An electrolytic aluminum load distributed cooperative control system based on edge cluster management

Abstract

This invention relates to the field of power distribution control technology and provides a distributed collaborative control system for electrolytic aluminum loads based on edge cluster management. The system includes: an edge computing cluster that collects local electrolytic aluminum cell operating parameters and network status in real time, constructs a dynamic digital model, and stores it locally; when communication is normal, the edge computing cluster receives a global scheduling strategy from the cloud and generates power adjustment commands based on the dynamic digital model to execute a first adjustment action; when communication is interrupted, the edge computing cluster automatically enters autonomous mode and independently generates power adjustment commands based on a locally pre-stored renewable energy output prediction model and the dynamic digital model to execute a second adjustment action; when communication is restored, the edge computing cluster compresses the operating data and adjustment command records stored during the interruption and uploads them to the cloud to complete data synchronization and strategy updates. This effectively improves the reliability of grid power balance response and renewable energy absorption efficiency in scenarios with a high proportion of renewable energy integration.

Description

Technical Field

[0001] This invention relates to the field of power distribution control technology, and in particular to a distributed collaborative control system for electrolytic aluminum loads based on edge cluster management. Background Technology

[0002] With a high proportion of new energy sources being integrated into regional power grids, electrolytic aluminum loads, due to their large power capacity, controllable thermal inertia, and rapid adjustment response, are gradually being regarded as an important flexible load resource for participating in power grid balance regulation. In the existing electrolytic aluminum load regulation system, a cloud-based dispatch center deployed on the power grid dispatch side typically collects the operating parameters of each electrolytic aluminum plant and the output data of new energy power plants. Based on the overall grid load demand and new energy output forecasts, a unified power allocation command is generated and directly sent to the load control terminals of each plant through communication channels such as optical fibers. The load control terminals adjust the on-load tap changers or rectifier transformer outputs of the electrolytic aluminum cells according to the received commands, thereby changing the DC power consumption of the electrolytic cells and achieving the following adjustment to power grid fluctuations.

[0003] In renewable energy grids covering remote areas, communication networks are prone to interruptions or latency fluctuations due to factors such as severe weather and equipment failures. When communication networks are interrupted, the load control terminals on the plant side cannot receive dispatch instructions from the cloud and can only maintain the current power status or execute pre-set fixed adjustment curves. They cannot proactively respond to the instantaneous fluctuations in local renewable energy output, leading to increased wind and solar curtailment and reduced renewable energy absorption rates. Summary of the Invention

[0004] This invention provides a distributed collaborative control system for electrolytic aluminum load based on edge cluster management, which is used to solve the problem of being unable to actively respond to the instantaneous fluctuations in local renewable energy output, resulting in increased wind and solar curtailment and reduced renewable energy consumption rate.

[0005] This invention provides a distributed collaborative control system for electrolytic aluminum load based on edge cluster management, comprising: a cloud-based dispatch center, edge computing clusters deployed in various electrolytic aluminum plant areas, and several load control terminals set up within the plant areas; each load control terminal is used to adjust the power consumption of the corresponding electrolytic aluminum cell and is communicatively connected to the edge computing cluster in its plant area; the edge computing cluster is communicatively connected to the cloud-based dispatch center.

[0006] The electrolytic aluminum load distributed collaborative control system based on edge cluster management is used to execute the following methods:

[0007] The edge computing cluster collects the operating parameters and local network status of the local electrolytic aluminum cell in real time, and constructs a dynamic digital model characterizing the real-time load adjustment capability of the plant area based on the operating parameters and local network status, and stores the dynamic digital model in the local storage module of the edge computing cluster.

[0008] When the communication connection between the edge computing cluster and the cloud scheduling center is in a normal state, the edge computing cluster receives the global scheduling strategy for the power allocation of the electrolytic aluminum load issued by the cloud scheduling center, and generates a power adjustment command in combination with the dynamic digital model to control the load control terminal to perform the first adjustment action.

[0009] When the edge computing cluster detects that the communication connection status between it and the cloud dispatch center changes from the normal state to the interrupted state, the edge computing cluster enters autonomous mode and generates a power adjustment command based on the new energy output prediction model and the dynamic digital model pre-stored in the local storage module to control the load control terminal to perform a second adjustment action.

[0010] After the edge computing cluster detects that the communication connection with the cloud scheduling center has been restored from the interrupted state to the normal state, the edge computing cluster will upload the operating data and power adjustment instruction records stored during the interrupted communication connection period to the cloud scheduling center after compression, so that the cloud scheduling center can perform data synchronization and policy updates.

[0011] Furthermore, the construction of a dynamic digital model characterizing the real-time load adjustment capability of the plant area based on the operating parameters and local network status includes:

[0012] Based on the cell type parameters and historical operating data of the local electrolytic aluminum cell, a corresponding basic load model component is matched from a preset model component library; the basic load model component is used to characterize the power response characteristics of the local electrolytic aluminum cell under standard operating conditions.

[0013] The real-time thermal balance state parameters of the local aluminum electrolysis cell are obtained, and a thermal state correction factor is generated based on the real-time thermal balance state parameters.

[0014] The basic load model components are modified according to the thermal state correction factor to generate the dynamic digital model characterizing the real-time load adjustment capability of the plant area.

[0015] Furthermore, the thermal state correction factor includes an adjustment rate attenuation coefficient and an adjustment depth constraint coefficient; the adjustment rate attenuation coefficient is determined based on the deviation between the current cell temperature and the target cell temperature of the local electrolytic aluminum cell, and is used to correct the power adjustment response time constant in the basic load model component; the adjustment depth constraint coefficient is determined based on the electrode distance state parameter of the local electrolytic aluminum cell, and is used to correct the maximum adjustable power amplitude in the basic load model component.

[0016] Furthermore, the process by which the cloud-based scheduling center generates a global scheduling strategy for the power allocation of electrolytic aluminum loads includes:

[0017] Obtain the real-time status information of the factory area uploaded by each of the edge computing clusters, wherein the real-time status information of the factory area includes the total adjustable power margin of the factory area in the dynamic digital model;

[0018] If the total adjustable power margin of any plant area is less than the preset minimum adjustment threshold, the cloud scheduling center marks the plant area as an adjustable-limited plant area, and when generating a power allocation scheme, reduces the load adjustment task weight of the adjustable-limited plant area, or transfers the adjustment task of the adjustable-limited plant area to other plants.

[0019] Furthermore, the process by which the cloud-based scheduling center generates a global scheduling policy also includes:

[0020] Obtain the local network status uploaded by the edge computing cluster;

[0021] If the local network status of any factory area meets the preset low-quality communication conditions, the cloud scheduling center will simultaneously issue a strategy execution period parameter for extending the validity period when issuing a global scheduling strategy to the edge computing cluster corresponding to that factory area, or simultaneously update the new energy output prediction model pre-stored in the edge computing cluster corresponding to that factory area.

[0022] Furthermore, the generation of power adjustment commands to control the load control terminal to perform a first adjustment action includes:

[0023] The global scheduling strategy is analyzed to obtain the total load adjustment target value for the factory area where the edge computing cluster is located; wherein, the total load adjustment target value is determined based on the adjustment task weight adjustment result of the cloud scheduling center for the factory area with limited adjustment.

[0024] Based on the real-time adjustable power margin corresponding to each of the electrolytic aluminum cells in the dynamic digital model, the total load adjustment target value is decomposed into the initial adjustment components corresponding to each of the electrolytic aluminum cells;

[0025] For any of the aforementioned electrolytic aluminum cells, based on the power regulation response time constant corresponding to the electrolytic aluminum cell in the dynamic digital model, the initial regulation component is subjected to regulation rate constraint processing to generate an executable regulation command, and the executable regulation command is sent to the load control terminal corresponding to the electrolytic aluminum cell.

[0026] Furthermore, the autonomous mode is a working mode in which the edge computing cluster independently makes load adjustment decisions during communication interruptions with the cloud scheduling center.

[0027] Furthermore, the generation of power regulation commands to control the load control terminal to perform a second regulation action includes:

[0028] The real-time thermal balance state parameters of the local aluminum electrolysis cell are obtained, and the adjustable power margin of the local aluminum electrolysis cell at the current moment is calculated based on the real-time thermal balance state parameters and the preset safe operating boundary of the aluminum electrolysis cell.

[0029] Based on the adjustable power margin and the predicted power curve output by the new energy output prediction model, a total power adjustment command is generated under the condition that the thermal balance safety constraint of the local electrolytic aluminum cell is met.

[0030] Furthermore, after generating the total power adjustment command, it also includes:

[0031] Based on the adjustable power margin of each electrolytic aluminum cell in the local electrolytic aluminum cell, the adjustment priorities of each electrolytic aluminum cell are sorted, and the corresponding adjustment components in the total power adjustment command are allocated to the corresponding load control terminal in descending order of adjustment priority.

[0032] Furthermore, the edge computing cluster will compress and upload the operational data and power adjustment command records stored during the period when the communication connection was interrupted to the cloud scheduling center, including:

[0033] After detecting that the communication connection status has recovered from the interrupted state to the normal state, the data to be uploaded stored during the interrupted communication connection status is acquired; the data to be uploaded includes operating data and power adjustment command records;

[0034] The data to be uploaded is classified into a first category of data and a second category of data; the first category of data is the power adjustment instruction record, and the second category of data is the data in the operation data other than the power adjustment instruction record.

[0035] The first type of data is compressed using a first compression strategy to generate a first compressed data packet; the second type of data is compressed using a second compression strategy to generate a second compressed data packet; wherein, the compression priority of the first compression strategy is higher than that of the second compression strategy.

[0036] The first compressed data packet and the second compressed data packet are uploaded to the cloud scheduling center in descending order of compression priority.

[0037] As can be seen from the above technical solutions, the present invention has the following advantages:

[0038] This invention constructs a three-layer distributed collaborative architecture consisting of a cloud-based scheduling center, an edge computing cluster, and a load control terminal. The edge computing cluster collects the operating parameters and network status of the local electrolytic aluminum cell in real time, constructs a dynamic digital model characterizing the plant's real-time load adjustment capability, and stores it locally. When communication is normal, the edge computing cluster receives the global scheduling strategy issued by the cloud and generates power adjustment commands based on the dynamic digital model to execute the first adjustment action. When communication is interrupted, the edge computing cluster automatically enters autonomous mode and independently generates power adjustment commands based on the locally pre-stored new energy output prediction model and dynamic digital model to execute the second adjustment action. When communication is restored, the edge computing cluster compresses the operating data and adjustment command records stored during the interruption and uploads them to the cloud to complete data synchronization and strategy updates. This invention decentralizes load regulation decision-making capabilities to the edge of the plant area, enabling the electrolytic aluminum load to continue flexible regulation based on local prediction models and real-time operating conditions even when communication networks are interrupted. This avoids regulation failures and wind / solar curtailment caused by communication outages. Simultaneously, by introducing a dynamic digital model, the generation of regulation commands can be adapted to the real-time thermal balance state and safe operating boundaries of each electrolytic aluminum cell, improving the safety and precision of the regulation process. This effectively reduces the dependence of electrolytic aluminum load regulation on centralized communication networks and significantly improves the reliability of grid power balance response and the efficiency of renewable energy absorption in scenarios with a high proportion of renewable energy integration. Attached Figure Description

[0039] Figure 1 This is a schematic flowchart of an embodiment of a distributed collaborative control system for electrolytic aluminum load based on edge cluster management in this invention.

[0040] Figure 2 This is a schematic diagram of the process for generating a dynamic digital model characterizing the real-time load adjustment capability of a plant area in this invention.

[0041] Figure 3 This is a schematic diagram of the process of generating power adjustment commands to control the load control terminal to perform the first adjustment action in this invention;

[0042] Figure 4 This is a schematic diagram of the process of generating power adjustment commands to control the load control terminal to perform the second adjustment action in this invention;

[0043] Figure 5 This is a schematic diagram of the data synchronization process after the edge computing cluster confirms the restoration of communication in this invention. Detailed Implementation

[0044] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0045] Example

[0046] This invention discloses a distributed collaborative control system for electrolytic aluminum loads based on edge cluster management, comprising a cloud-based dispatch center, edge computing clusters deployed in various electrolytic aluminum plant areas, and several load control terminals located within the plant areas. The cloud-based dispatch center is a centralized management and control platform deployed on the power grid dispatch side or a third-party aggregation service provider side. It possesses a high-performance computing server cluster and a database storage array, and is responsible for generating a global dispatch strategy for electrolytic aluminum load power allocation based on the forecast of renewable energy output across the entire network, the power grid operating status, and cross-plant load demand. The edge computing clusters are distributed computing node groups deployed in each electrolytic aluminum plant area. Each edge computing cluster consists of at least two edge computing servers for redundancy backup. The servers are connected to each load control terminal within the plant area via industrial Ethernet and establish a communication link with the cloud-based dispatch center via a dedicated fiber optic network. The load control terminals within the plant area are installed in the rectifier transformer control cabinet of each electrolytic aluminum cell, used to adjust the DC power consumption of the corresponding electrolytic aluminum cell. The number of load control terminals in a single electrolytic aluminum plant area corresponds one-to-one with the number of electrolytic aluminum cells in operation within that area, ranging from dozens to hundreds in this case. Each load control terminal interacts with the edge computing cluster in the plant area in real time through an industrial fieldbus communication protocol based on the transmission control protocol. After the edge computing cluster aggregates the operating parameters uploaded by each load control terminal, it completes data synchronization and policy distribution with the cloud dispatch center through a standardized interface protocol.

[0047] Please see Figure 1The edge cluster-managed distributed collaborative control system for electrolytic aluminum loads is used to execute the following methods:

[0048] S1. The edge computing cluster collects the operating parameters and local network status of the local electrolytic aluminum cell in real time, and constructs a dynamic digital model that characterizes the real-time load adjustment capability of the plant based on the operating parameters and local network status, and stores the dynamic digital model in the local storage module of the edge computing cluster.

[0049] Each electrolytic aluminum plant deploys an edge computing cluster, consisting of at least two edge computing servers interconnected via an industrial Ethernet network for redundancy. The edge computing cluster establishes communication with the load control terminal corresponding to each electrolytic aluminum cell via the plant's industrial Ethernet network. The load control terminal incorporates voltage transformers, current transformers, and temperature transmitters to collect real-time electrical and thermodynamic parameters of the corresponding electrolytic aluminum cell. A local electrolytic aluminum cell refers to the set of electrolytic aluminum cells deployed within the current plant area, in operation, and controlled by this edge computing cluster. The edge computing cluster polls each load control terminal at a set acquisition period, ranging from milliseconds to seconds depending on the plant's adjustment and response requirements. Operating parameters include the DC voltage, DC current, cell resistance, cell temperature, electrode spacing, and electrolyte temperature gradient of the aluminum electrolytic cell. DC voltage refers to the DC potential difference between the anode conductor and the cathode busbar, reflecting the electrical energy input state of the electrolytic cell. DC current refers to the DC current flowing through the aluminum electrolytic cell, output by a rectifier transformer and introduced into the cell through the anode. Cell resistance is the ratio of DC voltage to DC current, a comprehensive electrical parameter characterizing the electrochemical and thermal equilibrium states of the electrolytic cell. Cell temperature refers to the real-time temperature of the electrolyte layer within the aluminum electrolytic cell, obtained through measurement using thermocouples embedded in the cell sidewall. Electrode spacing refers to the vertical distance between the bottom of the anode and the mirror surface of the molten aluminum cathode, indirectly calculated by measuring the voltage fluctuation between the anode conductor and the cathode busbar. Electrolyte temperature gradient refers to the rate of temperature change per unit distance in the vertical direction of the electrolyte layer, used to assess the uniformity of the thermal field distribution within the electrolytic cell. The local network status includes the communication latency and packet loss rate between the edge computing cluster and each load control terminal, as well as the connectivity and signal strength of the communication link between the edge computing cluster and the cloud dispatch center. Communication latency refers to the time interval between the edge computing cluster sending a request message and receiving a response message from the load control terminal, used to assess the real-time performance of communication within the plant area. Packet loss rate refers to the ratio of the number of response messages that failed to be received within the statistical period to the total number of request messages sent, used to assess the reliability of communication within the plant area. The connectivity indicator is a connection status indicator of the communication link between the edge computing cluster and the cloud dispatch center, with values ​​ranging from a logical true value indicating link connectivity to a logical false value indicating link interruption. Signal strength refers to the power level of the communication signal received by the edge computing cluster from the cloud dispatch center, used to predict whether the communication link is showing signs of degradation.

[0050] Please see Figure 2 After obtaining the above operating parameters and local network status, the edge computing cluster caches the data in an in-memory database and executes subsequent steps S11-S13 based on the data:

[0051] S11. Based on the cell type parameters and historical operating data of the local electrolytic aluminum cell, match the corresponding basic load model component from the preset model component library; the basic load model component is used to characterize the power response characteristics of the local electrolytic aluminum cell under standard operating conditions.

[0052] The pre-defined model component library is a set of pre-built mathematical models stored locally on the edge computing cluster. This model set is pre-built according to the mainstream cell types in the electrolytic aluminum industry, including standard power response characteristic curves corresponding to different capacity levels, different rectifier transformer connection groups, and different electrolysis process types. Cell type parameters include the rated DC voltage, rated DC current, cell size, cathode carbon block type, and bus configuration of the electrolytic aluminum cell. Historical operating data includes DC power regulation records, cell temperature change curves, and cell resistance fluctuation range for the electrolytic aluminum cell in at least one complete production cycle. The edge computing cluster matches the cell type parameters of the local electrolytic aluminum cell with the index tags of each pre-stored basic load model component in the model component library. The index tags are generated by combining key fields of the cell type parameters. Upon successful matching, the corresponding basic load model component is retrieved. Mathematically, the basic load model component is expressed as a set of transfer functions, with the desired power regulation as input and the power response time constant and maximum regulation amplitude achievable by the electrolytic aluminum cell under standard operating conditions as output. The standard operating condition here refers to the typical operating range where the electrolytic aluminum cell is in normal thermal equilibrium, the electrode distance is stable, and there is no anode effect.

[0053] S12. Obtain the real-time thermal balance state parameters of the local aluminum electrolysis cell, and generate a thermal state correction factor based on the real-time thermal balance state parameters;

[0054] Real-time thermal balance parameters include the current cell temperature, electrolyte temperature, cell shell temperature, anode current distribution uniformity, and electrode gap voltage fluctuation. The current cell temperature is obtained in real-time through thermocouples embedded in the sidewall of the electrolytic cell. The electrolyte temperature is obtained through intermittent sampling by an online monitoring device. The electrode gap voltage fluctuation is obtained by collecting the voltage difference between the anode conductor and the cathode bus and calculating its standard deviation. The edge computing cluster compares the above parameters with the preset safe operating boundaries of the corresponding electrolytic aluminum cell. These preset safe operating boundaries include the upper limit of cell temperature, the lower limit of cell temperature, the safe electrode gap range, and the allowable deviation range of anode current distribution. In this embodiment, the thermal state correction factor is a set of dimensionless coefficients calculated by the edge computing cluster based on the degree to which the real-time thermal balance parameters deviate from their standard values. It is used to quantify the degree of attenuation and constraint tightening of the electrolytic aluminum cell's response to power regulation commands under the current thermal state. The thermal state correction factor is not a fixed value but a time-varying parameter that dynamically changes with the thermal balance state of the electrolytic aluminum cell.

[0055] S13. Correct the basic load model components according to the thermal state correction factor to generate a dynamic digital model characterizing the real-time load adjustment capability of the plant area.

[0056] In this embodiment, the thermal state correction factor includes the adjustment rate attenuation coefficient and the adjustment depth constraint coefficient. The adjustment rate attenuation coefficient is determined based on the deviation between the current cell temperature and the target cell temperature of the local electrolytic aluminum cell, and is used to correct the power adjustment response time constant in the basic load model component. The adjustment depth constraint coefficient is determined based on the electrode distance state parameter of the local electrolytic aluminum cell, and is used to correct the maximum adjustable power amplitude in the basic load model component.

[0057] Specifically, the current cell temperature is a real-time measurement value collected by the edge computing cluster at the moment of acquisition, while the target cell temperature is the preset optimal thermal balance operating temperature of the electrolytic aluminum cell in the current production stage, set by electrolytic process technicians based on cell age and production targets. The larger the deviation between the current cell temperature and the target cell temperature, the more serious the deviation in the thermal state of the electrolytic aluminum cell. If power regulation is still performed using the standard response time constant at this time, it may cause the cell temperature to deviate further or trigger the anode effect. Therefore, it is necessary to generate a regulation rate attenuation coefficient greater than one based on the deviation value, and multiply the original response time constant by this coefficient to obtain the actually usable extended response time constant.

[0058] The electrode gap state parameter is the ratio of the electrode gap voltage fluctuation value to the average electrode gap voltage. This ratio reflects the stability of the distance between the aluminum liquid layers of the anode and cathode. When the electrode gap state parameter exceeds the preset stability threshold, it indicates that the electrode gap is in an unstable state. At this time, performing large power adjustment may trigger electrolyte splashing or partial short circuit. Therefore, an adjustment depth constraint coefficient between zero and one is generated based on the electrode gap state parameter. Multiplying the original maximum adjustable power amplitude by this coefficient yields the actual maximum allowable power adjustment depth at the current moment.

[0059] The edge computing cluster replaces the power regulation response time constant in the basic load model component with an extended response time constant corrected by the regulation rate attenuation coefficient, and replaces the maximum adjustable power amplitude in the basic load model component with the actual maximum allowable power regulation depth corrected by the regulation depth constraint coefficient. The corrected model is the dynamic digital model characterizing the real-time load regulation capability of the plant area. The dynamic digital model uses each electrolytic aluminum cell as the smallest unit, outputting the adjustable power margin, maximum allowable regulation rate, and minimum safe operating power limit for each electrolytic aluminum cell at the current moment. These results are then aggregated by plant area dimension to form the total adjustable power margin and average regulation response time of the plant area. The edge computing cluster stores this dynamic digital model in a structured data format in a local storage module, which is a solid-state drive array built into the edge computing server. This ensures that the model data is not lost during communication interruptions and can be accessed at any time by the autonomous decision-making module of the edge computing cluster.

[0060] S2. When the communication connection between the edge computing cluster and the cloud scheduling center is in a normal state, the edge computing cluster receives the global scheduling strategy for the power allocation of the electrolytic aluminum load issued by the cloud scheduling center, and generates a power adjustment command in combination with the dynamic digital model to control the load control terminal to perform the first adjustment action.

[0061] A normal communication connection status refers to a working state where the communication link between the edge computing cluster and the cloud scheduling center is bidirectionally reachable, has stable latency, and a packet loss rate below a set threshold. In this embodiment, the normal status also requires that the communication latency is less than a first threshold and the number of consecutive packet losses is less than a second threshold. The first threshold is preferably 200 milliseconds, and the second threshold is preferably three data packets. When the communication connection status is normal, the edge computing cluster, as the execution unit of the cloud scheduling strategy, receives the global scheduling strategy issued by the cloud and generates power adjustment instructions based on the strategy and the local dynamic digital model. The first adjustment action refers to the power adjustment behavior performed by the edge computing cluster on each electrolytic aluminum cell in the plant area under the unified cloud scheduling framework, based on the global optimization objective.

[0062] In this embodiment, the process by which the cloud-based scheduling center generates a global scheduling strategy for the power allocation of electrolytic aluminum loads includes:

[0063] 1. Obtain real-time status information of the plant area uploaded by each edge computing cluster. The real-time status information of the plant area includes the total adjustable power margin of the plant area in the dynamic digital model.

[0064] 2. If the total adjustable power margin of any plant area is less than the preset minimum adjustment threshold, the cloud dispatch center will mark the plant area as an adjustable-limited plant area, and when generating the power allocation scheme, reduce the load adjustment task weight of the adjustable-limited plant area, or transfer the adjustment task of the adjustable-limited plant area to other plants area.

[0065] Specifically, the real-time status information of the plant area includes the total adjustable power margin of the plant area in the dynamic digital model, the number of online operating electrolytic aluminum cells in the plant area, the current total load of the plant area, and the average power factor of the plant area. The total adjustable power margin of the plant area is obtained by the edge computing cluster summing the adjustable power margins of each electrolytic aluminum cell in the dynamic digital model. The preset minimum adjustment threshold is a threshold value determined by the cloud dispatch center based on the reserve capacity requirements issued by the power grid dispatch department and the historical adjustment contribution rate of the plant area. The preferred range for this minimum adjustment threshold is 5% to 15% of the rated load of the plant area; in this embodiment, it is set to 10% of the rated load of the plant area. If the total adjustable power margin of any plant area is less than the preset minimum adjustment threshold, it indicates that the plant area does not have sufficient power adjustment space under the current thermal balance constraints. If a large adjustment task is continued to be issued to it, it may cause some electrolytic aluminum cells in the plant area to be forced to exceed the safe operating boundary, triggering the risk of thermal imbalance. Plants with limited adjustment capacity are those marked by the cloud-based scheduling center as having insufficient adjustment capacity in the current scheduling cycle and requiring a reduction in the weight of adjustment tasks.

[0066] When generating power allocation schemes, the cloud-based dispatch center reduces the load regulation task weight of restricted power plants by multiplying the plant's allocation ratio in the overall grid regulation task by an attenuation factor less than one. The attenuation factor is linearly determined based on the ratio of the plant's total adjustable power margin to the minimum regulation threshold; the smaller the ratio, the smaller the attenuation factor. Transferring regulation tasks from restricted power plants to other plants involves the cloud-based dispatch center iterating through the total adjustable power margins of other unrestricted power plants and redistributing the regulation tasks originally borne by the restricted power plants to the unrestricted plants according to their remaining adjustable margin ratios. These two operations can be performed individually or simultaneously, depending on grid dispatch requirements. First, the task weight of restricted power plants is reduced; if a gap still exists in the overall grid regulation task after the reduction, the gap is then transferred to other power plants.

[0067] In this embodiment, the process of generating a global scheduling policy by the cloud scheduling center also includes acquiring and responding to the local network status:

[0068] 1. Obtain the local network status uploaded by the edge computing cluster;

[0069] 2. If the local network status of any plant area meets the preset low-quality communication conditions, the cloud dispatch center will simultaneously issue a strategy execution period parameter for extending the validity period when issuing a global dispatch strategy to the edge computing cluster corresponding to that plant area, or simultaneously update the new energy output prediction model stored in the edge computing cluster corresponding to that plant area.

[0070] Specifically, the local network status includes the average latency, maximum latency, packet loss rate, and link jitter value of the communication link between the edge computing cluster and the cloud dispatch center over the past period. The preset low-quality communication conditions are defined as any of the above indicators exceeding a corresponding preset threshold: average latency exceeding 500 milliseconds, maximum latency exceeding one second, packet loss rate exceeding 5%, or link jitter value exceeding 100 milliseconds. When the local network status of any factory area meets any one or more of the above preset low-quality communication conditions, the cloud dispatch center determines that the edge computing cluster corresponding to that factory area is in a state of degraded communication quality, posing a risk of delay or failure in the subsequent dispatch strategy issuance.

[0071] In this scenario, when the cloud-based dispatch center issues a global dispatch policy to the edge computing cluster corresponding to the plant area, it simultaneously issues a policy execution period parameter extending the validity period. Specifically, the cloud-based dispatch center extends the policy execution period parameter from the default one scheduling cycle to two or three scheduling cycles. This extended validity period parameter is issued to the edge computing cluster along with the global dispatch policy. If a communication interruption occurs during the policy validity period, the edge computing cluster continues to execute adjustment actions according to the extended validity policy until the policy validity period expires or communication is restored and a new policy update is received. Alternatively, the cloud-based dispatch center can choose to simultaneously update the pre-stored renewable energy output prediction model in the edge computing cluster corresponding to the plant area. Specifically, the cloud-based dispatch center pushes the latest renewable energy output prediction model file to the local storage module of the edge computing cluster via incremental updates, overwriting the original pre-stored model. The above two operation methods can be selected based on the actual degree of communication quality degradation, or they can be executed simultaneously. When the network status is unstable, both the policy validity period and the local prediction model are extended to maximize the autonomous adjustment capability of the edge computing cluster during potential communication interruptions.

[0072] Please see Figure 3 Generating a power regulation command to control the load control terminal to perform a first regulation action includes steps S21-S23:

[0073] S21. Analyze the global scheduling strategy to obtain the total load adjustment target value for the factory area where the edge computing cluster is located; wherein, the total load adjustment target value is determined based on the adjustment task weight adjustment result of the cloud scheduling center for the factory area with limited adjustment;

[0074] S22. Based on the real-time adjustable power margin corresponding to each electrolytic aluminum cell in the dynamic digital model, the total load adjustment target value is decomposed into the initial adjustment components corresponding to each electrolytic aluminum cell;

[0075] S23. For any electrolytic aluminum cell, based on the power regulation response time constant corresponding to the electrolytic aluminum cell in the dynamic digital model, the initial regulation component is subjected to regulation rate constraint processing to generate an executable regulation command, and the executable regulation command is sent to the load control terminal corresponding to the electrolytic aluminum cell.

[0076] Specifically, after receiving the global scheduling policy from the cloud scheduling center, the edge computing cluster first parses the policy data packet to extract the total load adjustment target value corresponding to the plant area where the edge computing cluster is located. The total load adjustment target value represents the total power increase or decrease that the plant area needs to undertake within the current scheduling cycle. This value has been optimized and allocated across the entire network by the cloud scheduling center based on the total adjustable power margin and adjustment limitation flags of each plant area. The real-time adjustable power margin is the maximum increase or decrease that a single electrolytic aluminum cell can safely participate in power adjustment under the current thermal balance state. The edge computing cluster decomposes the total load adjustment target value into initial adjustment components for each electrolytic aluminum cell according to the proportion of the real-time adjustable power margin of each electrolytic aluminum cell to the total adjustable power margin of all electrolytic aluminum cells in the plant area. For electrolytic aluminum cells with a larger adjustable power margin, the initial adjustment component allocated to them is correspondingly larger; for electrolytic aluminum cells with an adjustable power margin of zero or close to zero, the initial adjustment component allocated to them is zero, meaning they do not participate in this adjustment task.

[0077] For any electrolytic aluminum cell, the edge computing cluster reads the corresponding power regulation response time constant from the dynamic digital model. This power regulation response time constant has been corrected by the regulation rate attenuation coefficient during the construction of the aforementioned dynamic digital model. The initial regulation component is multiplied by the reciprocal of the power regulation response time constant to obtain the allowable power change for the electrolytic aluminum cell within the current sampling interval. If this power change is greater than the initial regulation component, only the regulation amplitude corresponding to this power change is executed in this cycle, and the remaining regulation component is carried over to the next cycle. The instruction generated after regulation rate constraint processing is an executable regulation instruction, which includes the target power setpoint and regulation rate ramp parameters. The edge computing cluster sends this executable regulation instruction to the load control terminal of the corresponding electrolytic aluminum cell via industrial Ethernet. The load control terminal adjusts the on-load tap changer position or phase-controlled trigger angle of the rectifier transformer according to the received executable regulation instruction, thereby changing the DC power consumption of the electrolytic aluminum cell and completing the execution of the first regulation action.

[0078] S3. When the edge computing cluster detects that the communication connection status between the edge computing cluster and the cloud dispatch center has changed from normal to interrupted, the edge computing cluster enters autonomous mode and generates power adjustment instructions based on the new energy output prediction model and dynamic digital model pre-stored in the local storage module to control the load control terminal to perform the second adjustment action.

[0079] A change in communication connection status from normal to interrupted indicates that the edge computing cluster has sent three consecutive heartbeat probe packets to the cloud scheduling center without receiving a response, or that the edge computing cluster has detected a loss of physical layer signal in the communication link with the cloud scheduling center. When the edge computing cluster detects that the communication latency lasts for more than one second and the packet loss rate exceeds 10% for more than thirty seconds, it proactively marks the communication connection status as interrupted and triggers the autonomous mode preparation process in advance, thereby avoiding the gap in adjustment response caused by passive switching after a complete communication interruption.

[0080] In this embodiment, the autonomous mode is the working mode in which the edge computing cluster independently makes load adjustment decisions during the period when communication with the cloud scheduling center is interrupted.

[0081] Please see Figure 4 The process includes generating a power regulation command to control the load control terminal to perform a second regulation action, including steps S31-S33:

[0082] S31. Obtain the real-time thermal balance state parameters of the local electrolytic aluminum cell, and calculate the adjustable power margin of the local electrolytic aluminum cell at the current moment based on the real-time thermal balance state parameters and the preset safe operating boundary of the electrolytic aluminum cell.

[0083] The real-time thermal balance parameters and the preset safe operating boundary of the electrolytic aluminum cell here are defined in the same way as the real-time thermal balance parameters in step S12. In autonomous mode, the edge computing cluster uses the same parameters and boundaries to calculate the adjustable power margin of the local electrolytic aluminum cell at the current moment using the same algorithm logic as in normal mode. The adjustable power margin here is consistent with the real-time adjustable power margin involved in the global scheduling strategy in step S2 in terms of physical meaning and calculation method. Both refer to the maximum upward or downward adjustment range that a single electrolytic aluminum cell can safely participate in power regulation in the current thermal balance state. The only difference is the usage scenario: in step S2, this margin is used to allocate the total adjustment target in the cloud, while in step S31, this margin is used by the edge computing cluster to determine the total adjustment amount in autonomous mode.

[0084] S32. Based on the adjustable power margin and the predicted power curve output by the new energy output prediction model, generate a total power adjustment command under the condition of satisfying the thermal balance safety constraint of the local electrolytic aluminum cell.

[0085] The renewable energy output prediction model is a time-series prediction model pre-trained and stored in the local storage module of the edge computing cluster. This model is trained based on historical renewable energy output data and meteorological forecast data, and can output predicted renewable energy output values ​​for a future period based on meteorological parameters such as wind speed and irradiance at the current moment. The predicted power curve is a continuous curve output by the renewable energy output prediction model, with time as the horizontal axis and power as the vertical axis. This curve reflects the changing trend of renewable energy plant output within a preset time period, preferably fifteen minutes to two hours. The thermal balance safety constraints for the local electrolytic aluminum cells are that, during autonomous regulation, the power regulation behavior of each electrolytic aluminum cell must not cause its cell temperature to exceed the upper limit or fall below the lower limit, must not cause the electrode spacing to deviate from the safe range, and must not cause the anode current distribution deviation to exceed the allowable range. The edge computing cluster compares the fluctuation in renewable energy output represented by the predicted power curve with the current total load of the plant to obtain the total load adjustment demand required to balance the renewable energy fluctuations. Then, it compares the total load adjustment demand with the sum of the adjustable power margins of each electrolytic aluminum cell calculated in step S31, and takes the smaller absolute value as the value of the total power adjustment command. This ensures that the generated total power adjustment command meets the renewable energy consumption demand without exceeding the thermal balance safety constraints of the local electrolytic aluminum cells. This total power adjustment command is the total adjustment target value corresponding to the second adjustment action.

[0086] S33. Based on the adjustable power margin of each electrolytic aluminum cell in the local electrolytic aluminum cell, sort the adjustment priorities of each electrolytic aluminum cell, and allocate the corresponding adjustment component in the total power adjustment command to the corresponding load control terminal in descending order of adjustment priority.

[0087] Specifically, after receiving the total power regulation command, the edge computing cluster prioritizes the regulation based on the adjustable power margin of each electrolytic aluminum cell in the local electrolytic aluminum tank. Electrolytic aluminum cells with larger adjustable power margins indicate more stable current thermal conditions and stronger tolerance to power fluctuations, thus receiving higher regulation priority. Conversely, electrolytic aluminum cells with smaller or zero adjustable power margins are assigned lower priority or are not involved in the current regulation. For example, if a factory has four electrolytic aluminum cells with adjustable power margins of 15%, 10%, 5%, and 0% of the rated power, respectively, the regulation priorities from highest to lowest are: the first electrolytic aluminum cell, the second, the third, and the fourth, which is excluded from ranking and regulation due to its zero margin. The edge computing cluster then distributes the corresponding regulation components from the total power regulation command to the load control terminals of the corresponding electrolytic aluminum cells in descending order of priority until all total power regulation commands have been distributed. For each electrolytic aluminum cell assigned an adjustment component, the edge computing cluster also performs adjustment rate constraint processing on the assigned component based on the power adjustment response time constant corresponding to the electrolytic aluminum cell in the dynamic digital model, generates the final executable adjustment command and issues it, thereby completing the execution of the second adjustment action.

[0088] S4. After the edge computing cluster detects that the communication connection between the edge computing cluster and the cloud scheduling center has been restored from the interrupted state to the normal state, the edge computing cluster will upload the operation data and power adjustment instruction records stored during the communication connection interruption period to the cloud scheduling center after compression, so that the cloud scheduling center can perform data synchronization and policy updates.

[0089] The communication connection status recovers from an interrupted state to a normal state as follows: the heartbeat probe packets sent by the edge computing cluster to the cloud scheduling center receive responses three consecutive times, and the communication latency falls below the first threshold and the packet loss rate falls below the second threshold and remains below for more than 30 seconds. After confirming the communication recovery, the edge computing cluster initiates the data synchronization process. Please refer to [link to relevant documentation]. Figure 5 Perform the following steps S41 to S44:

[0090] S41. After detecting that the communication connection status has recovered from the interrupted state to the normal state, acquire the data to be uploaded stored during the period when the communication connection status was interrupted; the data to be uploaded includes operating data and power adjustment command records;

[0091] S42. Classify the data to be uploaded to obtain a first category of data and a second category of data; the first category of data is the power adjustment command record, and the second category of data is the running data excluding the power adjustment command record.

[0092] S43. The first type of data is compressed using a first compression strategy to generate a first compressed data packet; the second type of data is compressed using a second compression strategy to generate a second compressed data packet; wherein, the compression priority of the first compression strategy is higher than that of the second compression strategy;

[0093] S44. Upload the first compressed data packet and the second compressed data packet to the cloud scheduling center in descending order of compression priority.

[0094] Specifically, the operational data to be uploaded consists of electrical and thermodynamic parameters of the electrolytic aluminum cell continuously recorded by the edge computing cluster during communication interruptions according to a preset acquisition cycle. These parameters include DC voltage, DC current, cell resistance, cell temperature, and electrode gap voltage fluctuations. This data belongs to the same category as the historical operational data mentioned in step S11, but differs in that historical operational data refers to data accumulated over a long period before the interruption, while the operational data here specifically refers to real-time operational records collected and temporarily stored locally during communication interruptions. The power adjustment command record is a complete record of each executable adjustment command generated and issued to the load control terminal by the edge computing cluster in autonomous mode. This record includes the command generation time, target power setpoint, adjustment rate ramp parameters, and the corresponding electrolytic aluminum cell number.

[0095] When classifying the data to be uploaded, the edge computing cluster categorizes power regulation command records into Category 1 data and other operational data (excluding power regulation command records) into Category 2 data. Category 1 data has a smaller volume but higher information density, reflecting the decision-making behavior and regulation effects of the edge computing cluster in autonomous mode, and has the highest reference value for strategy evaluation and model calibration in the cloud scheduling center. Category 2 data has a larger volume, containing a large number of continuous electrical and thermodynamic sampling values, used for subsequent offline analysis and model iteration optimization, but its urgency for restoring the global perspective in the cloud scheduling center is slightly lower than that of Category 1 data. The first compression strategy uses a lossless compression algorithm with an added cyclic redundancy check code, ensuring that the compressed data can be completely restored and has integrity verification capabilities. The second compression strategy uses a high compression ratio lossy compression algorithm to remove redundant information and downsample the continuous sampling data, significantly reducing the amount of transmitted data while meeting the data analysis accuracy requirements of the cloud scheduling center. The compression priority of the first compression strategy is higher than that of the second compression strategy, meaning that in the upload queue scheduling of the edge computing cluster, the first compressed data packet is given higher sending priority and occupies communication bandwidth resources first.

[0096] After generating the first and second compressed data packets, the edge computing cluster initiates upload tasks sequentially according to compression priority, from highest to lowest. Specifically, the first compressed data packet is first fragmented and transmitted to the cloud scheduling center. Only after the first compressed data packet is fully delivered and confirmed by the cloud does the upload process for the second compressed data packet begin. If communication is interrupted again during the upload of the second compressed data packet, since the first compressed data packet has been successfully delivered and the cloud scheduling center has obtained the core decision data from the interruption, it can first conduct policy evaluation and parameter updates. Once communication is restored, it can continue receiving the remaining portion of the second compressed data packet. This categorized compression and priority upload method ensures data integrity while minimizing the waiting time required for the cloud scheduling center to restore its global scheduling capabilities.

[0097] It is understood that those skilled in the art can combine various implementation methods in the above embodiments under the guidance of the above examples to obtain technical solutions with multiple implementation methods.

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A distributed collaborative control system for electrolytic aluminum load based on edge cluster management, characterized in that, include: The system includes a cloud-based dispatch center, edge computing clusters deployed in various electrolytic aluminum plant areas, and several load control terminals located within the plant areas. Each load control terminal is used to adjust the power consumption of the corresponding electrolytic aluminum cell and is communicatively connected to the edge computing cluster in its plant area. The edge computing cluster is communicatively connected to the cloud-based dispatch center. The electrolytic aluminum load distributed collaborative control system based on edge cluster management is used to execute the following methods: The edge computing cluster collects the operating parameters and local network status of the local electrolytic aluminum cell in real time, and constructs a dynamic digital model characterizing the real-time load adjustment capability of the plant area based on the operating parameters and local network status, and stores the dynamic digital model in the local storage module of the edge computing cluster. When the communication connection between the edge computing cluster and the cloud scheduling center is in a normal state, the edge computing cluster receives the global scheduling strategy for the power allocation of the electrolytic aluminum load issued by the cloud scheduling center, and generates a power adjustment command in combination with the dynamic digital model to control the load control terminal to perform the first adjustment action. When the edge computing cluster detects that the communication connection status between it and the cloud dispatch center changes from the normal state to the interrupted state, the edge computing cluster enters autonomous mode and generates a power adjustment command based on the new energy output prediction model and the dynamic digital model pre-stored in the local storage module to control the load control terminal to perform a second adjustment action. After the edge computing cluster detects that the communication connection with the cloud scheduling center has been restored from the interrupted state to the normal state, the edge computing cluster will upload the operating data and power adjustment instruction records stored during the interrupted communication connection period to the cloud scheduling center after compression, so that the cloud scheduling center can perform data synchronization and policy updates.

2. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 1, characterized in that, The construction of a dynamic digital model characterizing the real-time load adjustment capability of the plant area based on the operating parameters and local network status includes: Based on the cell type parameters and historical operating data of the local electrolytic aluminum cell, a corresponding basic load model component is matched from a preset model component library; the basic load model component is used to characterize the power response characteristics of the local electrolytic aluminum cell under standard operating conditions. The real-time thermal balance state parameters of the local aluminum electrolysis cell are obtained, and a thermal state correction factor is generated based on the real-time thermal balance state parameters. The basic load model components are modified according to the thermal state correction factor to generate the dynamic digital model characterizing the real-time load adjustment capability of the plant area.

3. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 2, characterized in that, The thermal state correction factor includes an adjustment rate attenuation coefficient and an adjustment depth constraint coefficient; the adjustment rate attenuation coefficient is determined based on the deviation between the current cell temperature and the target cell temperature of the local electrolytic aluminum cell, and is used to correct the power adjustment response time constant in the basic load model component; the adjustment depth constraint coefficient is determined based on the electrode distance state parameter of the local electrolytic aluminum cell, and is used to correct the maximum adjustable power amplitude in the basic load model component.

4. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 3, characterized in that, The process by which the cloud-based scheduling center generates a global scheduling strategy for the power allocation of electrolytic aluminum loads includes: Obtain the real-time status information of the factory area uploaded by each of the edge computing clusters, wherein the real-time status information of the factory area includes the total adjustable power margin of the factory area in the dynamic digital model; If the total adjustable power margin of any plant area is less than the preset minimum adjustment threshold, the cloud scheduling center marks the plant area as an adjustable-limited plant area, and when generating a power allocation scheme, reduces the load adjustment task weight of the adjustable-limited plant area, or transfers the adjustment task of the adjustable-limited plant area to other plants.

5. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 4, characterized in that, The process of generating a global scheduling policy by the cloud-based scheduling center also includes: Obtain the local network status uploaded by the edge computing cluster; If the local network status of any factory area meets the preset low-quality communication conditions, the cloud scheduling center will simultaneously issue a strategy execution period parameter for extending the validity period when issuing a global scheduling strategy to the edge computing cluster corresponding to that factory area, or simultaneously update the new energy output prediction model pre-stored in the edge computing cluster corresponding to that factory area.

6. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 4 or 5, characterized in that, The generation of power adjustment commands to control the load control terminal to perform a first adjustment action includes: The global scheduling strategy is analyzed to obtain the total load adjustment target value for the factory area where the edge computing cluster is located; wherein, the total load adjustment target value is determined based on the adjustment task weight adjustment result of the cloud scheduling center for the factory area with limited adjustment. Based on the real-time adjustable power margin corresponding to each of the electrolytic aluminum cells in the dynamic digital model, the total load adjustment target value is decomposed into the initial adjustment components corresponding to each of the electrolytic aluminum cells; For any of the aforementioned electrolytic aluminum cells, based on the power regulation response time constant corresponding to the electrolytic aluminum cell in the dynamic digital model, the initial regulation component is subjected to regulation rate constraint processing to generate an executable regulation command, and the executable regulation command is sent to the load control terminal corresponding to the electrolytic aluminum cell.

7. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 1, characterized in that, The autonomous mode is the working mode in which the edge computing cluster independently makes load adjustment decisions during the period when communication with the cloud scheduling center is interrupted.

8. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 7, characterized in that, The generated power adjustment command controls the load control terminal to perform a second adjustment action, including: The real-time thermal balance state parameters of the local aluminum electrolysis cell are obtained, and the adjustable power margin of the local aluminum electrolysis cell at the current moment is calculated based on the real-time thermal balance state parameters and the preset safe operating boundary of the aluminum electrolysis cell. Based on the adjustable power margin and the predicted power curve output by the new energy output prediction model, a total power adjustment command is generated under the condition that the thermal balance safety constraint of the local electrolytic aluminum cell is met.

9. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 8, characterized in that, After generating the total power adjustment command, the method further includes: Based on the adjustable power margin of each electrolytic aluminum cell in the local electrolytic aluminum cell, the adjustment priorities of each electrolytic aluminum cell are sorted, and the corresponding adjustment components in the total power adjustment command are allocated to the corresponding load control terminal in descending order of adjustment priority.

10. The distributed collaborative control system for electrolytic aluminum load based on edge cluster management according to claim 1, characterized in that, The edge computing cluster will upload compressed records of operational data and power adjustment commands stored during the period when the communication connection was interrupted to the cloud scheduling center, including: After detecting that the communication connection status has recovered from the interrupted state to the normal state, the data to be uploaded stored during the interrupted communication connection status is acquired; the data to be uploaded includes operating data and power adjustment command records; The data to be uploaded is classified into a first category of data and a second category of data; the first category of data is the power adjustment instruction record, and the second category of data is the data in the operation data other than the power adjustment instruction record. The first type of data is compressed using a first compression strategy to generate a first compressed data packet; the second type of data is compressed using a second compression strategy to generate a second compressed data packet; wherein, the compression priority of the first compression strategy is higher than that of the second compression strategy. The first compressed data packet and the second compressed data packet are uploaded to the cloud scheduling center in descending order of compression priority.

Images