex working conditions through five core modules: multi-modal data fusion, feature processing and analysis, dynamic antibody optimization, adaptive energy routing, and distributed collaborative control.

[0017] Deploy a high-precision sensor array in the tower crane, including triaxial MEMS inertial sensors at the slewing mechanism, luffing wire rope, and hoisting motor. The sampling frequency is set to 1 kHz for real-time acquisition of the harmonic components of joint angular acceleration and the ripple characteristics of the motor three-phase current. The infrared thermal imager installed on the surface of the brake friction plate captures the temperature field gradient distribution at a rate of 30 frames per second, and the lidar scans the boom trajectory at a frequency of 10 Hz. The boom trajectory envelope surface is reconstructed through ICP registration. The mechanical vibration signal is processed by differentiable wavelet transform, and the Morlet wavelet basis function is selected for multi-scale decomposition. The time domain window is dynamically adjusted to 1 / 4 cycle of the vibration main frequency to generate a matrix containing time-frequency information. The thermal imaging map is detected by sliding a 7×7 asymmetric convolution kernel, and the local overheating area features are extracted. The threshold is set to the ambient temperature + 15°C. A dynamic coupling relationship is established between the boom movement trajectory and the instantaneous wind speed vector measured by the lidar.

[0018] The system projects mechanical dynamics characteristics, thermodynamic characteristics, and environmental disturbance characteristics onto a unified manifold through Lie group algebra. Specifically, the rotational motion of the boom is represented as an element on the SO(3) Lie group, and the translational motion is represented as an element on the SE(3) Lie group. Through the exponential map, the angular velocity and linear velocity of mechanical vibration are converted into corresponding group elements, enabling the fusion of different modal data within a unified framework. The weights of the gated recurrent unit are updated every 100 ms to dynamically adjust the contributions of each feature channel, and finally, a spatio-temporally continuous first feature tensor is generated. The sampling interval in the time dimension is 50 ms, and the spatial dimension contains 32 feature channels.

[0019] The hierarchical convolutional attention mechanism deeply processes the first feature tensor. The underlying convolutional layer uses a rotation-sensitive hexagonal topology as the convolution kernel to effectively capture the anisotropic vibration modes generated by the boom swing. In the middle processing stage, the dynamic sparse convolution strategy adaptively adjusts the azimuth weight distribution of the convolution kernel according to the real-time wind speed vector. When the angle between the wind direction and the boom axis exceeds 30° and the wind speed exceeds 5 m / s, the corresponding azimuth weight is increased by 20%. The dual-stream gated mechanism processes the time domain envelope and frequency domain harmonics of the mechanical vibration signal respectively, and generates comprehensive features through element-level fusion. The top-layer multi-head attention mechanism decomposes the thermal imaging temperature gradient field into a radial basis function expansion, and calculates the non-linear coupling coefficient between each feature channel. For example, when the coupling coefficient between a certain area in the temperature gradient field and its adjacent area exceeds 0.6, this area is marked as an energy-sensitive area. The high-dimensional feature space is compressed to 8 dimensions through the Riemannian manifold coordinate system, while retaining the eigenvalue distribution of the Jacobian matrix. The realization of the Riemannian manifold is based on the boom kinematic equation, which maps the feature vectors of mechanical energy, thermal energy, and electrical energy into a unified manifold space. By calculating the geodesic distance between feature vectors, the similarity between different energy modes is determined, thereby achieving feature dimensionality reduction. Finally, a second feature tensor is formed.

[0020] The tower crane motion feature quantities, the energy trajectory of auxiliary equipment, and the environmental interference factors in the second feature tensor are assigned to three agents to construct a collaborative optimization cluster. The tower crane motion feature quantity agent analyzes the mechanical vibration harmonic components. When the boom swing angle exceeds ±10° and the vibration frequency exceeds 2 Hz, it generates damping control parameters to suppress the end jitter and reduce the mechanical energy loss rate by 18%. The energy trajectory agent is based on the thermal imaging temperature gradient field. When the temperature gradient in the brake area is detected to exceed 6 °C / m, the priority weight is increased to 0.7 and tensor fusion is performed with the energy-sensitive features. The environmental interference agent uses the wind speed vector field reconstructed by lidar. When the instantaneous wind speed exceeds 6 m / s and the angle between the direction and the boom exceeds 30°, it generates a compensation coefficient matrix. The wind speed disturbance is mapped to the boom kinematic parameter space through Lie group manifold projection, and the end compensation value reaches 1.2 times the reference value to ensure the stability of the boom trajectory.

[0021] The communication link between agents is constructed based on the spatio-temporal continuity of the feature tensor, and the interaction weights are dynamically adjusted through the covariance matrix. When the mechanical energy fluctuation amplitude exceeds 15% of the nominal value within 10 seconds, collaborative decision-making is triggered, and the output parameters of each agent are orthogonally fused in the tangent space to generate a global energy consumption optimization strategy. In the fusion process, the interaction between the Lie group manifold and the Riemannian manifold is reflected in that the Lie group manifold provides the transformation basis for kinematic parameters, while the Riemannian manifold determines the priority of energy conversion through geodesic optimization. For example, when the phase difference between the mechanical vibration and the thermodynamic gradient is less than π / 6, the system determines that the energy conversion synchrony is relatively high, and preferentially triggers the energy conversion path from mechanical energy to thermal energy.

[0022] The adaptive energy routing topology maps the energy flow characteristics of mechanical energy, electrical energy, and thermal energy to a three-dimensional Riemannian manifold coordinate system. Mechanical energy corresponds to the rotational component, electrical energy corresponds to the translational component, and thermal energy corresponds to the dilation component. By calculating the vector integral of the energy gradient field, the path with the minimum transmission loss is found as the initial topology. When the path energy decay rate exceeds 15%, it is automatically excluded to ensure the energy transmission efficiency. During the cross-modal conversion process, when the infrared thermal imager detects that the temperature gradient in the brake area continuously exceeds 8 °C / m for 5 seconds and the mechanical vibration phase difference is less than π / 6, the energy conversion path from mechanical energy to thermal energy is activated, and the conversion efficiency is linearly adjusted with the temperature gradient, with an efficiency increase of 2% for each increase of 1 °C / m.

[0023] The hierarchical energy caching strategy divides the storage priorities according to the projection lengths of energy components on the manifold; the energy with a projection length greater than 0.6 directly drives the execution, the energy between 0.3 and 0.6 is stored in the cache, and the energy less than 0.3 is transferred to the standby energy storage; the energy decomposition adopts manifold orthogonalization processing to decompose the mechanical energy of the jib swing into three non-interfering components, and the storage priorities of each component are determined by the modulus length; when the energy manifold distortion index caused by the ambient wind speed exceeds 0.2, the system completes the path reconstruction within 100 ms to ensure that the transmission efficiency of each node is not less than 85%; during the reconstruction process, the Lie group manifold is used to update the compensation coefficient of the jib kinematic parameters, while the Riemannian manifold determines the new energy flow path through geodesic re-planning.

[0024] The distributed ledger takes the energy conversion efficiency as the consensus mechanism, and the conversion records of mechanical energy, electrical energy and thermal energy are stored in different nodes respectively; each node receives the second feature tensor data at an interval of 0.1 second, analyzes the mechanical vibration harmonics and the temperature gradient field, and establishes the energy conversion priority rule; for example, when the correlation coefficient between the 2.5 Hz vibration component and the temperature gradient field exceeds 0.6, this working condition is marked as a priority processing scenario; the underlying verification node compares the energy flow path with the preset topology in real time, and triggers an alarm when the matching degree is lower than 80%; the upper-layer verification node verifies the energy conservation through the Lie group manifold projection, and determines it as abnormal when the deviation between the input and output energies exceeds 3%; the energy reallocation process selects the optimal scheme from the rule table and completes the adjustment within 200 ms after double-layer verification.

[0025] For example, in the tower crane operation scenario at a construction site, when the instantaneous wind speed reaches 7 m / s and the direction forms a 45° angle with the jib, the environmental disturbance intelligent agent generates a compensation coefficient matrix to map the wind speed disturbance to the jib kinematic parameter space; at this time, the tower crane motion characteristic quantity intelligent agent detects that the jib swing angle exceeds ±12° and the vibration frequency reaches 2.8 Hz, and generates damping control parameters to suppress the end jitter; the energy trajectory intelligent agent detects that the temperature gradient of the brake exceeds 8 °C / m, and the priority weight is increased to 0.8, triggering the conversion path from mechanical energy to thermal energy, and the conversion efficiency is increased to 83%; the system fuses the parameters of each intelligent agent through the Riemannian manifold coordinate system to generate a dynamic routing table, and sets the conversion priorities of mechanical energy and thermal energy to the highest; the energy management module adjusts the energy flow path according to the routing table, converts the excess mechanical energy into thermal energy for storage, and records the energy conversion efficiency characteristics through the distributed ledger at the same time; the cooperative control module verifies that the matching degree between the energy distribution path and the preset topology reaches 92%, and the deviation of the energy conservation verification is controlled within 1.5% to ensure the balance of the global energy efficiency.

[0026] In the system of Lie group manifold and Riemannian manifold, the Lie group manifold is responsible for dealing with the continuous symmetries and transformations of dynamic systems, providing a mathematical basis for the kinematic parameters of the jib; the Riemannian manifold realizes the unified representation and efficient conversion of multi-modal energy features through geodesic optimization and feature dimensionality reduction; the two work together to ensure that the system achieves energy consumption optimization and refined control of energy management under complex working conditions; for example, under complex wind conditions, the system reduces the comprehensive energy consumption of the tower crane, improves the energy recovery efficiency, and reduces the wear of key components; the distributed ledger record shows that the standard deviation of the energy conversion efficiency feature is reduced and the system stability is improved; through the mathematical mapping of the Lie group manifold and the Riemannian manifold, the unified representation and efficient conversion of multi-modal energy features are realized, providing reliable technical support for the green and intelligent development of tower crane remote operation.

[0027] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An energy consumption optimization system for remote operation of a tower crane, characterized in that: include: The multimodal data fusion module collects the motion characteristics of the tower crane, the energy trajectory of the auxiliary equipment and the environmental interference factors through multi-source sensors, and constructs the first characteristic tensor of the tower crane; The feature processing and analysis module uses a layered convolutional attention mechanism to perform feature extraction and dimensionality reduction on the first feature tensor to extract the second feature tensor of energy-sensitive features; The dynamic confrontation body module takes the crane motion feature quantity, auxiliary equipment energy trajectory and environmental interference factors in the second feature tensor as multiple participants in the confrontation, builds an intelligent agent cluster, analyzes the energy relationship and interaction between the participants, and generates energy consumption optimization management strategies; The energy management module establishes an adaptive energy routing topology based on the energy consumption optimization management strategy. By dynamically adjusting the energy flow path, the mechanical energy, electrical energy and thermal energy generated by the tower crane under different working conditions are converted and utilized in a gradient manner. The collaborative control module builds a collaborative control mechanism to realize self-execution of energy scheduling among attached devices through smart contracts and distributed ledger technology, thus achieving global energy efficiency balance.

2. The energy consumption optimization system for remote operation of a tower crane according to claim 1 is characterized in that: The multimodal data fusion module deploys a three-axis MEMS inertial sensor array on the tower crane's slewing mechanism, luffing wire rope and hoisting motor to collect the joint angular acceleration harmonic components and motor phase current ripple characteristics in real time, captures the brake friction plate temperature field gradient distribution through an infrared thermal imager, and reconstructs the boom trajectory envelope surface in combination with the lidar point cloud data. The above data stream is decomposed into a time-frequency matrix by multi-scale decomposition of the mechanical vibration signal using a differentiable wavelet transform, and an asymmetric convolution kernel is used to extract the local overheating area features in the thermal imaging image, and a dynamic coupling relationship between the boom motion and the wind speed vector is established. Finally, in the latent space, the mechanical dynamics characteristics, thermodynamic characteristics and environmental disturbance characteristics are projected onto a unified Lie group manifold to form a first feature tensor with spatiotemporal continuity, in which each feature channel dynamically adjusts its contribution weight through a gated recurrent unit to achieve noise suppression and feature complementarity.

3. The energy consumption optimization system for remote operation of a tower crane according to claim 2 is characterized in that: The feature processing and analysis module is processed through a hierarchical convolution attention mechanism to construct a cascade processing structure consisting of a spatiotemporal interleaved convolution layer and a gated attention unit, wherein the bottom convolution kernel uses a rotation-sensitive hexagonal topology structure to capture the anisotropic vibration mode of the boom swing; the middle layer uses a dynamic sparse convolution strategy to adaptively adjust the azimuth weight distribution of the convolution kernel according to the real-time change of the ambient wind speed vector in the first feature tensor, and uses a dual-stream gating mechanism to process the time domain envelope characteristics and frequency domain harmonic components of the mechanical vibration signal respectively; The top-level design uses a multi-head attention mechanism with energy perception capability to decompose the thermal imaging temperature gradient field into a radial basis function expansion, and generates an attention probability map for energy sensitivity by calculating the nonlinear coupling coefficients between each feature channel. In the dimensionality reduction stage, the high-dimensional feature space is compressed to the Riemannian manifold coordinate system derived from the kinematic equations of the boom, while retaining the eigenvalue distribution of the Jacobian matrix related to the environmental interference factors, ultimately forming the second eigentensor.

4. The energy consumption optimization system for remote operation of a tower crane according to claim 3 is characterized in that: The dynamic pair antibody module constructs an agent cluster, wherein the crane motion feature quantity, the auxiliary equipment energy trajectory and the environmental interference factor in the second feature tensor are assigned to three agents, wherein the crane motion feature quantity agent generates action constraints by extracting the mechanical vibration harmonic component through the hierarchical convolution attention mechanism, and its output is associated with the energy loss rate of the boom swing trajectory envelope surface; the energy trajectory agent predicts the priority weight of the energy distribution path based on the thermal imaging temperature gradient field constructed by the multimodal data fusion module, and fuses the weight with the second feature tensor; The agent of environmental disturbance factors uses the wind speed vector field reconstructed by LiDAR to generate a compensation coefficient matrix linked to the kinematic parameters of the boom, which is aligned with the mechanical vibration characteristics through Lie group manifold projection; The communication links between the agents are constructed based on the spatiotemporal continuity characteristics of the first feature tensor. The interaction weights are adjusted by the correlation between the energy-sensitive features extracted from the hierarchical convolutional attention mechanism, and the noise signals irrelevant to the current working conditions are filtered out through dynamic sparse convolution, so that the agents trigger collaborative decision-making only when the mechanical energy fluctuation threshold is exceeded.

5. The energy consumption optimization system for remote operation of a tower crane according to claim 4, characterized in that: The method for generating the energy consumption optimization management strategy is to calculate the covariance matrix of the boom joint torque and energy-sensitive characteristics in real time through the intelligent agent of the tower crane motion characteristic quantity, and generate the mechanical energy loss suppression parameter; the intelligent agent of the energy trajectory generates the thermodynamic efficiency optimization coefficient based on the spatiotemporal distribution of the local overheating area in the thermal imaging temperature gradient field and the gated attention mechanism; The environmental interference agent generates an energy manifold compensation vector according to the dynamic coupling relationship between the boom trajectory envelope and the wind speed vector; the above parameters are projected and fused through the Riemann manifold coordinate system to form a dynamic routing table for cross-modal energy conversion, in which the conversion priority of mechanical energy and thermal energy is determined by the phase difference between the thermodynamic gradient distribution and the harmonic component of the mechanical vibration; the final energy consumption optimization management strategy adjusts the energy flow path in real time through the adaptive routing topology of the energy management module, and uses the distributed ledger of the collaborative control module to record the energy conversion efficiency characteristics under various working conditions.

6. The energy consumption optimization system for remote operation of a tower crane according to claim 1, characterized in that: The energy management module implements adaptive energy routing topology to map the energy flow characteristics of mechanical energy, electrical energy and thermal energy to the Riemannian manifold coordinate system, and constructs an initial routing topology by solving the optimal transmission path of the energy gradient field; For the cross-modal conversion process, a bidirectional gating mechanism based on thermodynamic gradient distribution and mechanical vibration phase difference is designed. When the thermal imaging temperature gradient exceeds the threshold, the conversion path of mechanical energy to thermal energy is triggered, and the conversion efficiency of electrical energy and mechanical energy is dynamically adjusted according to the motor phase current ripple characteristics; in the gradient utilization stage, the excess mechanical energy generated by the swing of the boom is decomposed into energy components of different priorities through Lie group manifold projection through a hierarchical energy caching strategy. The high-priority components are directly supplied to the current working conditions, and the low-priority components are stored in the energy pool; the dynamic adjustment of the routing topology is achieved through the real-time monitoring of the second feature tensor output by the feature processing and analysis module. When the environmental interference factor causes the energy manifold to be distorted, the routing path is immediately reconstructed and the energy allocation record in the distributed ledger is updated.

7. The energy consumption optimization system for remote operation of a tower crane according to claim 6, characterized in that: The collaborative control module constructs a distributed ledger based on the adaptive routing topology with energy conversion efficiency as the consensus mechanism. Each subchain in the distributed ledger corresponds to a conversion record of an energy mode; the conversion records of mechanical energy, electrical energy and thermal energy are stored in different nodes respectively; by analyzing the mechanical vibration harmonic components and the thermal imaging temperature gradient field in the second characteristic tensor, an energy conversion priority rule table is established, and when the boom swing trajectory envelope does not match the energy demand of the current working condition, the energy redistribution process is automatically triggered; a double-layer verification mechanism is adopted, in which the bottom-level verification node compares the matching degree of the actual energy flow path with the routing topology in real time, and the upper-level verification node evaluates the global efficiency through the energy conservation constraint on the Lie group manifold.

Images