Energy consumption monitoring optimization system based on digital twinning in intelligent warehousing

Abstract

The invention provides an energy consumption monitoring optimization system based on digital twinning in intelligent warehousing, and relates to the technical field of intelligent management and control, and the system comprises an obtaining module which is used for obtaining multi-source high-frequency dynamic data in real time, carrying out the real-time mapping and dynamic synchronization of a static three-dimensional visual model, and obtaining a dynamic digital twinning body; the construction module is used for constructing a dynamic parameterized spatial manifold based on the dynamic digital twin according to the distribution positions and respective data updating frequencies of the transient current harmonic acquisition points, the real-time path tracking and energy consumption mapping acquisition points and the temperature and humidity field dynamic response sensing acquisition points in the space; and mapping the multi-source high-frequency dynamic data to discrete nodes of a spatial manifold according to a timestamp and a spatial coordinate to obtain a synchronous operation state flow. The comprehensive energy consumption is reduced, and the energy efficiency management and control refinement level is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent management and control technology, and in particular to an energy consumption monitoring and optimization system based on digital twins for intelligent warehousing. Background Technology

[0002] As smart warehousing upgrades towards automation, intelligence, and large-scale operations, the energy consumption of core equipment such as stacker cranes, unmanned handling equipment, and cold chain refrigeration systems is increasing year by year. Energy consumption management has become the key to reducing warehousing operating costs and achieving green and low-carbon operations. Currently, most smart warehouses use traditional energy consumption monitoring methods.

[0003] For example, a large-scale intelligent cold chain warehouse only deploys a single type of energy consumption monitoring point at the stacker crane drive motor and the cold chain evaporator fan. It monitors energy consumption by collecting current, temperature and humidity data at regular intervals. However, it has technical defects: in actual operation, the warehouse often experiences transient current harmonic impacts due to frequent start-stop of the stacker crane and congestion on the path of unmanned handling equipment, which leads to a surge in energy consumption per unit mileage. Traditional monitoring systems collect discrete single-point data, making it difficult to build a spatiotemporal correlation model of energy consumption across the entire warehouse. This results in serious energy waste and low control accuracy, making it difficult to meet the refined and energy-saving operation requirements of intelligent warehousing. Summary of the Invention

[0004] This invention provides an energy consumption monitoring and optimization system based on digital twins for intelligent warehousing, which reduces overall energy consumption and improves the level of precision in energy efficiency management.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] The first aspect is the energy consumption monitoring and optimization system based on digital twins in smart warehousing, including:

[0007] The acquisition module is used to acquire multi-source high-frequency dynamic data in real time, and to perform real-time mapping and dynamic synchronization of static 3D visualization models to obtain dynamic digital twins;

[0008] The module is used to construct a dynamic parameterized spatial manifold based on a dynamic digital twin, according to the spatial distribution of transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensor acquisition points and their respective data update frequencies. The module then maps multi-source high-frequency dynamic data to discrete nodes of the spatial manifold according to timestamps and spatial coordinates to obtain a synchronous running state flow.

[0009] The computation module is used to perform isoparametric transformation on the dynamically parameterized spatial representation based on the synchronous running state flow to obtain the transformed projection domain; extract the boundary contours corresponding to each acquisition point on the transformed projection domain, and obtain multi-layer nested buffer boundaries by successively offsetting along the normal direction; divide the spatial representation into a core disturbance zone, a transition influence zone, and a peripheral stable zone according to each layer of buffer boundaries, and perform numerical simulation of the transient energy consumption surge trend in each zone to obtain the pre-simulation results of the abnormal energy consumption surge condition and the corresponding pre-control parameter set;

[0010] The processing module is used to perform dynamic adaptive control processing based on the pre-simulation results and the pre-control parameter set, and obtain the equipment operation feedback data after dynamic adaptive control processing.

[0011] The iterative module is used to perform closed-loop optimization iterative processing based on the equipment operation feedback data after dynamic adaptive control, so as to continuously optimize the operating condition prediction and pre-control parameters.

[0012] Furthermore, real-time acquisition of multi-source high-frequency dynamic data allows for real-time mapping and dynamic synchronization of the static 3D visualization model, resulting in a dynamic digital twin. This also enables real-time capture of the original transaction data stream, including:

[0013] Transient current harmonic acquisition points are deployed at the winding end of the drive motor of the high-speed stacker crane, real-time path tracking and energy consumption mapping acquisition points are deployed at the navigation and positioning point of the unmanned handling equipment, and temperature and humidity field dynamic response sensing acquisition points are deployed at the return air vent of the evaporator in the cold chain warehouse. The sampling frequency and data reporting protocol of each acquisition point are configured, and a real-time data transmission link with the digital twin platform is established.

[0014] Based on the real-time data transmission link, each acquisition point captures the three-phase current time-domain waveform of the stacker crane drive motor and extracts the harmonic distortion rate characteristic value, the real-time position coordinates and energy consumption per unit mileage of the unmanned handling equipment, and the temperature and humidity change rate of the return air vent, forming a multi-source high-frequency dynamic data stream.

[0015] Add timestamps and spatial coordinate labels to the multi-source high-frequency dynamic data streams. Based on the preset anchor point positions of each collection point in the static 3D visualization model, map the timestamp-aligned data streams to the corresponding component nodes of the static 3D visualization model according to the spatial coordinates to obtain the mapping results.

[0016] Based on the mapping results, the geometric structure, spatial pose, and operational attribute parameters of the stacker crane, unmanned handling equipment, and cold chain fan in the static 3D visualization model are dynamically refreshed, so that the representation content of the static 3D visualization model is synchronized with the physical entity, forming a dynamic digital twin.

[0017] Furthermore, based on the dynamic digital twin, a dynamic parameterized spatial manifold is constructed according to the spatial distribution and data update frequency of transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points, including:

[0018] Spatial coordinate values ​​of each acquisition point are extracted from the dynamic digital twin, and harmonic distortion rate characteristic value sequence, location coordinate and unit mileage energy consumption sequence, and temperature and humidity change rate sequence are associated with the data stream type corresponding to each acquisition point to form the spatiotemporal attribute vector of each acquisition point.

[0019] Using the spatial coordinates of each collection point as the node position and the data update frequency as the time axis sampling interval, a discrete node set is established. Based on the spatiotemporal attribute vectors associated with each node, a continuous attribute field covering the storage space is constructed. The isoparametric surface of the continuous attribute field is defined as the base surface of the dynamically parameterized spatial manifold.

[0020] On the base surface, based on the direction of harmonic distortion rate gradient change of transient current harmonic acquisition points, the density of motion trajectories of real-time path tracking and energy consumption mapping acquisition points, and the distribution of temperature and humidity field equipotential lines of temperature and humidity field dynamic response sensor acquisition points, a non-uniform scaling factor is applied along the three principal directions of the base surface to obtain a dynamically parameterized spatial manifold.

[0021] The dynamically parameterized spatial manifold is discretized into a mesh topology, and the corresponding spatiotemporal attribute vectors are retained for each vertex, thus completing the construction of the dynamically parameterized spatial manifold.

[0022] Furthermore, the multi-source high-frequency dynamic data is mapped to discrete nodes of the spatial manifold according to timestamps and spatial coordinates to obtain a synchronized running state stream, including:

[0023] The three-dimensional spatial coordinates of each discrete node are extracted from the grid topology. Based on the spatiotemporal attribute vector type of each node, the spatial affiliation between the node and the transient current harmonic acquisition point, the real-time path tracing and energy consumption mapping acquisition point, and the temperature and humidity field dynamic response sensing acquisition point are established to determine the dominant acquisition point affiliation of each discrete node in space.

[0024] Based on three parallel multi-source high-frequency dynamic data streams, and using the set data update frequency of each acquisition point as a benchmark, the data streams are aligned by a sliding window on the time axis to obtain a synchronized data frame sequence with a unified time benchmark.

[0025] For each synchronized data frame, based on the determined spatial affiliation between nodes and acquisition points, the harmonic distortion rate characteristic value, location coordinates, energy consumption per unit mileage, and temperature and humidity change rate corresponding to each acquisition point are assigned to each discrete node in the spatial region, so that each discrete node obtains a fusion attribute value that matches its spatial location.

[0026] The assigned discrete node fusion attribute values ​​are then fused with the retained spatiotemporal attribute vector using Kalman filtering to obtain the real-time state vector of all discrete nodes.

[0027] The real-time state vectors of all discrete nodes are arranged in time sequence according to the obtained synchronous data frame sequence and encapsulated into an iteratively updatable synchronous running state stream.

[0028] Furthermore, based on the synchronous running state flow, an isoparametric transformation is performed on the dynamically parameterized spatial representation to obtain the transformed projection domain; the boundary contours corresponding to each acquisition point are extracted on the transformed projection domain, and a multi-layered nested buffer boundary is obtained by successively offsetting along the normal direction, including:

[0029] Extract the mesh topology of the dynamic parameterized spatial representation and the real-time state vectors on each discrete node from the synchronous running state flow, and construct the mapping relationship from the three-dimensional physical space to the three-dimensional attribute space using the coordinates of each node and the real-time state vectors.

[0030] By applying an isoparametric transformation to the mapping relationship, the dynamic parameterized spatial representation is mapped to the parameterized projection domain in the three-dimensional attribute space, so that the spatial position and the attribute field quantity form a one-to-one correspondence of parameter coordinates, thus completing the parameterized dimensionality reduction expression.

[0031] In the parametric projection domain, with the parameter coordinates corresponding to each acquisition point as the center, the threshold is set according to the characteristic value of harmonic distortion rate, the peak energy consumption per unit mileage and the extreme value of temperature and humidity change rate captured by each acquisition point, and the contour lines are extracted as the initial boundary profile.

[0032] The initial boundary contour is inversely mapped back to the three-dimensional physical space to obtain the corresponding three-dimensional boundary contour lines. Each contour line is then discretized to obtain a polygonal boundary composed of an ordered sequence of points.

[0033] For each polygon boundary, the normal direction of each side segment is calculated, and the extended boundary is obtained successively outward along the normal direction with a preset offset step size. The original boundary and the extended boundary are numbered hierarchically according to the offset distance to obtain a multi-layer nested buffer boundary sequence.

[0034] Furthermore, based on the buffer boundaries of each layer, the spatial representation is divided into a core disturbance region, a transitional influence region, and a peripheral stable region. Numerical simulations of transient energy consumption surge trends are then performed in each region to obtain the pre-simulation results of abnormal energy consumption surge conditions and the corresponding pre-regulation parameter set, including:

[0035] The innermost boundary is extracted from the multi-layer buffer boundary sequence as the core perturbation boundary, and the region inside this boundary is marked as the core perturbation region; the middle boundary is extracted as the transition influence boundary, and the region between the core perturbation boundary and the transition influence boundary is marked as the transition influence region; the region outside the outermost boundary is marked as the peripheral stable region.

[0036] For the core disturbance area, the real-time state vector of discrete nodes in the area is extracted to simulate the instantaneous impact effect of stacker crane start-up and shutdown, equipment path intersection and fan load fluctuation on local energy consumption, and obtain the peak value and duration of energy consumption surge.

[0037] For the transitional impact zone, the state vectors of discrete nodes within the zone are extracted, and the attenuation law and delayed response characteristics of energy consumption impact propagating outward along the grid topology are simulated to obtain the predicted results of energy consumption impact range and lag time.

[0038] For the outer stable region, the steady-state operating parameters of discrete nodes within the region are extracted. The simulation results of the core disturbance region and the transition influence region are compared with the baseline to identify abnormal points exceeding the threshold and obtain the abnormal overflow pre-simulation results.

[0039] Based on the results of the pre-simulation in various regions, the optimized values ​​of stacker crane start-stop frequency, the adjustment values ​​of unmanned handling equipment path energy consumption weight, and the pre-control values ​​of cold chain fan speed are solved to obtain the pre-control parameter set.

[0040] Furthermore, based on the pre-simulation results and the pre-control parameter set, dynamic adaptive control processing is performed to obtain the equipment operation feedback data after dynamic adaptive control processing, including:

[0041] The optimized values ​​of stacker crane start-stop frequency, the adjustment values ​​of unmanned transport equipment path energy consumption weight, and the pre-control values ​​of cold chain fan speed are extracted from the pre-control parameters and encapsulated into control instruction frames for stacker crane drive controller, unmanned transport equipment navigation scheduler, and cold chain fan frequency converter driver, respectively.

[0042] The optimized values ​​of the stacker crane start-stop frequency are sent to the high-speed stacker crane drive controller to dynamically correct the stacker crane start-stop timing and acceleration / deceleration curves, thereby reducing harmonic current impact and mechanical energy consumption.

[0043] The path energy consumption weight adjustment value of the unmanned transport equipment is sent to the navigation scheduler of the unmanned transport equipment to re-plan the travel path and intersection traffic priority, avoid congestion and empty detours, and reduce the path energy consumption density.

[0044] The pre-regulation value of the cold chain fan speed is sent to the cold chain fan frequency converter, and the feedforward adjustment of the evaporator fan speed and start-stop cycle prevents the refrigeration unit from triggering full-load redundant operation due to slight temperature and humidity fluctuations.

[0045] The actual current harmonic sequence after the stacker crane drive controller is executed, the actual unit mileage energy consumption sequence after the unmanned handling equipment navigation scheduler is executed, and the actual temperature and humidity fluctuation sequence after the cold chain fan frequency converter is executed are collected. The collected data is compared with the corresponding raw data stream to calculate the deviation and obtain the equipment operation feedback data.

[0046] Furthermore, based on the equipment operation feedback data after dynamic adaptive control processing, closed-loop optimization iterative processing is performed to continuously optimize the operating condition prediction and pre-control parameters, including:

[0047] The harmonic suppression deviation sequence between the actual current harmonic sequence of the stacker crane and the original harmonic distortion rate, the path energy consumption deviation sequence between the actual unit mileage energy consumption of the unmanned handling equipment and the original unit mileage energy consumption, and the temperature control response deviation sequence between the actual temperature and humidity fluctuation of the cold chain fan and the original temperature and humidity change rate are extracted from the equipment operation feedback data. The three types of deviation sequences are aligned by timestamp to construct a closed-loop deviation vector set.

[0048] The closed-loop deviation vector set is reverse-mapped into the mesh topology of the dynamically parameterized spatial manifold. The deviation vector magnitude is used as a weighting coefficient to correct the non-uniform scaling factor applied along the three principal directions of the base surface, resulting in the updated dynamically parameterized spatial manifold.

[0049] Based on the updated dynamic parameterized spatial manifold, the threshold setting of the initial boundary profile is recalculated, and the offset step size of the successive offsets along the normal direction is adjusted using the statistical characteristics of the closed-loop deviation vector set as the offset.

[0050] The corrected non-uniform scaling factor, the adjusted threshold setting, and the updated offset step size are substituted into the working condition prediction and anomaly simulation processing, so that the simulation results and pre-control parameter set of the next iteration are adaptively calibrated based on the actual control feedback.

[0051] In a second aspect, a computing device includes:

[0052] One or more processors;

[0053] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to execute the system.

[0054] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, performs the system.

[0055] The above-described solution of the present invention has at least the following beneficial effects:

[0056] By constructing a dynamic digital twin that is synchronized in real time with the physical entity of the smart warehouse, and building a dynamic parameterized spatial manifold based on multi-source high-frequency dynamic data to generate a synchronized operating state flow, the spatiotemporal correlation representation of energy consumption across the entire warehouse domain and the precise alignment and fusion of multi-source heterogeneous data were achieved. By performing isoparametric transformation on the dynamic parameterized spatial representation based on the synchronized operating state flow and generating multi-layer nested buffer boundaries, and using the boundaries to divide the core disturbance zone, transition influence zone and peripheral stable zone, the transient energy consumption numerical simulation of each zone was carried out. This enabled the pre-simulation and accurate prediction of transient energy consumption surges such as harmonic impacts during stacker crane start-up and shutdown and path congestion of unmanned handling equipment. Based on the simulation results, a pre-control parameter set was generated and dynamic adaptive control was executed. At the same time, closed-loop optimization iteration was carried out based on equipment operation feedback data, which reduced the overall energy consumption of the smart warehouse and improved the refinement, intelligence and long-term self-optimization level of energy efficiency management. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of an energy consumption monitoring and optimization system based on digital twins in an intelligent warehouse, provided by an embodiment of the present invention.

[0058] Figure 2 This is a schematic diagram illustrating the process of an energy consumption monitoring and optimization system based on digital twins in an intelligent warehouse, provided by an embodiment of the present invention. The system performs dynamic adaptive control processing based on the pre-simulation results and the pre-control parameter set to obtain the equipment operation feedback data after dynamic adaptive control processing. Detailed Implementation

[0059] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0060] like Figure 1 As shown, an embodiment of the present invention proposes an intelligent diagnostic and processing system for abnormal states of urban electronic police monitoring support poles, comprising:

[0061] The acquisition module is used to acquire multi-source high-frequency dynamic data in real time, and to perform real-time mapping and dynamic synchronization of static 3D visualization models to obtain dynamic digital twins;

[0062] The module is used to construct a dynamic parameterized spatial manifold based on a dynamic digital twin, according to the spatial distribution of transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensor acquisition points and their respective data update frequencies. The module then maps multi-source high-frequency dynamic data to discrete nodes of the spatial manifold according to timestamps and spatial coordinates to obtain a synchronous running state flow.

[0063] The computation module is used to perform isoparametric transformation on the dynamically parameterized spatial representation based on the synchronous running state flow to obtain the transformed projection domain; extract the boundary contours corresponding to each acquisition point on the transformed projection domain, and obtain multi-layer nested buffer boundaries by successively offsetting along the normal direction; divide the spatial representation into a core disturbance zone, a transition influence zone, and a peripheral stable zone according to each layer of buffer boundaries, and perform numerical simulation of the transient energy consumption surge trend in each zone to obtain the pre-simulation results of the abnormal energy consumption surge condition and the corresponding pre-control parameter set;

[0064] The processing module is used to perform dynamic adaptive control processing based on the pre-simulation results and the pre-control parameter set, and obtain the equipment operation feedback data after dynamic adaptive control processing.

[0065] The iterative module is used to perform closed-loop optimization iterative processing based on the equipment operation feedback data after dynamic adaptive control, so as to continuously optimize the operating condition prediction and pre-control parameters.

[0066] In this embodiment of the invention, multi-source high-frequency dynamic data is acquired in real time to map and dynamically synchronize a static three-dimensional visualization model, constructing a dynamic digital twin that is synchronized in real time with the physical entity of the intelligent warehouse. Based on the dynamic digital twin, a dynamic parameterized spatial manifold is constructed, and multi-source high-frequency dynamic data is mapped to discrete nodes according to timestamps and spatial coordinates to generate a synchronized running state flow. This achieves the spatiotemporal correlation representation of energy consumption across the entire warehouse and the precise alignment and fusion of multi-source heterogeneous high-frequency data. Furthermore, isoparametric transformations are performed on the dynamic parameterized spatial representation based on the synchronized running state flow, generating multi-layered nested buffer boundaries and dividing the kernel. Transient energy consumption numerical simulations were conducted in the central disturbance zone, transitional influence zone, and peripheral stable zone to obtain pre-simulation results and pre-control parameter sets for abnormal operating conditions with energy consumption surges. Based on the simulation results, dynamic adaptive control was performed, and closed-loop optimization iterations were carried out based on equipment operation feedback data. This enabled the pre-judgment and precise control of transient energy consumption anomalies such as harmonic impacts during stacker crane start-up and shutdown and path congestion of unmanned handling equipment. This avoided energy waste and control accuracy decay caused by the poor adaptability and lack of long-term self-optimization capability of static digital twin models, reduced the overall energy consumption of intelligent warehousing, and improved the refinement, intelligence, and long-term self-optimization level of energy efficiency management.

[0067] In a preferred embodiment of the present invention, multi-source high-frequency dynamic data is acquired in real time, and a static three-dimensional visualization model is mapped and dynamically synchronized in real time to obtain a dynamic digital twin. The original transaction data stream is captured in real time, including:

[0068] Transient current harmonic acquisition points are deployed at the winding end of the drive motor of the high-speed stacker crane; real-time path tracking and energy consumption mapping acquisition points are deployed at the navigation and positioning points of unmanned handling equipment; and temperature and humidity dynamic response sensing acquisition points are deployed at the return air vent of the evaporator in the cold chain storage area. The sampling frequency and data reporting protocol of each acquisition point are configured, and a real-time data transmission link with the digital twin platform is established. Specifically, in view of the shortcomings of existing traditional warehouses that only deploy a small number of single-type acquisition points and cannot cover multi-dimensional energy consumption sources, the first step is to carry out the deployment of full-dimensional accurate acquisition points and the construction of transmission links. First, determine the physical installation locations of the three types of dedicated function data collection points. A transient current harmonic acquisition point is fixedly installed at the end of the winding inside the drive motor of the high-speed stacker crane. This location can directly capture the fluctuations of the motor's native power supply current, avoiding interference from external lines. A real-time path tracking and energy consumption mapping acquisition point is fixedly installed at the center of the navigation and positioning module on the top of the unmanned handling equipment to accurately match the equipment's movement and positioning benchmark. A temperature and humidity field dynamic response sensor acquisition point is fixedly installed on the side shell of the return air inlet behind the evaporator outlet in the cold chain storage area to directly sense the real environmental parameters of the refrigeration return air.

[0069] Different sampling frequencies are configured for the three types of data collection points, with values ​​set based on energy consumption fluctuation characteristics. For extremely rapid changes in transient current harmonics, a high-frequency sampling frequency of 1,000 times per second is configured. For moderate real-time path and energy consumption fluctuations, a conventional high-frequency sampling frequency of 100 times per second is configured. For slow changes in temperature and humidity, a low-frequency stable sampling frequency of 10 times per second is configured. At the same time, a unified industrial real-time data reporting protocol is configured, with fixed data packet length, cyclic verification rules, and automatic retransmission mechanism for packet loss set to ensure error-free data transmission. A dedicated real-time data transmission link is established between all data collection points and the digital twin platform, relying on industrial Ethernet hardware lines.

[0070] Based on the real-time data transmission link, each acquisition point captures the three-phase current time-domain waveform of the stacker crane drive motor and extracts harmonic distortion rate characteristic values, real-time position coordinates and energy consumption per unit mileage of the unmanned transport equipment, and temperature and humidity change rate of the return air vent, forming a multi-source high-frequency dynamic data stream. Specifically, relying on the established real-time data transmission link, three types of acquisition points are driven to capture raw operating data in parallel and calculate core characteristic parameters, integrating them into a unified data stream. The transient current harmonic acquisition point continuously captures the current time-domain waveform of the three-phase power supply line of the stacker crane drive motor, records the real-time current magnitude corresponding to each sampling moment, and performs harmonic distortion rate characteristic value calculation. First, the effective value of the fundamental frequency current in the current waveform is extracted, and then the effective values ​​of all harmonic currents from the 2nd to the 50th are counted sequentially. The effective values ​​of all harmonic currents are squared and summed. The square root of the sum is taken to obtain the total effective value of the harmonic current. The total effective value of the harmonic current is divided by the effective value of the fundamental current and then multiplied by 100%. Finally, accurate harmonic distortion rate characteristic values ​​are obtained. Secondly, the real-time spatial coordinates and real-time power consumption values ​​of the unmanned transport equipment are captured by the real-time path tracking and energy consumption mapping acquisition points at each sampling moment. The travel distance is calculated by the difference between the position coordinates of two adjacent sampling moments. The total travel distance and total power consumption of the equipment within a fixed period are accumulated and statistically analyzed. The total power consumption is divided by the total travel distance to obtain the real-time energy consumption value per unit mile of the unmanned transport equipment. The real-time temperature and humidity values ​​of the cold chain return air vent are continuously captured by the temperature and humidity field dynamic response sensor acquisition points at each sampling moment. The temperature change rate is obtained by dividing the temperature difference between two adjacent sampling moments by the sampling time interval, and the humidity change rate is obtained by dividing the humidity difference between two adjacent sampling moments by the sampling time interval. The two types of values ​​are integrated to form temperature and humidity change rate parameters. The original waveforms, position coordinates and all calculated characteristic parameters captured by the above three types of points are sequentially spliced ​​according to the acquisition time sequence to form a multi-source high-frequency dynamic data stream with continuous time sequence and complete parameters.

[0071] This process involves attaching timestamps and spatial coordinate labels to multi-source high-frequency dynamic data streams. Based on the preset anchor point positions of each collection point in the static 3D visualization model, the timestamp-aligned data streams are mapped to the corresponding component nodes of the static 3D visualization model according to their spatial coordinates to obtain the mapping results. Specifically, this includes addressing the shortcomings of existing traditional data lacking spatiotemporal markers and unable to match 3D model positions, performing spatiotemporal label binding and precise mapping matching processing on the generated multi-source high-frequency dynamic data streams. For each independent data point in the multi-source high-frequency dynamic data stream, a precise millisecond-level timestamp and a 3D absolute spatial coordinate label are uniformly attached. The timestamp strictly corresponds to the actual collection time of the data point, and the spatial coordinates strictly match the fixed 3D position of the corresponding collection point in the physical warehouse. The process also involves retrieving pre-constructed static 3D visualization models of the warehouse. The visualization model pre-determines the 3D coordinates of fixed anchor points corresponding to the physical locations of stacker crane motors, unmanned transport equipment navigation modules, and cold chain evaporator return air vents, based on the actual physical warehouse layout. It then performs timestamp alignment and sorting on all tagged data streams, arranging all data in chronological order to smooth out minor sampling start time discrepancies among the three types of data collection points. This ensures consistent temporal sequence of multi-dimensional data at the same moment. Each data point after time alignment is precisely matched and mapped to the corresponding fixed anchor point and associated equipment component node pre-defined in the static 3D visualization model, based on its own 3D spatial coordinate label. The matching relationship between each data point and node is verified to eliminate misalignment and omissions, ultimately yielding a complete and accurate global data mapping result.

[0072] Based on the mapping results, the geometric structure, spatial pose, and operational attribute parameters of the stacker crane, unmanned transport equipment, and cold chain fan in the static 3D visualization model are dynamically updated to keep the representation content of the static 3D visualization model synchronized with the physical entities, forming a dynamic digital twin. Specifically, this includes: First, based on the obtained full-domain data mapping results, performing full-parameter dynamic updates on the static 3D visualization model to solve the problem that traditional static models cannot follow the changes in physical entity operation. A synchronous twin model is constructed, and geometric parameters are dynamically updated. Based on the mapping results, real-time changes in the stacker crane's telescopic lifting dimensions, the unmanned transport equipment's posture and external dimensions, and the cold chain fan's impeller opening and closing dimensions are synchronously acquired and directly replaced with the corresponding geometrical values ​​such as side length, height, and volume of the equipment components in the static 3D visualization model, updating the model's physical external structure in real time. Second, spatial pose parameters are updated, extracting the real-time 3D spatial coordinates of the equipment from the mapping data. By calculating the horizontal offset angle through the difference in horizontal coordinates and the pitch angle through the difference in vertical coordinates, the placement, rotation angle, and spatial attitude of the stacker crane, unmanned handling equipment, and cold chain fan in the static 3D visualization model are corrected in real time. This accurately replicates the real-time positioning status of the equipment and dynamically updates the operating attribute parameters. Real-time operating characteristic values ​​such as the harmonic distortion rate of the stacker crane, the energy consumption per unit mile of the unmanned handling equipment, and the temperature and humidity change rate of the cold chain warehouse are imported synchronously, and the operating attribute text and numerical labels bound to the model components are dynamically updated. Through frame-by-frame dynamic updates of three types of parameters—geometric structure, spatial pose, and operating attributes—all appearance, spatial positioning, and operating status representations of the static 3D visualization model are completely synchronized in real time with the actual physical entities of the stacker crane, unmanned handling equipment, and cold chain fan in the physical warehouse, without delay, deviation, or distortion. Ultimately, a stable dynamic digital twin that can replicate the physical operating status in real time is constructed.

[0073] In this embodiment of the invention, transient current harmonic acquisition points, real-time path tracking and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points are deployed at the winding end of the high-speed stacker crane drive motor, the navigation and positioning point of the unmanned handling equipment, and the return air vent side of the evaporator in the cold chain storage area, respectively. The sampling frequency and data reporting protocol of each acquisition point are configured, and a real-time data transmission link with the digital twin platform is established, thus constructing a multi-source high-frequency data acquisition network covering the core energy-consuming equipment of the warehouse. By capturing the three-phase current time-domain waveform of the stacker crane drive motor based on the real-time data transmission link and extracting the harmonic distortion rate characteristic value, the real-time position coordinates and energy consumption per unit mileage of the unmanned handling equipment, and the temperature and humidity change rate of the return air vent, a multi-source high-frequency data acquisition network is formed. Dynamic data streams provide multi-dimensional, high-frequency data support for energy consumption characterization across the entire warehouse. By attaching timestamps and spatial coordinate labels to the multi-source, high-frequency dynamic data streams, and based on the preset anchor point positions of each collection point in the static 3D visualization model, the timestamp-aligned data streams are mapped to the corresponding component nodes according to spatial coordinates. Based on the mapping results, the geometric structure, spatial pose, and operating attribute parameters of stacker cranes, unmanned handling equipment, and cold chain fans are dynamically updated, ensuring that the static 3D visualization model and physical entities remain synchronized in real time and form a dynamic digital twin. This avoids energy consumption characterization distortion caused by the asynchrony between the model and the entity, and improves the digital twin's ability to accurately depict the energy consumption status of the entire warehouse in real time.

[0074] In a preferred embodiment of the present invention, based on a dynamic digital twin, a dynamically parameterized spatial manifold is constructed according to the spatial distribution and data update frequency of transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points, including:

[0075] The spatial coordinates of each acquisition point are extracted from the dynamic digital twin, and harmonic distortion rate characteristic value sequence, location coordinate and unit mileage energy consumption sequence, and temperature and humidity change rate sequence are associated according to the data stream type corresponding to each acquisition point to form the spatiotemporal attribute vector of each acquisition point. Specifically, this includes: accurately extracting the fixed three-dimensional spatial coordinate values ​​of three types of core acquisition points from the completed dynamic digital twin. These coordinates strictly match the actual installation position of the acquisition point in the physical warehouse and digital model without offset error. According to the specific data acquisition type of different acquisition points, the corresponding complete time-series characteristic value sequence is bound to complete the accurate data association and matching; for transient current harmonic acquisition points, all time-by-time acquisition meters are arranged in a fixed sampling time order. The calculated harmonic distortion rate characteristic values ​​are spliced ​​together to form a continuous and uninterrupted time series of harmonic distortion rate characteristic values. For real-time path tracking and energy consumption mapping acquisition points, the three-dimensional position coordinate values ​​of the unmanned handling equipment accurately captured at each sampling time are synchronously bound, along with the energy consumption value per unit mileage calculated at the corresponding time, forming a dual-parameter time series of position coordinates and energy consumption per unit mileage. For temperature and humidity field dynamic response sensing acquisition points, all the combined values ​​of temperature change rate and humidity change rate calculated at each time are sorted by time to form a complete temperature and humidity change rate time series. The fixed three-dimensional spatial coordinates extracted from each acquisition point, the bound complete time series characteristic value sequence, and the time dimension index identifier are uniformly integrated to generate a unique spatiotemporal attribute vector for each acquisition point.

[0076] Using the spatial coordinates of each collection point as the node position and the data update frequency as the time axis sampling interval, a discrete node set is established. A continuous attribute field covering the storage space is constructed based on the spatiotemporal attribute vectors associated with each node. The isoparametric surface of the continuous attribute field is defined as the base surface of the dynamically parameterized spatial manifold. Specifically, this includes: building basic discrete nodes based on the generated spatiotemporal attribute vectors of all collection points and generating a continuous data field through global interpolation, defining the manifold's base surface; directly using the precise three-dimensional spatial coordinate values ​​of the three types of collection points and setting them as the fixed spatial location of each independent node within the basic discrete node set, without additional coordinate conversion correction; matching the predetermined data update frequency of each discrete node's corresponding collection point, accurately calculating the node's exclusive time axis sampling interval value; the data update frequency of the transient current harmonic collection point is 1000 times per second, and the time axis adjacent sampling interval value is equal to 0.001; the data update frequency of the real-time path tracking and energy consumption mapping collection point is 100 times per second, and the time axis adjacent sampling interval value is equal to 0.01; the temperature and humidity field dynamic... The data update frequency of the state response sensor acquisition points is ten times per second, and the adjacent sampling interval of the time axis is equal to 0.1. All independent nodes with dedicated spatiotemporal references are aggregated to form a complete and standardized discrete node set. Based on the spatiotemporal attribute vector bound to each node in the discrete node set, a continuous attribute field covering the entire three-dimensional space of the warehouse is constructed by global linear interpolation. Between any two adjacent discrete nodes, the difference in the three-dimensional spatial straight-line distance between the two nodes is calculated first, and then the difference in the energy consumption characteristic values ​​of the two nodes is calculated. The spatial distance between the intermediate position to be calculated and the starting node is divided by the total distance between the two nodes, multiplied by the difference in the energy consumption characteristic values ​​of the two nodes, and finally the original energy consumption characteristic value of the starting node is added to accurately calculate the transition energy consumption characteristic value of any blank point in the space, filling the data blind spot between discrete measurement points. The surface with the smoothest numerical change and the lowest degree of distortion is selected and extracted from the global continuous attribute field. This surface is defined as an isoparametric surface, and at the same time, this isoparametric surface is fixed as the unique basic base surface of the dynamic parameterized spatial manifold, completing the construction of the underlying geometric reference of the manifold.

[0077] On the base surface, based on the direction of harmonic distortion rate gradient change at transient current harmonic acquisition points, the density of motion trajectories at real-time path tracking and energy consumption mapping acquisition points, and the distribution of equipotential lines of temperature and humidity fields at dynamic response sensing acquisition points, non-uniform scaling factors are applied along the three principal directions of the base surface to obtain a dynamically parameterized spatial manifold. Specifically, this includes: on the determined base surface, differentially calculating and applying non-uniform scaling factors to optimize and adapt the dynamically parameterized spatial manifold to heterogeneous data; determining three mutually orthogonal reference principal directions on the base surface, corresponding to the horizontal lateral extension direction, horizontal longitudinal travel direction, and vertical height rise and fall direction of the storage site, respectively; generating exclusive scaling factors for each of the three types of acquisition point characteristics and precisely applying adjustments along the three principal directions; for transient current harmonic acquisition points, calculating the spatial variation gradient value of harmonic distortion rate between adjacent discrete nodes segment by segment. The gradient value is equal to the difference in harmonic distortion rate between two adjacent nodes divided by the spatial distance between the two nodes. The larger the gradient value, the more local the harmonic distortion rate changes. The more drastic the energy consumption fluctuations, the smaller the scaling factor should be set in the horizontal principal direction of the base surface to reduce the local grid spacing and improve the accuracy of detail representation. For real-time path tracking and energy consumption mapping acquisition points, the cumulative number of times the unmanned handling equipment traverses within a unit storage area is counted. The more cumulative number of trajectories, the more frequent the changes in regional energy consumption. Correspondingly, a smaller scaling factor should be set in the horizontal principal direction of the base surface to densify the geometric grid in densely packed trajectory areas. For temperature and humidity dynamic response sensor acquisition points, the density of temperature and humidity equipotential lines is counted. The denser the equipotential lines, the greater the difference in temperature and humidity energy consumption stratification in the vertical direction. Correspondingly, a smaller scaling factor should be set in the vertical principal direction of the base surface to accurately characterize the dynamic changes in temperature and humidity in layers. The calculated three sets of differentiated non-uniform scaling factors are then used to simultaneously stretch or compress the geometric dimensions of different regions of the surface along the three reference principal directions of the base surface to adaptively adapt to the changing characteristics of three types of heterogeneous energy consumption data, ultimately forming a non-uniform, highly adaptable, dynamically parameterized spatial manifold.

[0078] The dynamic parameterized spatial manifold is discretized into a grid topology, and the corresponding spatiotemporal attribute vectors are retained for each vertex. This completes the construction of the dynamic parameterized spatial manifold. Specifically, this includes: uniformly dividing the overall geometric surface of the completed dynamic parameterized spatial manifold along three main reference directions—horizontal transverse, horizontal longitudinal, and vertical height—according to a preset unified basic segmentation step size, decomposing the continuous surface into countless tiny regular geometric bodies, forming a complete and orderly three-dimensional grid topology. All intersection points of geometric bodies are uniformly defined as grid vertices. The three-dimensional spatial position values ​​corresponding to each grid vertex are accurately retrieved, and the exclusive spatiotemporal attribute vector of the nearest original acquisition point is matched. All original parameter data within the vector are completely retained, including fixed three-dimensional spatial coordinates, complete temporal feature value sequences, and time dimension index identifiers, without deletion or tampering with any original acquisition and calculation results, ensuring that the grid vertices inherit the real original energy consumption spatiotemporal characteristics. The three-dimensional grid topology is fully verified, checking that all grid vertices have no missing data, no spatial position misalignment, and no parameter matching errors. After confirming that the structure is regular and the data is complete and error-free, the overall construction of the dynamic parameterized spatial manifold is completely completed.

[0079] In this embodiment of the invention, by extracting the spatial coordinate values ​​of each collection point from the dynamic digital twin, and associating the harmonic distortion rate characteristic value sequence, the location coordinate and unit mileage energy consumption sequence, and the temperature and humidity change rate sequence according to the data stream type corresponding to each collection point, a spatiotemporal attribute vector for each collection point is formed, thereby achieving precise binding between the spatial location of the collection point and multi-source heterogeneous time-series energy consumption data; by establishing a discrete node set with the spatial coordinate values ​​of each collection point as node positions and the data update frequency as the time axis sampling interval, a continuous attribute field covering the storage space is constructed based on the spatiotemporal attribute vector associated with each node, and the isoparametric surface of the continuous attribute field is defined as the base surface of the dynamically parameterized spatial manifold, thereby realizing continuous energy consumption attributes from discrete measurement points to the entire storage area. The construction of the dynamic parameterized spatial manifold overcomes the limitation of traditional single-point sampling, which cannot cover the energy consumption status of the entire storage area. By applying non-uniform scaling factors along the three principal directions of the base surface based on the direction of harmonic distortion rate gradient change of transient current harmonic acquisition points, the density of motion trajectories of real-time path tracking and energy consumption mapping acquisition points, and the distribution of temperature and humidity field equipotential lines of temperature and humidity field dynamic response sensor acquisition points, a dynamic parameterized spatial manifold is obtained. This adapts to the heterogeneous data characteristics of different types of acquisition points and avoids the problem of uneven accuracy in representing the energy consumption status of different devices caused by traditional uniform grid modeling. The construction of the dynamic parameterized spatial manifold is completed by discretizing the dynamic parameterized spatial manifold into a grid topology and retaining the corresponding spatiotemporal attribute vector for each vertex.

[0080] In a preferred embodiment of the present invention, multi-source high-frequency dynamic data is mapped to discrete nodes of a spatial manifold according to timestamps and spatial coordinates to obtain a synchronized running state stream, including:

[0081] The three-dimensional spatial coordinates of each discrete node are extracted from the mesh topology. Based on the spatiotemporal attribute vector type of each node, the spatial affiliation between the node and transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points is established. This determines the dominant acquisition point affiliation of each discrete node in space. Specifically, this includes: extracting the precise three-dimensional spatial coordinate values ​​of all discrete nodes in batches from the constructed dynamic parameterized spatial manifold mesh topology. These coordinates strictly correspond to the absolute position of the node in the digital twin warehouse space, including specific values ​​for three dimensions: horizontal coordinate, horizontal coordinate, and vertical height coordinate, without coordinate offset or missing dimensions; and extracting the spatiotemporal attribute vectors bound to each discrete node. The vector type identifier distinguishes which of the three types of acquisition points the node is initially associated with: transient current harmonics, real-time path tracing and energy consumption mapping, and temperature and humidity dynamic response sensing. Based on the spatial distance determination rule, the spatial membership relationship between the node and the three types of acquisition points is established: First, the spatial straight-line distance between the current discrete node's three-dimensional spatial coordinates and the three-dimensional coordinates of all transient current harmonic acquisition points in the warehouse is calculated. The calculation method is to add the square of the difference between the horizontal coordinates of the two points to the square of the difference between the horizontal coordinates and the square of the difference between the vertical coordinates, and then take the square root of the summation result to obtain the spatial straight-line distance. The same method is used to calculate the spatial straight-line distance between the node and all real-time path tracing and energy consumption mapping acquisition points, and all temperature and humidity dynamic response sensing acquisition points.

[0082] The minimum spatial distance from the node to each of the three types of acquisition points is calculated, and the type of acquisition point corresponding to the minimum distance is determined as the spatial affiliation type of the node. If the minimum distance from a node to the transient current harmonic acquisition point is less than the minimum distance to the other two types of acquisition points, then the node belongs to the spatial region to which the transient current harmonic acquisition point belongs. At the same time, the nearest acquisition point of the same type to the node is determined as the dominant acquisition point affiliation of the node in space, thus completing the binding of the spatial affiliation relationship between all discrete nodes and the three types of acquisition points.

[0083] Based on three parallel multi-source high-frequency dynamic data streams, and using the set data update frequency of each acquisition point as a benchmark, a sliding window alignment process is performed on the time axis of each data stream to obtain a synchronized data frame sequence with a unified time benchmark. Specifically, this includes: addressing the deficiency of asynchronous time axis analysis and inability to uniformly analyze multi-source high-frequency dynamic data streams due to different sampling frequencies, a sliding window alignment process is performed on the three parallel data streams to unify the time benchmark. The core benchmark rule for time alignment is determined, using the highest sampling frequency of transient current harmonic acquisition points (1000 times per second) among the three types of acquisition points as the core time benchmark. The base time length of the sliding window is set to 1 millisecond, and the sliding step size is consistent with the window length, both being 1 millisecond, to ensure that there are no omissions or overlaps in the time axis. The three original data streams are processed separately: for the transient current harmonic data stream, whose sampling frequency is consistent with the benchmark frequency, the original sequence is directly split according to the rule of one data frame every 1 millisecond; for the real-time path tracking... The tracking and energy consumption mapping data stream is sampled 100 times per second. During sliding window alignment, each original data frame is copied and filled into 10 consecutive 1-millisecond sliding windows to ensure that the data covers all windows in the corresponding time interval. For the temperature and humidity field dynamic response sensing data stream, the sampling frequency is 10 times per second. Each original data frame is copied and filled into 100 consecutive 1-millisecond sliding windows to cover the corresponding time interval. During the filling process, the consistency of repeated data in each sliding window is checked. If different data values ​​of the same collection point appear in the same window, the data value with the timestamp closest to the center of the window is taken as the final value of the window. After all windows are filled, the window data of the three data streams are matched and integrated one by one in chronological order. Each time window corresponds to a complete data frame containing data from the three types of collection points. All data frames are arranged from early to late according to the time axis to form a synchronous data frame sequence with a unified time reference.

[0084] For each synchronized data frame, based on the determined spatial affiliation between nodes and acquisition points, the harmonic distortion rate characteristic value, location coordinates, energy consumption per unit mileage, and temperature and humidity change rate corresponding to each acquisition point are assigned to discrete nodes within their respective spatial regions. This ensures that each discrete node obtains a fusion attribute value that matches its spatial location. Specifically, this involves: relying on the determined node-acquisition point spatial affiliation and the generated synchronized data frame sequence, accurately allocating multi-source data to discrete nodes across the entire domain, assigning fusion attribute values ​​to nodes that match their spatial locations, and traversing each data frame in the synchronized data frame sequence. The core data of three types of acquisition points within the frame are extracted: harmonic distortion rate characteristic values ​​of transient current harmonic acquisition points, location coordinates and energy consumption per unit mileage of real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity change rate values ​​of temperature and humidity field dynamic response sensing acquisition points. For each type of acquisition point data, it is allocated to the corresponding discrete nodes according to spatial affiliation: for discrete nodes belonging to transient current harmonic acquisition points, a distance weighting allocation rule is used to calculate the spatial straight-line distance from the node to its dominant acquisition point, and at the same time, the sum of the total distances from all subordinate nodes within the radiation range of the dominant acquisition point to the acquisition point is calculated. , It is the characteristic value of harmonic distortion rate at the dominant acquisition point. It is the first Harmonic distortion rate allocation values ​​for each node It is the sum of the total distances within the radiation range. No. The discrete nodes are spatially linearly distant; for discrete nodes belonging to the real-time path tracing and energy consumption mapping acquisition points, the position coordinates of the dominant acquisition point in this frame and the energy consumption value per unit mileage are directly assigned to the node. If multiple nodes belong to the same dominant acquisition point, all nodes are assigned this set of data; for discrete nodes belonging to the temperature and humidity field dynamic response sensing acquisition points, a linear interpolation allocation rule is adopted. With the dominant acquisition point as the center, the temperature and humidity change rate values ​​are linearly adjusted according to the spatial distance between the node and the acquisition point. The farther the distance, the lower the temperature and humidity change rate values ​​are proportional to the distance.

[0085] After the single-frame data allocation is completed, each discrete node obtains a fusion attribute value that is highly matched with its own spatial location, including one or more parameters such as harmonic distortion rate, location coordinates and energy consumption per unit mileage, and temperature and humidity change rate, so as to realize the accurate landing of multi-source data to the spatial nodes of the whole domain.

[0086] The assigned discrete node fusion attribute values ​​are fused with the retained spatiotemporal attribute vector using Kalman filtering to obtain the real-time state vector of all discrete nodes. Specifically, this involves: fully retrieving the measured fusion attribute values ​​of any discrete node at the current sampling moment after assignment; simultaneously, accurately reading the unique spatiotemporal attribute vector permanently stored for that discrete node. This vector not only contains the node's fixed three-dimensional physical space coordinates and the original complete time-series energy consumption characteristic sequence, but also continuously iteratively stores the node's final steady-state operating state after historical filtering convergence, and core filtering foundation data such as the unique historical error benchmark. Then, based on the acquisition type corresponding to the discrete node, preset fixed noise benchmark parameters are matched, and unique benchmark ranges are configured to distinguish three different data noise characteristics: transient current harmonics, real-time path energy consumption, and temperature and humidity field. This completes all pre-filtering foundation data preparation. Afterwards, the Kalman filtering prediction stage begins. Relying on the historical steady-state foundation data retained in the spatiotemporal attribute vector, combined with fixed state influence weight rules that conform to the slow inertial change law of energy consumption in storage equipment, the theoretical predicted steady state of the node, unaffected by the current instantaneous sampling noise, is smoothly deduced. Simultaneously, combined with the preset noise benchmark range for the corresponding data type, precise measurement... To mitigate the inherent minor fluctuations in the theoretical deduction model, a reliable theoretical reference standard is established. This is followed by an adaptive Kalman filter update phase. First, based on the inherent fluctuations in the theoretical deduction and the factory-defined measurement accuracy of the acquisition equipment, the fusion priority weights of the theoretical steady-state data and the measured data are adaptively assigned. This prioritizes measured values ​​with lower noise and historical steady-state values ​​with lower theoretical fluctuations. Then, combining these adaptive priority weights, the precise measured attribute values ​​of the current node are fused, gently correcting minor deviations generated during the theoretical deduction process. This thoroughly filters out random acquisition noise and instantaneous data jumps caused by various external interferences, converging to obtain a high-precision, distortion-free final steady-state energy consumption value for the node. Simultaneously, the node's unique error reference data is iteratively updated, and the new reference is permanently stored in the node's spatiotemporal attribute vector as the core basis for filtering iteration at the next sampling time. Finally, the entire domain data is integrated and output. The high-precision final steady-state energy consumption core parameters of a single discrete node after filtering convergence are completely encapsulated along with the node's fixed three-dimensional physical space coordinates, the iteratively updated new error reference, and the original time-series unique identifier, generating a standardized real-time state vector for the node.

[0087] The real-time state vectors of all discrete nodes are sequentially arranged according to the obtained synchronization data frame sequence and encapsulated into an iteratively updatable synchronous running state stream. Specifically, this includes: addressing the problem that the real-time state vectors of discrete nodes lack a unified temporal structure and cannot support subsequent iterative simulations, encapsulating the real-time state vectors of all nodes into an iteratively updatable synchronous running state stream along the time axis; extracting the complete time axis index of the generated synchronization data frame sequence, containing all millisecond-level timestamps from the start to the end of data acquisition, and arranging them in chronological order to form an ordered time index table; traversing the synchronization data frame corresponding to each timestamp, extracting the real-time state vectors of all discrete nodes under that frame after Kalman filtering fusion, and arranging the real-time state vectors of all nodes under the same timestamp... State vectors are integrated into a single state frame, which includes a timestamp identifier, a list of real-time state vectors for all nodes, and a data integrity verification identifier. All state frames are sequentially concatenated according to the time index table, forming a continuous, time-axis-arranged sequence. To enable iterative updates, a data appending interface and data update rules are reserved within this sequence: when a new synchronization data frame is generated and node vector fusion is completed, it is automatically appended to the end of the sequence in timestamp order. If a state frame at a certain timestamp requires data correction, the frame can be located via its timestamp index, and the real-time state vectors within it can be replaced. Simultaneously, the state transition calculation baseline for subsequent frames is automatically updated. This ordered, appendable, and updatable state frame sequence is encapsulated as a synchronous running state stream.

[0088] In this embodiment of the invention, the three-dimensional spatial coordinates of each discrete node are extracted from the mesh topology, and the spatial affiliation between the node and transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points are established according to the spatiotemporal attribute vector type of each node, thus determining the dominant acquisition point affiliation of each discrete node in space. Based on three parallel multi-source high-frequency dynamic data streams, a sliding window alignment process is performed on the time axis of each data stream using a set data update frequency as a benchmark, resulting in a synchronized data frame sequence with a unified time benchmark. This achieves precise time axis alignment of multi-source heterogeneous high-frequency data, avoiding fusion distortion caused by data time misalignment. For each synchronized data frame, based on the determined node and... The spatial affiliation between collection points assigns the harmonic distortion rate characteristic value, location coordinates, energy consumption per unit mileage, and temperature and humidity change rate of each collection point to discrete nodes within their respective spatial regions. This allows each discrete node to obtain a fusion attribute value that matches its spatial location, achieving precise spatial fusion of multi-source energy consumption data from discrete nodes across the entire warehouse. By fusing the assigned fusion attribute values ​​of each discrete node with the retained spatiotemporal attribute vector using Kalman filtering, the real-time state vector of all discrete nodes is obtained, suppressing noise interference in the multi-source data and improving the real-time performance and accuracy of the node state data. Finally, by arranging the real-time state vectors of all discrete nodes in a temporal sequence according to the obtained synchronous data frame sequence, it is encapsulated into an iteratively updatable synchronous running state stream.

[0089] In a preferred embodiment of the present invention, based on the synchronous running state flow, an isoparametric transformation is performed on the dynamically parameterized spatial representation to obtain the transformed projection domain; the boundary contours corresponding to each acquisition point are extracted on the transformed projection domain, and multi-layered nested buffer boundaries are obtained by successively offsetting along the normal direction, including:

[0090] The mesh topology and real-time state vectors of the dynamic parameterized spatial representation are extracted from the synchronous running state stream. A mapping relationship from 3D physical space to 3D attribute space is constructed using the node coordinates and real-time state vectors. Specifically, this includes: extracting the core basic data of the dynamic parameterized spatial representation from the completed synchronous running state stream: First, the global mesh topology, containing complete topological information such as the 3D physical space coordinates of all discrete nodes, the connection relationships between nodes, and the geometric dimensions of mesh cells; second, the real-time state vector bound to each discrete node, containing all energy consumption data such as harmonic distortion rate characteristic values, location coordinates and energy consumption per unit distance, temperature and humidity change rate values, and error covariance. Attribute parameters ensure data integrity and authenticity; based on extracted core data, a precise mapping relationship is constructed from 3D physical space to 3D attribute space: First, a dual-space coordinate system is defined for each discrete node. The 3D physical space coordinates are the node's horizontal, vertical, and height coordinates in the real physical scenario of the warehouse; the 3D attribute space coordinates are constructed from the core energy consumption parameters in the node's real-time state vector, using the harmonic distortion rate characteristic value as the first dimension coordinate, the energy consumption per unit mileage as the second dimension coordinate, and the temperature and humidity change rate as the third dimension coordinate; a one-to-one mapping rule is established between the two spaces: for any discrete node, its 3D physical space coordinates... A unique set of three-dimensional attribute space coordinates =Harmonic distortion rate, =Energy consumption per unit distance = Rate of change of temperature and humidity; For any grid cell in the grid topology, its physical spatial geometry uniquely corresponds to the energy consumption attribute distribution in the attribute space.

[0091] Applying isoparametric transformations to the mapping relationship maps the dynamic parameterized spatial representation to the parameterized projection domain in the three-dimensional attribute space, creating a one-to-one correspondence between spatial positions and attribute field quantities, thus achieving parameterized dimensionality reduction. Specifically, to address the redundancy and low efficiency of direct simulation in three-dimensional physical space, isoparametric transformations are applied to the constructed mapping relationship to achieve parameterized dimensionality reduction in high-dimensional space. The core of the isoparametric transformation is to map the coordinates of both the three-dimensional physical space and the three-dimensional attribute space to the same set of parameterized coordinates using unified parameterized basis functions. , , The range of parameter coordinates is uniformly set to 0≤ , , ≤1, the specific calculation process is as follows:

[0092] The cubic B-spline basis functions are used as the core transformation basis functions, and the basis function values ​​are derived from the parameter coordinates ( , , The result is obtained by calculating the cubic polynomial of ), where the parameter coordinates are... Corresponding basis functions The calculation formula is: Parameter coordinates Corresponding basis functions Parameter coordinates Corresponding basis functions Calculation method and Completely identical, that is: , .

[0093] For any discrete node in the dynamic parameterized spatial representation mesh topology, its three-dimensional physical space coordinates The calculation formula is as follows: Mapping to the parameter space via isoparametric transformation. ; ; ;

[0094] This represents the total number of control points in the grid cells to which the discrete node belongs. , , These are the first and second cells within the grid cell, respectively. The three-dimensional physical space coordinate components of each control point For the first The cubic B-spline basis function values ​​corresponding to each control point. This indicates all within the grid cell. The calculation results for each control point are summed.

[0095] Synchronized with the physical space transformation, the three-dimensional attribute space coordinates corresponding to this discrete node The formula for mapping to the parameter space is: ; ; ; , , These are the first and second cells within the grid cell, respectively. The three-dimensional attribute space coordinate components of each control point; the definitions of the remaining parameters are consistent with the physical space transformation formula.

[0096] Through the aforementioned isoparametric transformation, the dynamic parameterized spatial representation is synchronously mapped from the three-dimensional physical space and the three-dimensional attribute space to the three-dimensional parameterized projection domain. Each set of parameter coordinates within the projection domain It uniquely corresponds to a location in the three-dimensional physical space and also uniquely corresponds to a set of energy consumption attribute values ​​in the three-dimensional attribute space, realizing a one-to-one correspondence between spatial location and attribute field quantity, and completing the parameterized dimensionality reduction expression.

[0097] In the parametric projection domain, with the parameter coordinates corresponding to each acquisition point as the center, thresholds are set based on the characteristic values ​​of harmonic distortion rate, peak energy consumption per unit mileage, and extreme values ​​of temperature and humidity change rate captured by each acquisition point. Contour lines are extracted as the initial boundary contour. Specifically, this includes: determining the core parameter coordinates of each acquisition point in the parametric projection domain; extracting the discrete node parameter coordinates corresponding to transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points. , serving as the central coordinate for various energy consumption anomaly analyses.

[0098] Then, a differentiation threshold is set:

[0099] Harmonic distortion rate threshold: The maximum and average values ​​of the historical harmonic distortion rate characteristic values ​​at this sampling point are statistically analyzed. The threshold is equal to the average value plus 0.5 times (maximum value minus average value), that is, threshold = average value + 0.5 × (maximum value - average value).

[0100] Peak energy consumption threshold per unit mileage: The 95th percentile of the historical energy consumption per unit mileage at this collection point is used as the threshold.

[0101] Extreme threshold for temperature and humidity change rate: Calculate the maximum absolute value of the historical temperature and humidity change rate at this collection point and set this maximum value as the threshold.

[0102] Contour lines are extracted based on the set thresholds as the initial boundary profile: Within the parameterized projection domain, with the core parameter coordinates of the acquisition points as the center, all parameter coordinate points within the projection domain are traversed, and all parameter coordinate points whose harmonic distortion rate, energy consumption per unit mileage, and temperature and humidity change rate are selected are selected. These points are then connected sequentially according to spatial continuity to form closed contour lines. The contour lines of the three types of acquisition points are marked as harmonic distortion rate boundaries, energy consumption peak boundaries, and temperature and humidity change rate boundaries, respectively. The three types of boundaries are then integrated to form a complete initial boundary profile.

[0103] The initial boundary contour is inversely mapped back to 3D physical space to obtain the corresponding 3D boundary contour lines. These contour lines are then discretized to obtain polygonal boundaries composed of ordered point sequences. Specifically, to ensure that the energy consumption anomaly boundary closely matches the actual warehouse physical space layout, the initial boundary contour extracted in step 3.3 is inversely mapped back to 3D physical space and discretized. The core of the inverse mapping is to convert the contour line parameter coordinates within the parameterized projection domain... The specific calculation process for restoring the coordinates to three-dimensional physical space is as follows:

[0104] For any parametric coordinate point on the initial boundary contour Its corresponding three-dimensional physical space coordinates The result is equal to the sum of the physical coordinates of all control points in the grid cell to which the parameter point belongs, multiplied by their corresponding basis function values. The basis function values ​​are derived from the parameter coordinates. It is calculated using the cubic B-spline basis function formula.

[0105] By performing inverse mapping on all parameter coordinate points on the initial boundary contour one by one, a continuous curve composed of three-dimensional physical space coordinate points is obtained, namely the three-dimensional boundary contour line. The three-dimensional boundary contour line is then discretized: points are uniformly selected along the three-dimensional boundary contour line according to a preset dispersion length, and the three-dimensional physical space coordinates of each selected point are recorded. These coordinate points are arranged sequentially according to the direction of the contour line to form an ordered point sequence. The first and last coordinate points of the ordered point sequence are connected to form a closed polygonal boundary. The number of side segments of each polygonal boundary is determined by the contour line length and the dispersion length.

[0106] For each polygon boundary, the normal direction of each side segment is calculated. The extended boundary is obtained by successively moving outward along the normal direction with a preset offset step size. The original boundary and the extended boundary are numbered hierarchically according to the offset distance to obtain a multi-layer nested buffer boundary sequence. Specifically, the normal direction of each side segment of the polygon boundary is calculated. The three-dimensional physical space coordinates of the two endpoints of a certain side segment are extracted, namely endpoint A (X1, Y1, Z1) and endpoint B (X2, Y2, Z2). The direction vector of the side segment is calculated: direction vector X component = X2 - X1, direction vector Y component = Y2 - Y1, direction vector Z component = Z2 - Z1.

[0107] Calculate the normal vector of the edge segment: normal vector X component = -direction vector Y component, normal vector Y component = direction vector X component, normal vector Z component = 0 (normal vector in the horizontal plane of the storage). Then, normalize the normal vector: normalized normal vector = normal vector ÷ normal vector magnitude, normal vector magnitude = √(square of normal vector X component + square of normal vector Y component + square of normal vector Z component), ensuring that the normal vector length is 1 and the direction points to the outside of the polygon boundary.

[0108] Repeat the above calculation for all sides of the polygon boundary to obtain the normalized outer normal vector of each side.

[0109] Then, the extended boundary is generated by successively offsetting along the normal direction: a fixed offset step size is preset (set to the side length of the smallest grid unit in the physical space of the warehouse). First, all vertices of the polygon boundary are moved by one offset step size along the normalized normal vector of the corresponding edge segment. After the movement, the vertex coordinates = original vertex coordinates + offset step size × normalized normal vector. The moved vertices are connected in the original order to obtain the first layer of extended boundary. Then, all vertices of the first layer of extended boundary are moved by one offset step size along the normal vector to obtain the second layer of extended boundary. This process is repeated to generate an extended boundary with a preset number of layers (set to three layers).

[0110] Finally, all boundaries are numbered hierarchically: the original polygonal boundaries are numbered as level 0 (core boundaries), the first layer of extended boundaries are numbered as level 1 (transition boundaries), and the second layer of extended boundaries are numbered as level 2 (outer boundaries). All boundaries are integrated in order of number to form a multi-layered nested buffer boundary sequence. This sequence accurately delineates the gradient range of energy consumption disturbances, laying a solid geometric foundation for subsequent partitioned energy consumption simulation.

[0111] In this embodiment of the invention, by extracting the mesh topology of the dynamic parameterized spatial representation and the real-time state vectors of each discrete node from the synchronous running state flow, a precise mapping relationship from three-dimensional physical space to three-dimensional attribute space is constructed. This binds the geometric location of the storage space with the multi-dimensional energy consumption attribute state, overcoming the defect of the separation between spatial layout and energy consumption parameters in traditional energy consumption boundary partitioning. By applying an isoparametric transformation to the mapping relationship, the spatial representation is mapped to the parameterized projection domain, achieving a one-to-one correspondence between spatial location and attribute field quantity in parameterized dimensionality reduction expression. This simplifies the computational complexity of high-dimensional global energy consumption simulation and avoids the computational complexity of three-dimensional complex objects. This paper addresses the problems of redundant computing power and low computational efficiency in direct simulation of physical space. By extracting contour lines centered on the coordinates of the collected points in the projection domain and combining them with harmonic distortion rate, peak energy consumption, and extreme temperature and humidity thresholds as the initial boundary profile, the core range of various energy consumption anomaly sources is accurately located. By inversely mapping the initial contour back to three-dimensional physical space and discretizing it into ordered polygonal boundaries, the paper ensures that the abnormal boundaries are highly consistent with the actual warehouse physical space layout without geometric distortion. By calculating the normal of the polygonal side segments and successively offsetting them according to a preset step size to generate a nested buffer boundary sequence with hierarchical numbering, a gradient energy consumption disturbance boundary hierarchy system is accurately constructed.

[0112] In a preferred embodiment of the present invention, the spatial representation is divided into a core disturbance region, a transitional influence region, and a peripheral stable region based on the buffer boundaries of each layer. Numerical simulations of transient energy consumption surge trends are then performed on each region to obtain the pre-simulation results of abnormal energy consumption surge conditions and the corresponding pre-regulation parameter set, including:

[0113] The innermost boundary is extracted from the multi-layered buffer boundary sequence as the core perturbation boundary, and the region inside this boundary is marked as the core perturbation zone. The middle layer boundaries are extracted as transitional influence boundaries, and the region between the core perturbation boundary and the transitional influence boundary is marked as the transitional influence zone. The region outside the outermost boundary is marked as the peripheral stable zone. Specifically, this includes: accurately dividing the storage space into three zones based on the generated multi-layered nested buffer boundary sequence; extracting the geometric information of all levels of boundaries from the buffer boundary sequence, including complete data such as the three-dimensional physical space coordinates of the polygon vertices of each boundary and the boundary closure range.

[0114] Core disturbance boundary extraction: The original polygon boundary numbered 0 in the sequence is selected as the innermost core disturbance boundary, which is the core range of the energy consumption anomaly disturbance source;

[0115] Transitional influence boundary extraction: The first layer extension boundary numbered 1 in the sequence is selected as the transitional influence boundary of the intermediate layer. This boundary is the transition range of the core disturbance spreading outward.

[0116] Extraction of the outermost stable boundary: The second-level extended boundary numbered 2 in the sequence is selected as the outermost outer stable boundary. The area outside this boundary is the steady-state operating region of warehouse energy consumption.

[0117] Subsequently, a region attribution determination rule is established, which determines the region to which a discrete node belongs based on its three-dimensional physical space coordinates:

[0118] Determining the core disturbance zone: Calculate the spatial relationship between the three-dimensional coordinates of a discrete node and the core disturbance boundary. If the node coordinates are completely inside the closed polygon of the core disturbance boundary, the determination method is to draw a ray from the node in any direction. If the number of intersections with the boundary segment is odd, then the node is marked as a core disturbance zone node. The spatial range covered by all core disturbance zone nodes is the core disturbance zone.

[0119] Determination of Transitional Influence Zone: If a node's coordinates are located in the annular region between the core disturbance boundary and the transitional influence boundary, and are neither inside the core disturbance boundary nor outside the transitional influence boundary, then it is marked as a node in the transitional influence zone. The area covered by all such nodes is the transitional influence zone.

[0120] Determining the outer stable region: If the node coordinates exceed the closed range of the transition influence boundary, it is marked as an outer stable region node, and the area covered by all such nodes is the outer stable region.

[0121] After completing the regional affiliation labeling of all discrete nodes, the grid topology of each regional node is integrated to form a three-zone spatial division system with clear boundaries and well-defined affiliations.

[0122] For the core disturbance zone, the real-time state vectors of discrete nodes within the zone are extracted to simulate the instantaneous impact of stacker crane start-up and shutdown, equipment path intersection, and fan load fluctuations on local energy consumption. This yields the predicted peak and duration of energy consumption surges. Specifically, the real-time state vectors of all nodes are extracted from the grid topology of the core disturbance zone, including complete parameters such as harmonic distortion rate characteristic values, energy consumption per unit mileage, temperature and humidity change rate values, and timestamps, ensuring that the data covers all dimensions of energy consumption characteristics required for simulation.

[0123] The instantaneous impact of local energy consumption was then simulated in three scenarios:

[0124] Stacker crane start-stop impact simulation: Extract the harmonic distortion rate time sequence of the corresponding nodes of the stacker crane drive motor in the core disturbance area to simulate the current impact during start-stop operation. The impact effect is calculated as follows: , It is a natural exponential function. The instantaneous harmonic distortion rate at a certain moment. As the reference harmonic distortion rate, The start-stop impact coefficient. For the time difference, The impact decay time is defined as follows: the start-stop impact coefficient is set according to the rated power of the stacker crane (coefficient value = rated power ÷ 1000), the time difference is the time interval between the current moment and the start-stop action trigger moment, and the impact decay time is set to 5 seconds.

[0125] Equipment path intersection impact simulation: Extract the unit mileage energy consumption time series of the corresponding nodes of unmanned transport equipment, and simulate the surge in energy consumption when multiple equipment paths are congested. The number of congested equipment is the total number of unmanned transport equipment with intersecting paths in the core disturbance area at the same time.

[0126] Simulation of fan load fluctuation impact: Extract the time series of temperature and humidity change rates of corresponding nodes of cold chain fans to simulate the energy consumption impact of fan load fluctuation.

[0127] After completing the superposition calculation of the three types of impact effects, the peak value and duration of the energy consumption surge in the core disturbance zone were statistically analyzed:

[0128] Calculation of peak energy consumption surge: Iterate through the instantaneous energy consumption values ​​of all nodes during the simulation period, namely harmonic distortion rate, energy consumption per unit distance, and temperature and humidity change rate, and sum them by weight, with weights of 0.5, 0.3, and 0.2 respectively. Take the maximum value as the peak energy consumption surge.

[0129] Duration calculation: Determine the start and end times when the energy consumption value exceeds 1.5 times the baseline value. The baseline value is the average energy consumption value when there is no impact in the core disturbance area.

[0130] The final output shows the peak value of energy consumption surge in the core disturbance region, the time of peak occurrence, and the duration of the impact, solving the problem that traditional methods cannot predict the extreme value of transient energy consumption.

[0131] For the transitional influence zone, the state vectors of discrete nodes within the zone are extracted. The attenuation law and delayed response characteristics of energy consumption shock propagating outward along the grid topology are simulated to obtain the predicted energy consumption range and lag time. Specifically, this includes: extracting the state vectors of all discrete nodes within the transitional influence zone, including parameters such as the node's spatial coordinates, energy consumption characteristic values, timestamp, and spatial distance from the core disturbance zone; calculating the attenuation coefficient of the energy consumption shock propagating outward along the grid topology. , Let be the energy consumption value of a discrete node after decay. The peak energy consumption in the core disturbance area This represents the shortest propagation distance between the node and the core disturbance boundary. To preset the energy consumption attenuation length, where the propagation distance is the shortest straight-line distance between the node and the core disturbance boundary, and the attenuation length is set to 10 times the side length of a single grid cell in the storage area, all nodes in the transition influence zone are traversed, and the energy consumption value at different propagation distances is calculated according to this formula. The maximum propagation distance when the energy consumption value drops to 50% of the peak value is statistically analyzed, and this distance is the energy consumption impact range. The time delay of energy consumption impact propagation is calculated, and the delay time is positively correlated with the propagation distance. The energy consumption propagation speed is set to 0.5 m / s based on the layout of the storage equipment. For each node in the transition influence zone, the lag time of receiving the energy consumption impact is calculated, and the maximum value and distribution pattern of the lag time of all nodes are statistically analyzed. The lag time prediction result of the energy consumption impact is output.

[0132] For the outer stable zone, steady-state operating parameters of discrete nodes within the zone are extracted. The simulation results of the core disturbance zone and the transitional influence zone are compared with the baseline to identify anomalies exceeding the threshold, thus obtaining anomaly overflow pre-simulation results. Specifically, this includes: extracting steady-state operating parameters of all discrete nodes within the outer stable zone. The steady-state parameters are the average energy consumption values ​​during the simulation period without energy consumption impacts, including steady-state harmonic distortion rate, steady-state energy consumption per unit mileage, and steady-state temperature and humidity change rate; establishing an energy consumption baseline threshold for the outer stable zone, which is set at 1.2 times the steady-state parameters. This threshold takes into account normal warehouse operations. The energy consumption fluctuation range and anomaly identification sensitivity of the row are analyzed. The simulated energy consumption results of the core disturbance area and the transition influence area, including the energy consumption values ​​of each node at different times, are mapped to the grid topology of the outer stable area. The simulated energy consumption value of each node is compared with the baseline threshold. If the energy consumption deviation value of a node is >0 and the simulated energy consumption value is > the baseline threshold, the node is marked as an abnormal overflow point. If the deviation value is ≤0 or the simulated energy consumption value is ≤ the threshold, it is determined to be a normal steady-state node. The number, spatial distribution location and the magnitude of exceeding the threshold of abnormal overflow points in the outer stable area are counted to form the abnormal overflow pre-simulation results.

[0133] Based on the results of the pre-simulation in various regions, the optimized values ​​of the stacker crane start-stop frequency, the adjustment values ​​of the energy consumption weight of the unmanned handling equipment path, and the pre-regulation values ​​of the cold chain fan speed are calculated to obtain the pre-regulation parameter set. Specifically, it includes: based on the peak energy consumption surge and duration in the core disturbance area, the optimized value of the start-stop frequency is calculated with the goal of reducing the peak to within 1.2 times the baseline. The ratio of the peak energy consumption under the original start-stop frequency to the baseline is calculated. If the peak ratio is > 1.2, the optimized value of the start-stop frequency = original start-stop frequency × (1 - (peak ratio - 1.2) ÷ peak ratio); if the peak ratio is ≤ 1.2, the optimized value of the start-stop frequency remains unchanged from the original frequency. For example, if the original start-stop frequency is 10 times / hour and the peak ratio is 1.5, then the optimized value = 10 × (1 - (1.5 - 1.2) ÷ 1.5) = 10 × 0.8 = 8 times / hour.

[0134] Based on the energy consumption spillover range of the transitional impact zone, the path energy consumption weight is adjusted with the goal of reducing the spillover range to within 80% of the original range. The energy consumption spillover range under the original weight is calculated, and the original range is the maximum propagation distance of the transitional impact zone. The path energy consumption weight adjustment value = original weight × (1 + (original range - target range) ÷ original range), where the target range = 0.8 × original range. The path energy consumption weight is used for path planning by the navigation scheduler. The higher the weight, the greater the probability that the path will be preferentially avoided during planning.

[0135] Based on the number of abnormal overflow points in the peripheral stable zone, the pre-regulation value of the fan speed is calculated with the goal of reducing the number of abnormal overflow points to 0. The number of abnormal overflow points at the original speed is calculated as N. If N > 0, the pre-regulation value of the fan speed = original speed × (1 - N ÷ total number of nodes), where the total number of nodes is the total number of nodes in the peripheral stable zone. If N = 0, the pre-regulation value of the fan speed remains unchanged at the original speed. For example, if the original speed is 1500 rpm, the number of abnormal overflow points is 10, and the total number of nodes is 100, then the pre-regulation value = 1500 × (1 - 10 ÷ 100) = 1500 × 0.9 = 1350 rpm.

[0136] The calculated optimized values ​​for stacker crane start-stop frequency, unmanned transport equipment path energy consumption weight adjustment values, and cold chain fan speed pre-control values ​​are integrated, classified and packaged according to equipment type, to form a complete set of pre-control parameters.

[0137] In this embodiment of the invention, a gradient-based spatial hierarchical system for energy consumption anomalies is constructed by accurately dividing the core disturbance zone, transition influence zone, and peripheral stable zone from a multi-layered buffer boundary sequence. By extracting real-time state vectors of nodes in the core disturbance zone, the instantaneous energy consumption impacts of stacker crane start-up and shutdown, equipment path intersection, and fan load fluctuations are simulated, and the peak and duration prediction results of energy consumption surges are accurately quantified and output, overcoming the technical challenge of predicting sudden transient energy consumption extremes of equipment. By simulating the propagation attenuation law and delayed response characteristics of energy consumption impacts along the grid topology in the transition influence zone, the energy consumption impact range and lag time are accurately deduced. This technology overcomes the shortcomings of traditional energy consumption prediction, which ignores the delayed effects of cascading disturbances. By comparing the stable operating baseline with the surrounding stable zone, it identifies abnormal overflow points and conducts a comprehensive investigation of hidden energy consumption imbalances, preventing the spread of minor local disturbances from causing global energy waste. By integrating the multi-dimensional and precise simulation results from the three zones, it quantitatively solves the start-stop frequency of stacker cranes, the energy consumption weight of unmanned handling equipment paths, and the exclusive optimized values ​​of cold chain fan speeds, and forms a pre-control parameter set. This enables precise output of the basis for pre-emptive, layered, quantitative prediction and customized control of transient energy consumption anomalies, improving the refinement, quantification, and practical adaptability of numerical simulation of sudden surges in warehouse energy consumption.

[0138] like Figure 2 As shown, in another preferred embodiment of the present invention, dynamic adaptive control processing is performed based on the pre-simulation results and the pre-control parameter set to obtain equipment operation feedback data after dynamic adaptive control processing, including:

[0139] The optimized start-stop frequency values ​​for stacker cranes, the path energy consumption weight adjustment values ​​for unmanned transport equipment, and the pre-controlled speed values ​​for cold chain fans are extracted from the pre-controlled parameter set. These are then encapsulated into control command frames for the stacker crane drive controller, the unmanned transport equipment navigation scheduler, and the cold chain fan frequency converter. Specifically, this involves: comprehensively analyzing the generated pre-controlled parameter set to extract precise control parameters corresponding to the three core equipment types: the optimized start-stop frequency value for the stacker crane (calculated as the number of times the stacker crane starts and stops per hour to reduce transient harmonic impact); the path energy consumption weight adjustment value for unmanned transport equipment (weight coefficients used to optimize path planning and avoid congestion); and the pre-controlled speed value for the cold chain fan (fan speed per minute used to avoid redundant operation). Following the dedicated communication protocols of the three types of equipment controllers, these parameters are encapsulated into corresponding control command frames for each equipment, ensuring that the command frames can be accurately recognized and executed by each equipment controller. The encapsulated content of each command frame is complete and specifically adapted to the equipment characteristics.

[0140] The instruction frame for the stacker crane drive controller includes the unique identifier of the stacker crane equipment, the optimized value of the stacker crane start and stop frequency, the start time of instruction execution, the execution cycle is set to 1 hour, which is consistent with the start and stop frequency statistical cycle, and the instruction check code is used to verify that the instruction transmission is error-free. The instruction check code is calculated by adding the equipment identifier value and the optimized value of the start and stop frequency, and then taking the last 4 digits.

[0141] Command frame for unmanned transport equipment navigation scheduler: The encapsulated content includes unmanned transport equipment group identifier, path energy consumption weight adjustment value, path planning update interval set to 5 minutes to ensure timely avoidance of congestion, and command check code. The check code is calculated by multiplying the equipment group identifier value by the weight adjustment value and then taking the last 4 digits.

[0142] The instruction frame for the variable frequency drive of the cold chain fan includes the cold chain fan number, the fan speed pre-adjustment value, the speed adjustment smoothness coefficient set to 0.1 to avoid sudden speed changes that could damage the equipment, the instruction execution start time, and the instruction check code. The check code is calculated by adding the fan number value to the speed pre-adjustment value, then dividing by 100 and taking the remainder.

[0143] After encapsulating the three types of instruction frames, each instruction frame is subjected to integrity verification to confirm that there are no missing parameters, no checksum errors, and no protocol mismatches, ensuring that the instructions can be accurately received and executed by each device controller after they are issued.

[0144] The optimized start-stop frequency value of the stacker crane is sent to the high-speed stacker crane drive controller to dynamically correct the start-stop timing and acceleration / deceleration curves of the stacker crane, reducing harmonic current impact and mechanical energy consumption. Specifically, this involves sending the encapsulated stacker crane start-stop frequency optimization value instruction frame to the high-speed stacker crane drive controller via an industrial real-time transmission link. After receiving the instruction, the drive controller automatically initiates the dynamic correction process. The specific implementation process is as follows:

[0145] Start-stop timing correction: First, extract the current original start-stop timing of the stacker crane, i.e., the time interval between two start-stop operations. Calculate the corrected start-stop time interval using the formula: Corrected start-stop time interval = 60 minutes ÷ Stacker crane start-stop frequency optimization value. For example, if the original start-stop frequency is 10 times / hour and the optimization value is 8 times / hour, then the corrected start-stop time interval = 60 ÷ 8 = 7.5 minutes. This ensures that the stacker crane starts and stops smoothly at the optimized frequency, reducing the number of frequent start-stop operations.

[0146] Acceleration / deceleration curve correction: Based on the optimized start / stop frequency value, the acceleration / deceleration curve parameters of the stacker crane are dynamically adjusted to avoid harmonic impacts caused by excessively rapid acceleration / deceleration. The specific correction method is as follows: Corrected acceleration time = original acceleration time × (1 + optimized start / stop frequency value ÷ original start / stop frequency), corrected deceleration time = original deceleration time × (1 + optimized start / stop frequency value ÷ original start / stop frequency). For example, if the original acceleration time is 2 seconds, the original start / stop frequency is 10 times / hour, and the optimized value is 8 times / hour, then the corrected acceleration time = 2 × (1 + 8 ÷ 10) = 3.6 seconds. By extending the acceleration / deceleration time, the rate of current change is reduced.

[0147] Energy consumption optimization effect verification: Real-time monitoring of the current change of the stacker crane drive motor, calculation of the corrected harmonic distortion rate, comparison of the difference between the harmonic distortion rate before and after correction, difference = original harmonic distortion rate - corrected harmonic distortion rate, ensuring that the reduction in harmonic distortion rate is not less than 20%; at the same time, the reduction in mechanical energy consumption is calculated, mechanical energy consumption reduction = original mechanical energy consumption × (original start-stop frequency - optimized start-stop frequency) ÷ original start-stop frequency, ensuring that mechanical energy consumption is reduced synchronously, solving the energy waste problem caused by the start-stop of the stacker crane.

[0148] The entire correction process is synchronized in real time to the dynamic digital twin, ensuring that the stacker crane's operating status in the digital twin is consistent with the physical entity.

[0149] The energy consumption weight adjustment value of the unmanned transport equipment path is sent to the unmanned transport equipment navigation scheduler to replan the travel path and intersection traffic priority, avoid congestion and empty detours, and reduce path energy consumption density. Specifically, this includes: sending the encapsulated unmanned transport equipment path energy consumption weight adjustment value instruction frame to the unmanned transport equipment navigation scheduler. After receiving the instruction, the navigation scheduler immediately starts the path replanning process. The specific implementation process is as follows:

[0150] Path energy consumption weight update: First, the original path energy consumption weight in the navigation scheduler is replaced with the adjusted value. The larger the weight value, the higher the energy cost of the path, and the greater the probability that it will be avoided first during path planning. At the same time, the path evaluation system is updated, increasing the proportion of path energy consumption weight to 60%, with the remaining 40% being parameters such as path distance and passage difficulty, to ensure that path planning is centered on energy consumption optimization.

[0151] Route replanning: Based on the updated path energy consumption weights, the navigation scheduler traverses all passable paths within the warehouse and calculates the comprehensive energy consumption cost of each path. The comprehensive energy consumption cost = path length × path energy consumption weight + passage difficulty coefficient. The coefficient for congested sections is 1.5, and the coefficient for unobstructed sections is 1.0. The path with the lowest comprehensive energy consumption cost is selected as the final path. When multiple devices plan paths simultaneously, priority is given to paths with higher energy consumption weights to avoid congestion at intersections.

[0152] Empty-load detour avoidance: Real-time collection of the load status of unmanned transport equipment. For empty equipment, the navigation scheduler plans the shortest energy-efficient return path based on the path energy consumption weight to avoid empty equipment detouring on long-distance paths. It calculates the reduction value of energy consumption per unit mileage of empty equipment. The reduction value = original empty unit mileage energy consumption × (original path energy consumption weight - adjusted path energy consumption weight) ÷ original path energy consumption weight, ensuring that the energy waste caused by empty detours is reduced.

[0153] Path execution monitoring: The navigation scheduler tracks the movement trajectory of the unmanned transport equipment in real time. If a sudden congestion is detected on a certain path, it immediately replans alternative paths based on the current path energy consumption weight to ensure that the equipment always moves along the low-energy path, ultimately reducing the path energy consumption density. The path energy consumption density reduction value = original path energy consumption density × (1 - adjusted path energy consumption weight ÷ original path energy consumption weight).

[0154] The pre-regulation value of the cold chain fan speed is sent to the cold chain fan inverter driver. The feedforward adjustment of the evaporator fan speed and start-stop cycle prevents the refrigeration unit from triggering full-load redundant operation due to minor temperature and humidity fluctuations. Specifically, the packaged cold chain fan speed pre-regulation value command frame is sent to the cold chain fan inverter driver. After receiving the command, the inverter driver starts the feedforward adjustment process to adjust the fan speed and start-stop cycle in advance. The specific implementation process is as follows:

[0155] Speed ​​feedforward adjustment: First, extract the current original speed of the cold chain fan, compare it with the speed pre-control value, and calculate the speed adjustment amount. Adjustment amount = original speed - speed pre-control value. If the adjustment amount is positive, the frequency converter gradually reduces the fan speed, and the adjustment rate is set to 50 rpm to avoid sudden speed changes that could damage the equipment. If the adjustment amount is negative, the speed is gradually increased to ensure a smooth transition to the pre-control value. At the same time, the temperature and humidity change rate at the return air vent of the cold chain storage area is collected in real time. If the temperature and humidity change rate exceeds the steady-state range, the speed is fine-tuned in advance. Fine-tuning amount = speed pre-control value × temperature and humidity change rate deviation ÷ steady-state deviation threshold, where the temperature and humidity change rate deviation = actual temperature and humidity change rate - steady-state temperature and humidity change rate. The steady-state deviation threshold is set to 0.2℃ / min and 1.0% / min.

[0156] Start-stop cycle correction: Based on the speed pre-regulation value, the start-stop cycle of the cold chain fan is corrected. The calculation formula is: Corrected start-stop cycle = Original start-stop cycle × (Original speed ÷ Speed ​​pre-regulation value). For example, if the original speed is 1500 rpm, the pre-regulation value is 1350 rpm, and the original start-stop cycle is 30 minutes, then the corrected start-stop cycle = 30 × (1500 ÷ 1350) ≈ 33.3 minutes. This extends the continuous operation time of the fan and avoids energy loss caused by frequent start-stop.

[0157] Redundancy avoidance: Real-time monitoring of the refrigeration unit's operating load. If the refrigeration unit load is below 70%, the inverter driver maintains the fan speed at the pre-controlled value. If the load is above 70%, the speed is finely adjusted to 1.1 times the pre-controlled value based on the temperature and humidity change rate. This ensures the cooling effect while avoiding full-load redundant operation of the fan. The fan energy consumption reduction value is calculated as follows: Energy consumption reduction value = Original fan energy consumption × (Original speed - Pre-controlled speed value) ÷ Original speed. This ensures a significant reduction in fan energy consumption while maintaining stable temperature and humidity in the cold chain storage area.

[0158] The system collects the actual current harmonic sequence after the stacker crane drive controller executes, the actual energy consumption per unit mileage after the unmanned transport equipment navigation scheduler executes, and the actual temperature and humidity fluctuation sequence after the cold chain fan frequency converter driver executes. The collected data is then compared with the corresponding original data streams to calculate the deviation, yielding equipment operation feedback data. Specifically, this includes: establishing a dedicated feedback data acquisition link, running in parallel with the established real-time data transmission link, maintaining the same acquisition frequency as the original data acquisition frequency: 1000 times per second for transient current harmonics, 100 times per second for path and energy consumption, and 10 times per second for temperature and humidity. The collected content includes: the actual current harmonic sequence after the stacker crane drive controller executes: the three-phase current time-domain waveform and harmonic distortion rate characteristic value at each moment; the actual energy consumption per unit mileage after the unmanned transport equipment navigation scheduler executes: the equipment position coordinates, mileage, energy consumption, and energy consumption per unit mileage at each moment; and the actual temperature and humidity fluctuation sequence after the cold chain fan frequency converter driver executes: the return air inlet temperature, humidity, temperature and humidity change rate at each moment, and the actual fan speed.

[0159] Deviation Calculation: The deviation between the acquired data and the corresponding raw data stream of the three types of devices is calculated time-by-time to ensure accurate deviation calculation. The specific calculation method is as follows:

[0160] Stacker crane harmonic deviation calculation: Harmonic deviation value at each moment = Original harmonic distortion rate - Actual harmonic distortion rate. Calculate the average harmonic deviation value within the simulation period. Average deviation value = Arithmetic mean of harmonic deviation values ​​at all moments within the simulation period. Simultaneously calculate the harmonic suppression rate. Harmonic suppression rate = Average harmonic deviation value ÷ Original average harmonic distortion rate × 100%.

[0161] Energy consumption deviation calculation for unmanned transport equipment: Energy consumption deviation per unit mileage at each moment = Original energy consumption per unit mileage - Actual energy consumption per unit mileage. Calculate the average energy consumption deviation during the simulation period. Average deviation = Arithmetic mean of energy consumption deviations at all moments during the simulation period. Simultaneously calculate the energy consumption reduction rate. Energy consumption reduction rate = Average energy consumption deviation ÷ Original average energy consumption per unit mileage × 100%.

[0162] Cold chain fan temperature and humidity deviation calculation: Temperature deviation value at each moment = Actual temperature - Original temperature, Humidity deviation value at each moment = Actual humidity - Original humidity. Calculate the average temperature and humidity deviation value within the simulation period, and simultaneously calculate the temperature and humidity fluctuation control rate: Control rate = (Original temperature and humidity fluctuation amplitude - Actual temperature and humidity fluctuation amplitude) ÷ Original temperature and humidity fluctuation amplitude × 100%;

[0163] Feedback data integration: The actual operation data, time-by-time deviation value, average deviation value, inhibition rate / reduction rate / control rate and other parameters of the three types of equipment are aligned by timestamp and integrated into unified equipment operation feedback data.

[0164] In this embodiment of the invention, the optimized start-stop frequency value of the stacker crane, the path energy consumption weight adjustment value of the unmanned transport equipment, and the pre-controlled speed value of the cold chain fan are parsed from the pre-controlled parameter set and encapsulated into dedicated control instruction frames adapted to each equipment controller. This ensures that the control instructions are accurately matched with the control requirements of each equipment, avoiding the drawbacks of poor adaptability and insufficient targeting of traditional general control instructions. By sending the optimized start-stop frequency value of the stacker crane to the drive controller, the start-stop timing and acceleration / deceleration curves are dynamically corrected, reducing harmonic current impact and mechanical energy consumption. By sending the path energy consumption weight adjustment value of the unmanned transport equipment, the navigation scheduler is prompted to replan the travel path and intersection traffic priority, avoiding equipment clustering and congestion and empty detours, thus reducing path energy consumption density. By sending the pre-controlled speed value of the cold chain fan, the evaporator fan speed and start-stop cycle are adjusted forward, avoiding full-load redundant operation of the refrigeration unit triggered by small temperature and humidity fluctuations, thus reducing ineffective energy consumption. By collecting the actual operating data of each equipment after control and calculating the deviation with the original data stream, equipment operation feedback data is obtained.

[0165] In a preferred embodiment of the present invention, based on the equipment operation feedback data after dynamic adaptive control processing, closed-loop optimization iterative processing is performed to form continuous optimization of operating condition prediction and pre-control parameters, including:

[0166] This study extracts three types of deviation sequences from equipment operation feedback data: a harmonic suppression deviation sequence between the actual current harmonic sequence of the stacker crane and the original harmonic distortion rate; a path energy consumption deviation sequence between the actual unit mileage energy consumption of the unmanned transport equipment and the original unit mileage energy consumption; and a temperature control response deviation sequence between the actual temperature and humidity fluctuation of the cold chain fan and the original temperature and humidity change rate. These three types of deviation sequences are then aligned by timestamps to construct a closed-loop deviation vector set. Specifically, this involves: clarifying the extraction range and calculation logic of the three types of deviation sequences to ensure that the deviation data truly reflects the control effect; selecting comparative data before and after control for the three core equipment types—stack crane, unmanned transport equipment, and cold chain fan—from the equipment operation feedback data; and extracting the corresponding deviation sequences for each: Firstly, the harmonic suppression deviation sequence, focusing on the stacker crane's current harmonic control effect. Using the original harmonic distortion rate before control as a benchmark, the actual current harmonic sequence collected after control is compared. The deviation value at each sampling time is the original harmonic distortion rate minus the actual current harmonic distortion rate at the corresponding time. The system employs three main methods: First, it presents a sequence of deviations. First, it lists the deviations at each sampling point in chronological order, forming a complete harmonic suppression deviation sequence that directly reflects the actual harmonic suppression effect of the stacker crane. Second, it presents a path energy consumption deviation sequence, focusing on the energy consumption optimization effect of unmanned transport equipment. Using the original energy consumption per unit mileage before adjustment as a benchmark, it compares the actual energy consumption per unit mileage after adjustment. The deviation value at each sampling point is calculated as the original energy consumption per unit mileage minus the actual energy consumption per unit mileage at that corresponding moment. Arranging all deviation values ​​in chronological order forms the path energy consumption deviation sequence, accurately quantifying the energy consumption reduction after path optimization for unmanned transport equipment. Third, it presents a temperature control response deviation sequence, focusing on the temperature and humidity control effect of the cold chain fan. Using the original temperature and humidity change rate before adjustment as a benchmark, it compares the actual temperature and humidity fluctuation data after adjustment, calculating the deviation value at each sampling point, i.e., the original temperature and humidity change rate minus the actual temperature and humidity change rate. Arranging these values ​​in chronological order forms the temperature control response deviation sequence, reflecting the control effect of cold chain fan speed regulation on temperature and humidity fluctuations.

[0167] Timestamp alignment processing was carried out for three types of deviation sequences. Due to the different data acquisition frequencies of the three types of equipment—1000 times per second for the transient current harmonics of the stacker crane, 100 times per second for the path and energy consumption of the unmanned handling equipment, and 10 times per second for the temperature and humidity of the cold chain fan—the timestamp densities of the three types of deviation sequences were inconsistent, making direct combination impossible. Therefore, using the timestamp of the stacker crane harmonic data (the highest acquisition frequency) as a benchmark, interpolation was performed on the other two types of deviation sequences: For the path energy consumption deviation sequence, nine deviation values ​​were added between two adjacent original timestamps using linear interpolation, ensuring that each timestamp in the completed path energy consumption deviation sequence corresponds one-to-one with the timestamp of the stacker crane harmonic data; for the temperature control response deviation sequence, 99 deviation values ​​were added between two adjacent original timestamps using linear interpolation, similarly achieving complete matching with the timestamp of the stacker crane harmonic data. During the interpolation process, the principle of linear transition between adjacent deviation values ​​was strictly followed to ensure that the completed deviation values ​​closely match the actual control trend without distortion.

[0168] The three types of deviation sequences after timestamp alignment are combined according to the same timestamp. Each timestamp corresponds to a set of three-dimensional deviation data, namely the harmonic suppression deviation value, path energy consumption deviation value, and temperature control response deviation value at that moment. These three deviation values ​​are combined to form a deviation vector. This vector fully contains the control deviation of the three types of equipment at the same moment. The deviation vectors corresponding to all timestamps are integrated one by one in chronological order to form a closed-loop deviation vector set. This vector set system quantifies the deviation between the control effect of the three types of equipment and the original operating conditions, and clearly presents the advantages and disadvantages in the control process.

[0169] The closed-loop deviation vector set is reverse-mapped to the mesh topology of the dynamically parameterized spatial manifold. Using the deviation vector amplitude as a weighting coefficient, the non-uniform scaling factor applied along the three principal directions of the base surface is corrected to obtain the updated dynamically parameterized spatial manifold. Specifically, this involves completing the reverse mapping of the closed-loop deviation vector set to the mesh topology. The mesh topology of the dynamically parameterized spatial manifold completely corresponds to the three-dimensional physical space of the warehouse. Each mesh node corresponds to a specific physical location within the warehouse and is bound to the energy consumption characteristics and spatial coordinates of the equipment at that location. The core of the reverse mapping is to accurately match the deviation information in the closed-loop deviation vector set to the corresponding mesh node. Based on the spatial coordinates of the mesh node and its affiliation with the equipment, the deviation amplitude corresponding to the timestamp is assigned to that node. For example, mesh nodes in the area where the stacker crane is located are mainly bound to the amplitude corresponding to harmonic suppression deviation; mesh nodes along the path of unmanned handling equipment are mainly bound to the amplitude corresponding to path energy consumption deviation; and mesh nodes in the area covered by cold chain fans are mainly bound to the amplitude corresponding to temperature control response deviation.

[0170] Using the magnitude of the deviation vector as weight, the non-uniform scaling factor is corrected. The base surface of the dynamically parameterized spatial manifold has three mutually orthogonal principal directions, corresponding to the horizontal, vertical, and latitudinal directions of the warehouse, respectively. Initially, the non-uniform scaling factor in all three principal directions is set to 1, representing the spatial scaling ratio without adjustment deviation. During the correction process, the average deviation magnitude of all grid nodes in each principal direction is first calculated. Then, according to the equipment distribution characteristics of each principal direction, corresponding directional weights are set: stacker cranes are mainly distributed in the horizontal direction, with a weight of 0.5; unmanned handling equipment is mainly distributed in the horizontal direction, with a weight of 0.4; and cold chain fans are mainly distributed in the vertical direction, with a weight of 0.1. The corrected non-uniform scaling factor is then calculated. , This is the scaling factor after correction in a certain direction. Let be the initial scaling factor in a certain direction. The direction weight is a certain direction. The average deviation amplitude in a certain direction is considered. If the average deviation amplitude in a certain direction is large, it indicates that the control deviation in that direction is large. The corrected scaling factor will be greater than 1, so that the spatial manifold is adaptively stretched in that direction, thereby improving the representation accuracy of that region. If the average deviation amplitude is small, it indicates that the control effect is good. The corrected scaling factor will be close to 1, maintaining the spatial representation ratio of that region.

[0171] The modified non-uniform scaling factors of the three principal directions are substituted into the generation process of the dynamically parameterized spatial manifold to update the grid node coordinates of the original manifold. The updated node coordinates are equal to the original node coordinates multiplied by the modified scaling factors of the corresponding directions. Through this update, the spatial manifold will adaptively stretch towards the region with large control deviation and adaptively shrink towards the region with good control effect and small deviation, so that the grid topology of the manifold is highly consistent with the spatiotemporal distribution of energy consumption in the entire storage area after actual control.

[0172] Based on the updated dynamic parameterized spatial manifold, the threshold setting value of the initial boundary profile is recalculated, and the offset step size along the normal direction is adjusted using the statistical characteristics of the closed-loop deviation vector set as the offset. Specifically, this includes: recalculating the threshold setting value of the initial boundary profile. The threshold values ​​of the initial boundary profile include three types: harmonic distortion rate threshold, peak energy consumption per unit mileage threshold, and extreme value threshold of temperature and humidity change rate. These thresholds are the core basis for delineating abnormal energy consumption boundaries. Traditional thresholds are fixed and cannot adapt to changes in operating conditions after regulation. During recalculation, the latest energy consumption data on the manifold grid nodes is extracted based on the updated dynamic parameterized spatial manifold, and the mean characteristics of the closed-loop deviation vector set are combined for correction: for the harmonic distortion rate threshold, the new threshold is equal to the original threshold multiplied by (1). The new threshold is calculated by subtracting the mean of the harmonic suppression deviation sequence and dividing it by the mean of the original harmonic distortion rate. For the peak energy consumption threshold per unit mileage, the new threshold is equal to the original threshold multiplied by (1 minus the mean of the path energy consumption deviation sequence divided by the mean of the original energy consumption per unit mileage). For the extreme threshold of temperature and humidity change rate, the new threshold is equal to the original threshold multiplied by (1 minus the mean of the temperature control response deviation sequence divided by the mean of the original temperature and humidity change rate). If the mean of the deviation sequence is positive, it indicates that the energy consumption has decreased after regulation, and the corresponding threshold will be appropriately lowered to make the judgment standard for abnormal energy consumption more stringent and accurate, avoiding misjudging normal energy consumption as abnormal. If the mean of the deviation sequence is negative, it indicates that the regulation effect is not good, and the energy consumption has not decreased or even increased. The corresponding threshold will be appropriately raised to avoid misjudging normal energy consumption fluctuations as abnormal, ensuring that the threshold setting is compatible with the actual regulation effect.

[0173] The offset step size is a key parameter for generating multi-layered nested buffer boundaries. Traditional fixed step sizes cannot adapt to changes in the energy consumption disturbance range after regulation, and need to be adjusted based on the statistical characteristics of the closed-loop deviation vector set. Calculating the statistical characteristics of the closed-loop deviation vector set focuses on calculating the mean and standard deviation of all deviation vector amplitudes. The standard deviation reflects the degree of fluctuation in the deviation values; the greater the fluctuation, the more unstable the regulation effect and the greater the change in the energy consumption disturbance range. Subsequently, the offset step size is adjusted using the ratio of the standard deviation to the mean, i.e., the deviation fluctuation coefficient, as the offset amount. The adjusted offset step size is equal to the original offset step size multiplied by (1 plus the deviation fluctuation coefficient). If the deviation fluctuation coefficient is large, it indicates that the regulation effect is unstable and the energy consumption disturbance range changes significantly. The adjusted offset step size will be appropriately increased to expand the coverage of the multi-layer buffer boundary and ensure that the gradient changes of energy consumption disturbance can be fully captured. If the deviation fluctuation coefficient is small, it indicates that the control effect is stable and the energy consumption disturbance range does not change much. The adjusted offset step size will be appropriately decreased to reduce the coverage of the buffer boundary and improve the accuracy of abnormal partitioning. The original offset step size is the side length of the smallest grid unit in the physical space of the warehouse. The adjusted step size still maintains the adaptability to the side length of the grid unit to ensure that the generation of multi-layer nested buffer boundaries conforms to the layout of the warehouse space. By recalculating the threshold and adjusting the offset step size, the initial boundary outline is divided more accurately, and the gradient distribution of the multi-layer nested buffer boundary is more in line with the actual energy consumption disturbance characteristics after control.

[0174] The corrected non-uniform scaling factor, the adjusted threshold setting, and the updated offset step size are substituted into the working condition prediction and anomaly simulation process, so that the simulation results and pre-control parameter set of the next iteration are adaptively calibrated based on the actual control feedback. Specifically, all corrected parameters are fully substituted into the working condition prediction and anomaly simulation process to ensure that the simulation model is highly adapted to the actual controlled working conditions. The modified non-uniform scaling factor is substituted into the construction stage of the dynamically parameterized spatial manifold, replacing the original initial scaling factor. This ensures that the newly constructed spatial manifold accurately reflects the spatiotemporal distribution of energy consumption across the entire storage area after regulation. The adjusted initial boundary contour threshold is substituted into the division stage of the core disturbance zone, transition influence zone, and peripheral stable zone, replacing the original fixed threshold. This makes the boundary division of the three regions more consistent with the actual energy consumption anomaly characteristics, avoiding missed or misjudged cases. The updated offset step size is substituted into the generation stage of the multi-layer nested buffer boundary, replacing the original fixed step size. This makes the gradient distribution of the buffer boundary more suitable for the range of energy consumption disturbance after regulation, ensuring that the propagation attenuation law and delayed response characteristics of energy consumption impact can be accurately simulated. During the substitution process, the compatibility of each modified parameter with the corresponding stage is strictly ensured, and the model running status after parameter substitution is verified one by one to avoid parameter mismatch and abnormal model running.

[0175] Based on the spatial manifold, boundary delineation, and buffer boundary after incorporating the corrected parameters, the operating condition prediction and anomaly simulation were re-conducted, focusing on calibrating core simulation results such as energy consumption surge peaks, affected areas, lag times, and anomaly overflow points. During the calibration process, the simulation results were corrected based on the average deviation amplitude of the closed-loop deviation vector set: for the energy consumption surge peak in the core disturbance zone, the calibrated peak value was equal to the original simulation peak value multiplied by (1 minus the average deviation amplitude divided by the maximum deviation amplitude), making the simulation peak value closer to the actual energy consumption level after control, and avoiding excessive deviation between the simulation and actual values; for the energy consumption affected area and lag time in the transitional influence zone, calibration was performed in conjunction with the deviation fluctuation coefficient. When the fluctuation coefficient was large, the affected area was appropriately expanded and the lag time was extended; when the fluctuation coefficient was small, the affected area was appropriately reduced and the lag time was shortened; for the anomaly overflow points in the peripheral stable zone, the number and spatial location of the anomaly points were calibrated in conjunction with the distribution of temperature control response deviation and path energy consumption deviation, ensuring that the anomaly overflow situation in the simulation was consistent with the actual control feedback.

[0176] Based on the calibrated operating condition prediction results, the optimized values ​​for stacker crane start-stop frequency, the adjusted values ​​for path energy consumption weights of unmanned transport equipment, and the pre-regulation values ​​for cold chain fan speed are recalculated to form a calibrated set of pre-regulation parameters. The calibration logic revolves around the deviation of the control feedback: if the mean of the harmonic suppression deviation sequence is positive, it indicates that the stacker crane has a good harmonic suppression effect, and the calibrated optimized start-stop frequency value is appropriately reduced from the original optimized value to further enhance the harmonic suppression effect; if the mean of the deviation is negative, it indicates that the harmonic suppression effect is poor, and the calibrated optimized start-stop frequency value is appropriately increased to avoid over-regulation leading to abnormal equipment operation. For the adjusted values ​​for path energy consumption weights of unmanned transport equipment, if the mean of the path energy consumption deviation sequence is positive, it indicates that the path optimization effect is good, and the calibrated weight value is appropriately increased to further avoid congestion and empty detours; if the mean of the deviation is negative, it indicates that the path optimization effect is poor, and the calibrated weight value is appropriately decreased to optimize the rationality of path planning. For the pre-adjustment value of the cold chain fan speed, if the mean of the temperature control response deviation sequence is positive, it indicates that the temperature and humidity control effect is good. The speed after calibration should be appropriately reduced to further reduce energy consumption. If the mean of the deviation is negative, it indicates that the temperature and humidity control effect is poor. The speed after calibration should be appropriately increased to ensure the temperature and humidity stability of the cold chain storage area.

[0177] In this embodiment of the invention, by accurately extracting three types of time-series sequences from equipment operation feedback data—stacking crane harmonic suppression deviation, unmanned handling equipment path energy consumption deviation, and cold chain fan temperature control response deviation—and aligning them by timestamp to construct a closed-loop deviation vector set, unified quantitative collection and precise time-series matching of multi-device multi-dimensional control errors are achieved. By inversely mapping the closed-loop deviation vector set to a dynamically parameterized spatial manifold mesh topology, and using the deviation vector amplitude as weight to correct the non-uniform scaling factor of the three principal directions of the base surface and update the spatial manifold, the limitations of traditional static digital twin modeling parameters being fixed and unable to adapt to the dynamic drift of warehousing conditions are overcome, enabling full-domain... The spatial representation of energy consumption closely matches the actual distribution of control errors. By recalculating the initial boundary contour threshold based on the updated spatial manifold and dynamically adjusting the contour normal offset step size in combination with the deviation statistical characteristics, the rules for judging abnormal energy consumption boundaries are optimized, avoiding the problems of distortion, omission, and misjudgment caused by fixed thresholds and step sizes. By substituting all the correction parameters into the working condition prediction and abnormal simulation process to achieve adaptive calibration iteration, a long-term closed loop of physical control feedback driving digital model self-updating is formed. This overcomes the shortcomings of traditional energy consumption prediction in terms of long-term accuracy decay and the year-by-year decline in the adaptability of control parameters, and continuously improves the accuracy of transient energy consumption simulation and the adaptability of dynamic control.

[0178] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0179] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0180] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An energy consumption monitoring and optimization system based on digital twins in intelligent warehousing, characterized in that, include: The acquisition module is used to acquire multi-source high-frequency dynamic data in real time, and to perform real-time mapping and dynamic synchronization of static 3D visualization models to obtain dynamic digital twins; The module is used to construct a dynamic parameterized spatial manifold based on a dynamic digital twin, according to the spatial distribution of transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensor acquisition points and their respective data update frequencies. The module then maps multi-source high-frequency dynamic data to discrete nodes of the spatial manifold according to timestamps and spatial coordinates to obtain a synchronous running state flow. The computation module is used to perform isoparametric transformation on the dynamically parameterized spatial representation based on the synchronous running state flow to obtain the transformed projection domain; extract the boundary contours corresponding to each acquisition point on the transformed projection domain, and obtain multi-layer nested buffer boundaries by successively offsetting along the normal direction; divide the spatial representation into a core disturbance zone, a transition influence zone, and a peripheral stable zone according to each layer of buffer boundaries, and perform numerical simulation of the transient energy consumption surge trend in each zone to obtain the pre-simulation results of the abnormal energy consumption surge condition and the corresponding pre-control parameter set; The processing module is used to perform dynamic adaptive control processing based on the pre-simulation results and the pre-control parameter set, and obtain the equipment operation feedback data after dynamic adaptive control processing. The iterative module is used to perform closed-loop optimization iterative processing based on the equipment operation feedback data after dynamic adaptive control, so as to continuously optimize the operating condition prediction and pre-control parameters.

2. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 1, characterized in that, Real-time acquisition of multi-source, high-frequency dynamic data; real-time mapping and dynamic synchronization of static 3D visualization models to obtain dynamic digital twins; real-time capture of raw transaction data streams, including: Transient current harmonic acquisition points are deployed at the winding end of the drive motor of the high-speed stacker crane, real-time path tracking and energy consumption mapping acquisition points are deployed at the navigation and positioning point of the unmanned handling equipment, and temperature and humidity field dynamic response sensing acquisition points are deployed at the return air vent of the evaporator in the cold chain warehouse. The sampling frequency and data reporting protocol of each acquisition point are configured, and a real-time data transmission link with the digital twin platform is established. Based on the real-time data transmission link, each acquisition point captures the three-phase current time-domain waveform of the stacker crane drive motor and extracts the harmonic distortion rate characteristic value, the real-time position coordinates and energy consumption per unit mileage of the unmanned handling equipment, and the temperature and humidity change rate of the return air vent, forming a multi-source high-frequency dynamic data stream. Add timestamps and spatial coordinate labels to the multi-source high-frequency dynamic data streams. Based on the preset anchor point positions of each collection point in the static 3D visualization model, map the timestamp-aligned data streams to the corresponding component nodes of the static 3D visualization model according to the spatial coordinates to obtain the mapping results. Based on the mapping results, the geometric structure, spatial pose, and operational attribute parameters of the stacker crane, unmanned handling equipment, and cold chain fan in the static 3D visualization model are dynamically refreshed, so that the representation content of the static 3D visualization model is synchronized with the physical entity, forming a dynamic digital twin.

3. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 2, characterized in that, Based on a dynamic digital twin, a dynamically parameterized spatial manifold is constructed according to the spatial distribution and data update frequency of transient current harmonic acquisition points, real-time path tracing and energy consumption mapping acquisition points, and temperature and humidity field dynamic response sensing acquisition points. This includes: Spatial coordinate values ​​of each acquisition point are extracted from the dynamic digital twin, and harmonic distortion rate characteristic value sequence, location coordinate and unit mileage energy consumption sequence, and temperature and humidity change rate sequence are associated with the data stream type corresponding to each acquisition point to form the spatiotemporal attribute vector of each acquisition point. Using the spatial coordinates of each collection point as the node position and the data update frequency as the time axis sampling interval, a discrete node set is established. Based on the spatiotemporal attribute vectors associated with each node, a continuous attribute field covering the storage space is constructed. The isoparametric surface of the continuous attribute field is defined as the base surface of the dynamically parameterized spatial manifold. On the base surface, based on the direction of harmonic distortion rate gradient change of transient current harmonic acquisition points, the density of motion trajectories of real-time path tracking and energy consumption mapping acquisition points, and the distribution of temperature and humidity field equipotential lines of temperature and humidity field dynamic response sensor acquisition points, a non-uniform scaling factor is applied along the three principal directions of the base surface to obtain a dynamically parameterized spatial manifold. The dynamically parameterized spatial manifold is discretized into a mesh topology, and the corresponding spatiotemporal attribute vectors are retained for each vertex, thus completing the construction of the dynamically parameterized spatial manifold.

4. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 3, characterized in that, Multi-source high-frequency dynamic data is mapped to discrete nodes of a spatial manifold according to timestamps and spatial coordinates to obtain a synchronized running state stream, including: The three-dimensional spatial coordinates of each discrete node are extracted from the grid topology. Based on the spatiotemporal attribute vector type of each node, the spatial affiliation between the node and the transient current harmonic acquisition point, the real-time path tracing and energy consumption mapping acquisition point, and the temperature and humidity field dynamic response sensing acquisition point are established to determine the dominant acquisition point affiliation of each discrete node in space. Based on three parallel multi-source high-frequency dynamic data streams, and using the set data update frequency of each acquisition point as a benchmark, the data streams are aligned by a sliding window on the time axis to obtain a synchronized data frame sequence with a unified time benchmark. For each synchronized data frame, based on the determined spatial affiliation between nodes and acquisition points, the harmonic distortion rate characteristic value, location coordinates, energy consumption per unit mileage, and temperature and humidity change rate corresponding to each acquisition point are assigned to each discrete node in the spatial region, so that each discrete node obtains a fusion attribute value that matches its spatial location. The assigned discrete node fusion attribute values ​​are then fused with the retained spatiotemporal attribute vector using Kalman filtering to obtain the real-time state vector of all discrete nodes. The real-time state vectors of all discrete nodes are arranged in time sequence according to the obtained synchronous data frame sequence and encapsulated into an iteratively updatable synchronous running state stream.

5. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 4, characterized in that, Based on the synchronous running state flow, an isoparametric transformation is performed on the dynamically parameterized spatial representation to obtain the transformed projection domain. The boundary contours corresponding to each acquisition point are extracted from the transformed projection domain, and multi-layered nested buffer boundaries are obtained by successively offsetting along the normal direction, including: Extract the mesh topology of the dynamic parameterized spatial representation and the real-time state vectors on each discrete node from the synchronous running state flow, and construct the mapping relationship from the three-dimensional physical space to the three-dimensional attribute space using the coordinates of each node and the real-time state vectors. By applying an isoparametric transformation to the mapping relationship, the dynamic parameterized spatial representation is mapped to the parameterized projection domain in the three-dimensional attribute space, so that the spatial position and the attribute field quantity form a one-to-one correspondence of parameter coordinates, thus completing the parameterized dimensionality reduction expression. In the parametric projection domain, with the parameter coordinates corresponding to each acquisition point as the center, the threshold is set according to the characteristic value of harmonic distortion rate, the peak energy consumption per unit mileage and the extreme value of temperature and humidity change rate captured by each acquisition point, and the contour lines are extracted as the initial boundary profile. The initial boundary contour is inversely mapped back to the three-dimensional physical space to obtain the corresponding three-dimensional boundary contour lines. Each contour line is then discretized to obtain a polygonal boundary composed of an ordered sequence of points. For each polygon boundary, the normal direction of each side segment is calculated, and the extended boundary is obtained successively outward along the normal direction with a preset offset step size. The original boundary and the extended boundary are numbered hierarchically according to the offset distance to obtain a multi-layer nested buffer boundary sequence.

6. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 5, characterized in that, Based on the buffer boundaries of each layer, the spatial representation is divided into a core disturbance region, a transitional influence region, and a peripheral stable region. Numerical simulations of transient energy consumption surge trends are performed in each region to obtain the pre-simulation results of abnormal energy consumption surge conditions and the corresponding pre-regulation parameter set, including: The innermost boundary is extracted from the multi-layer buffer boundary sequence as the core perturbation boundary, and the region inside this boundary is marked as the core perturbation region; the middle boundary is extracted as the transition influence boundary, and the region between the core perturbation boundary and the transition influence boundary is marked as the transition influence region; the region outside the outermost boundary is marked as the peripheral stable region. For the core disturbance area, the real-time state vector of discrete nodes in the area is extracted to simulate the instantaneous impact effect of stacker crane start-up and shutdown, equipment path intersection and fan load fluctuation on local energy consumption, and obtain the peak value and duration of energy consumption surge. For the transitional impact zone, the state vectors of discrete nodes within the zone are extracted, and the attenuation law and delayed response characteristics of energy consumption impact propagating outward along the grid topology are simulated to obtain the predicted results of energy consumption impact range and lag time. For the outer stable region, the steady-state operating parameters of discrete nodes within the region are extracted. The simulation results of the core disturbance region and the transition influence region are compared with the baseline to identify abnormal points exceeding the threshold and obtain the abnormal overflow pre-simulation results. Based on the results of the pre-simulation in various regions, the optimized values ​​of stacker crane start-stop frequency, the adjustment values ​​of unmanned handling equipment path energy consumption weight, and the pre-control values ​​of cold chain fan speed are solved to obtain the pre-control parameter set.

7. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 6, characterized in that, Based on the pre-simulation results and the pre-control parameter set, dynamic adaptive control processing is performed to obtain equipment operation feedback data after dynamic adaptive control processing, including: The optimized values ​​of stacker crane start-stop frequency, the adjustment values ​​of unmanned transport equipment path energy consumption weight, and the pre-control values ​​of cold chain fan speed are extracted from the pre-control parameters and encapsulated into control instruction frames for stacker crane drive controller, unmanned transport equipment navigation scheduler, and cold chain fan frequency converter driver, respectively. The optimized values ​​of the stacker crane start-stop frequency are sent to the high-speed stacker crane drive controller to dynamically correct the stacker crane start-stop timing and acceleration / deceleration curves, thereby reducing harmonic current impact and mechanical energy consumption. The path energy consumption weight adjustment value of the unmanned transport equipment is sent to the navigation scheduler of the unmanned transport equipment to re-plan the travel path and intersection traffic priority, avoid congestion and empty detours, and reduce the path energy consumption density. The pre-regulation value of the cold chain fan speed is sent to the cold chain fan frequency converter, and the feedforward adjustment of the evaporator fan speed and start-stop cycle prevents the refrigeration unit from triggering full-load redundant operation due to slight temperature and humidity fluctuations. The actual current harmonic sequence after the stacker crane drive controller is executed, the actual unit mileage energy consumption sequence after the unmanned handling equipment navigation scheduler is executed, and the actual temperature and humidity fluctuation sequence after the cold chain fan frequency converter is executed are collected. The collected data is compared with the corresponding raw data stream to calculate the deviation and obtain the equipment operation feedback data.

8. The energy consumption monitoring and optimization system based on digital twin in intelligent warehousing according to claim 7, characterized in that, Based on the equipment operation feedback data processed by dynamic adaptive control, closed-loop optimization iterative processing is performed to continuously optimize the operating condition prediction and pre-control parameters, including: The harmonic suppression deviation sequence between the actual current harmonic sequence of the stacker crane and the original harmonic distortion rate, the path energy consumption deviation sequence between the actual unit mileage energy consumption of the unmanned handling equipment and the original unit mileage energy consumption, and the temperature control response deviation sequence between the actual temperature and humidity fluctuation of the cold chain fan and the original temperature and humidity change rate are extracted from the equipment operation feedback data. The three types of deviation sequences are aligned by timestamp to construct a closed-loop deviation vector set. The closed-loop deviation vector set is reverse-mapped into the mesh topology of the dynamically parameterized spatial manifold. The deviation vector magnitude is used as a weighting coefficient to correct the non-uniform scaling factor applied along the three principal directions of the base surface, resulting in the updated dynamically parameterized spatial manifold. Based on the updated dynamic parameterized spatial manifold, the threshold setting of the initial boundary profile is recalculated, and the offset step size of the successive offsets along the normal direction is adjusted using the statistical characteristics of the closed-loop deviation vector set as the offset. The corrected non-uniform scaling factor, the adjusted threshold setting, and the updated offset step size are substituted into the working condition prediction and anomaly simulation processing, so that the simulation results and pre-control parameter set of the next iteration are adaptively calibrated based on the actual control feedback.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the system as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, performs the system as described in any one of claims 1 to 8.

Images