Unit vibration area intelligent collaborative crossing system and method based on dynamic game and self-adaptive prediction

Abstract

The invention discloses a unit vibration area intelligent collaborative crossing system and method based on dynamic game and adaptive prediction, and belongs to the technical field of power system control, and the method comprises the steps: deploying a sensor on each generator set, and collecting the operation state data of the generator set; the edge node preprocesses locally acquired data, and predicts whether the generator set enters a vibration area in a preset time window by using a locally deployed adaptive prediction model; the central control unit collects prediction results of all edge nodes and constructs a dynamic game model; the central control unit issues an adjusting instruction with limited step length to the related generator set; adjusting a control strategy and parameters according to the simulation deviation; after control operation is executed, feedback data are continuously collected, the self-adaptive prediction model is updated, and closed-loop optimization of parameters is achieved. According to the invention, the safety and efficiency of vibration area crossing in an actual hydraulic power plant or intelligent micro-grid environment can be significantly improved.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, specifically to a unit vibration zone intelligent cooperative crossing system and method based on dynamic game theory and adaptive prediction. Background Technology

[0002] In traditional power generation systems, generating units have vibration zones, which are unstable areas where mechanical or electrical oscillations are prone to occur within certain load or speed ranges. Passing through these vibration zones can lead to severe vibrations, unstable operation, or even shutdowns. To avoid crossing these zones, it is generally necessary to gradually change the output or alternately start and stop the unit to ensure a smooth passage. However, this approach is inefficient and lacks flexibility.

[0003] Existing research has proposed scheduling optimization models that consider the risk of vibration zone crossings, aiming to reduce unnecessary crossings and water consumption. For example, in hydropower plant systems, optimal scheduling models that consider vibration zone risk can effectively reduce the number of unnecessary vibration zone crossings. However, these methods typically optimize at the planning level, lacking dynamic coordinated control of the real-time operating status of the units, and are ill-equipped to handle the complexities brought about by renewable energy fluctuations and the heterogeneity of multiple units. Current technologies have not fully explored the potential of combining dynamic game theory and adaptive prediction in coordinating the control of multiple units crossing vibration zones, nor have they constructed a structured and easily deployable intelligent system. Therefore, a new system architecture and methodology are urgently needed to coordinate the control of multiple units and achieve safe and efficient crossing of vibration zones. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is: how to address the fact that existing technologies have not fully explored the potential of combining dynamic game theory with adaptive prediction in the coordinated multi-unit travel control.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: an intelligent collaborative traversal system for generator vibration zones based on dynamic game theory and adaptive prediction, comprising a generator monitoring unit, an edge computing unit, a communication unit, a central control unit, a digital twin simulation unit, and a federated learning training unit; the generator monitoring unit installs multiple sensors on each generator set to collect generator status data in real time and send it to the local edge computing unit and the central control unit; the edge computing unit includes deploying edge computing nodes next to each generator set for performing local data preprocessing and predictive calculations; the communication unit includes constructing a bidirectional communication network using industrial communication protocols to connect each edge module and the central control unit, enabling real-time transmission of monitoring data, prediction results, and control commands; the central control unit includes summarizing data and prediction results from each generator set and executing a dynamic game theory algorithm; the digital twin simulation unit is used to establish a simulation model for each generator set and run it in real time; the federated learning training unit includes training the prediction model using a federated learning approach.

[0007] As a preferred embodiment of the intelligent collaborative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction described in this invention, the edge computing unit and the unit monitoring unit are deployed in a one-to-one correspondence, responsible for running the adaptive prediction model locally to accurately predict future vibration trends; the communication network is used to connect the edge computing unit and the central control unit to realize real-time exchange of data and instructions; the central control unit is used to establish a dynamic game model based on the prediction data of each unit, solve the Nash equilibrium, and generate collaborative control instructions for each unit.

[0008] As a preferred embodiment of the intelligent collaborative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction described in this invention, the central control unit includes a game analysis module, a Shapley calculation module, and an optimization solution module.

[0009] As a preferred embodiment of the intelligent cooperative traversal system for generator vibration zones based on dynamic game theory and adaptive prediction described in this invention, the game analysis module is used to construct a game utility function with cooperative terms based on the predicted output and vibration index of each generator unit; the Shapley calculation module is used to calculate the Shapley value for each participating generator unit. Represented as:

[0010]

[0011] in, For each participating unit, there is a Shapley value, N is the set of all participating units in the cross-traffic collaboration, S is a subset excluding the i-th unit, i is the i-th unit, f(S) is the collaboration benefit that the set of units S can achieve, and f(S∪{i}) is the collaboration benefit that the set of all participating units in the cross-traffic collaboration can achieve. The optimization solution module is used to solve the Nash equilibrium of the game and determine the optimal output adjustment value for all units.

[0012] As a preferred embodiment of the intelligent collaborative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction described in this invention, the adaptive prediction model integrates the advantages of long short-term memory networks and autoregressive integral moving average models, comprehensively analyzes nonlinear and linear time series characteristics, and the dynamic game model uses the Shapley value of each unit as the revenue distribution mechanism.

[0013] As a preferred embodiment of the intelligent collaborative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction described in this invention, the autoregressive integral moving average model linearly models the time series using differencing, AR, and MA components, and is expressed as:

[0014] Δ d Y t =φ1Y t-1 +…+φ p Y t-p +ε t +θ1ε t-1 +…+θ q ε t-q

[0015] Where, Δ d Y t For a stationary sequence after d differencing, φ1...φ p The coefficients of the autoregressive term are θ1...θ2. q ε is the coefficient of the moving average term. t ...ε t-q Y is the random error term. t-1 Let be the value of the t-1 time series, p be the order of the autoregressive term, and q be the order of the moving average term.

[0016] As a preferred embodiment of the intelligent collaborative vibration zone crossing system for generator units based on dynamic game theory and adaptive prediction described in this invention, the central control unit adopts an event-triggered small-step control strategy, and sets a trigger threshold e for each generator unit i based on real-time error monitoring. i (t), when ||e i (t)||≥γ i ||e i (t kWhen ||, the i-th unit is triggered to perform a small-amplitude output pulse adjustment.

[0017] This invention provides a method for intelligent collaborative traversal of generator vibration zones based on dynamic game theory and adaptive prediction.

[0018] To address the aforementioned technical problems, this invention provides the following technical solution: a method for intelligent collaborative traversal of generator vibration zones based on dynamic game theory and adaptive prediction, comprising: deploying sensors on each generator unit to collect operating status data and sending the data to edge nodes and a central control unit; preprocessing the locally collected data at the edge nodes and using locally deployed adaptive prediction models to predict whether the generator unit will enter the vibration zone within a predetermined time window; each edge node trains its adaptive prediction model locally, uploading model parameters only to the central control unit, which then fuses the uploaded model parameters to generate a unified model and distributes iterative training to each edge node; the central control unit collects the prediction results from all edge nodes and constructs a dynamic game theory model; when trigger control conditions are met, the central control unit issues step-limited adjustment commands to the relevant generator units to avoid excessively rapid adjustments; simulating the response of each generator unit using a digital twin model running synchronously with the physical system, and adjusting the control strategy and parameters based on simulation deviations; continuing to collect feedback data after executing control operations, and jointly updating the adaptive prediction model with the digital twin model through a federated learning mechanism to achieve closed-loop optimization of parameters.

[0019] As a preferred embodiment of the intelligent collaborative crossing method for unit vibration zones based on dynamic game theory and adaptive prediction described in this invention, the adaptive prediction model is trained independently on its respective edge nodes and then fused and optimized by the central control unit.

[0020] As a preferred embodiment of the intelligent collaborative crossing method for generator vibration zones based on dynamic game theory and adaptive prediction described in this invention, the dynamic game model is constructed based on dynamic game theory, and the game relationship is constructed with the vibration stability and power regulation cost of each generator unit as the payoff function, and the control responsibility of each generator unit in the regulation process is allocated based on the Shapley value.

[0021] The beneficial effects of this invention are as follows: This invention can significantly improve the safety and efficiency of vibration zone crossing in actual hydropower plant or smart microgrid environments. This invention enhances the generalization ability and data security of the model. Through a federated learning framework, it enables collaborative training of models across power plants without uploading original data, protecting data privacy. The federated mechanism can dynamically aggregate knowledge from multiple sites, improving adaptability to different regions, turbine models, and operating conditions. It supports long-term evolution and online self-learning. The invention introduces digital twin technology to achieve simulation verification and online optimization of control strategies. A virtual twin model of the generator unit is established to simulate possible outcomes before strategy execution. Real-time verification of the deviation between prediction and actual response forms a feedback loop. This achieves a safe control process of simulation before control, greatly reducing the risk of miscontrol. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of a scheme for an intelligent collaborative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction, provided as an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides an intelligent collaborative crossing system for generator vibration zones based on dynamic game theory and adaptive prediction, including: a generator monitoring unit, an edge computing unit, a communication unit, a central control unit, a digital twin simulation unit, and a federated learning training unit.

[0026] The system of this invention is deployed in modern power dispatching networks.

[0027] Each generator set (hydropower unit) is equipped with multiple sensors in the unit monitoring module to collect status parameters such as unit speed, power, and vibration acceleration in real time, and send the data to the local edge computing node and the central control unit.

[0028] Edge computing units deploy edge computing nodes next to each generator set to perform local data preprocessing and predictive calculations. These edge nodes utilize a hybrid LSTM+ARIMA prediction model in real time to accurately predict short-term vibration indicators and power output trends. The prediction model is trained using local historical data and incorporates global model parameters learned from other nodes through federated learning to continuously update the model weights.

[0029] The communication unit employs reliable industrial communication protocols (such as IEC 608705104 or IEC 61850) to construct a bidirectional communication network, connecting each edge node and the central control unit to achieve real-time transmission of monitoring data, prediction results, and control commands. The communication network supports both wired and wireless mesh network configurations, flexibly adapting to wide-area or distributed application scenarios.

[0030] The central control unit aggregates data and forecasts from each unit and executes a dynamic game algorithm.

[0031] The central control unit or backend server of the digital twin simulation unit establishes digital twin models of each unit, covering detailed mechanism and electromechanical system simulation. The real-time running digital twin operates in parallel with the actual units, enabling pre-verification of the effectiveness of control strategies or online parameter adjustment. According to literature reports, "digital twins of microgrids have become a key tool for predicting renewable energy output and load demand." In this invention, the digital twin module uses experimental data and physical models to improve the accuracy of prediction and control, further enhancing system robustness.

[0032] To address data privacy and communication bandwidth limitations, the federated learning training unit employs a federated learning approach to train the prediction model. Each edge computing node locally updates its LSTM+ARIMA model parameters using the most recent data, then uploads only the model gradients or parameters to the central server. The central server aggregates these parameters to generate a new global model, which is then distributed to all nodes. This process improves model generalization ability without sharing the original data, making it suitable for renewable energy scenarios such as wind power and photovoltaics where data is widely distributed.

[0033] Furthermore, several generator monitoring units are used to collect the status parameters of each unit in real time. Edge computing units are deployed one-to-one with the unit monitoring units and are responsible for running adaptive prediction models locally to accurately predict future vibration trends. The communication network is used to connect the edge computing units and the central control unit to realize the real-time exchange of data and instructions. The central control unit is used to establish a dynamic game model based on the prediction data of each unit, solve the Nash equilibrium, and generate coordinated control instructions for each unit.

[0034] Furthermore, the central control unit specifically includes the following steps A1-A4:

[0035] A1. Cooperative game modeling: Treating each unit as a player in a game, a non-zero-sum game model with cooperative elements is constructed. The utility function of each player includes a stability coefficient (such as the negative quantification of vibration amplitude) and an economic coefficient (such as power regulation cost), which comprehensively reflects the overall goal of cooperative crossing.

[0036] A2. Shapley apportionment calculation: Based on the above utility function, calculate the Shapley value of each participant to evaluate its marginal contribution to the total utility, thereby determining the role weight and game position of each unit in the coordinated control.

[0037] A3. Nash equilibrium solution: Apply dynamic non-cooperative game theory to find the optimal response of each unit at that moment, and obtain the Nash equilibrium solution and the corresponding output change scheme.

[0038] A4. Pulse-triggered control: Based on the calculated control increment, determine ΔP. i Check if the event triggering conditions are met. If so, issue small-step control commands to the corresponding units to achieve smooth output adjustment; otherwise, maintain the current state and wait for reassessment at the next moment.

[0039] Furthermore, the central control unit includes a game analysis module, a Shapley calculation module, and an optimization solution module.

[0040] The game analysis module is used to construct a game utility function with cooperative terms based on the predicted output and vibration index of each unit;

[0041] The Shapley calculation module is used to calculate the Shapley value for each participating unit. Represented as:

[0042]

[0043] in, Let N be the set of all participating units, S be the subset excluding the i-th unit, i be the i-th unit, f(S) be the collaboration benefit that the set of units S can achieve, and f(S∪{i}) be the collaboration benefit that the set of all participating units can achieve.

[0044] The optimization solution module is used to solve the Nash equilibrium of the game and determine the optimal output adjustment value for all units.

[0045] Furthermore, edge computing nodes are deployed inside or near each generator unit. Each node runs an adaptive prediction model, which integrates the advantages of Long Short-Term Memory (LSTM) networks and Auto-Regressive Integrated Moving Average (ARIMA) models. This allows for comprehensive analysis of nonlinear and linear time series characteristics. The dynamic game model uses the Shapley value of each unit as the payout mechanism. Through the hybrid prediction model of LSTM and ARIMA, the timing of when a unit is about to enter the vibration zone can be identified in advance. By utilizing Shapley value game allocation and Lagrange optimization adjustment strategies, multi-unit coordination and smooth transition are achieved, reducing the probability of strong vibrations. An event-triggered small-step control mechanism is introduced to further control the response rate and stability under sudden conditions.

[0046] Dynamic game theory and deep learning prediction are combined for hydropower unit crossing control; multiple units are modeled as independent game participants, avoiding traditional control strategies that rely heavily on manual settings; fair and interpretable allocation of regulation responsibilities is achieved through Nash equilibrium and marginal contribution calculation (Shapley value).

[0047] The formula for LSTM networks is expressed as follows:

[0048] f t =σ(W f [h t-1 ,x t ]+b f )

[0049] i t =σ(W i [h t-1 ,x t ]+b i )

[0050]

[0051] o t =σ(W o [h t-1 ,x t ]+b o )

[0052] h t =o t ⊙tanh(C t )

[0053] Where, x t Given the current input (such as load, wind speed, irradiance, and other multi-dimensional features), h t-1C is the output of the hidden layer at the previous time step. t As the current state, f t For the output of the forget gate, i t For the output of the input gate, o t W is the output of the output gate. f For the weight of the forget gate, b f For the bias of the forget gate, h t W represents the output of the hidden layer at the current moment. i b represents the weights of the input gate. i σ is the bias of the input gate, and σ is the sigmoid activation function. W is the new state information to be written after the candidate state undergoes a nonlinear transformation. c b represents the weight of the candidate state. c W is the bias for the candidate state. o b represents the weight of the output gate. o This is the bias of the output gate.

[0054] Meanwhile, the ARIMA model performs linear modeling of time series through differencing, AR, and MA components. Its general form after differencing is:

[0055] Δ d Y t =φ1Y t-1 +…+φ p Y t-p +ε t +θ1ε t-1 +…+θ q ε t-q

[0056] Where, Δ d Y t For a stationary sequence after d differencing, φ1...φ p The coefficients of the autoregressive term are θ1...θ2. q ε is the coefficient of the moving average term. t ...ε t-q Y is the random error term. t-1 Let be the value of the t-1 time series, p be the order of the autoregressive term, and q be the order of the moving average term.

[0057] This invention uses LSTM to capture nonlinear dynamic features and ARIMA to capture the linear trend of the sequence. The outputs of the two are added or fused together as the final prediction result, thereby improving prediction accuracy and stability.

[0058] Furthermore, the communication network adopts a real-time bidirectional communication architecture, ensuring that the central control unit sends instructions to edge computing nodes and receives operational data and model parameters uploaded by each node. Simultaneously, a federated learning-based data collaboration mechanism is established between the node level and the central control unit. Each node trains its model locally and uploads its model parameters to the central control unit, which then aggregates these parameters to form a global model before distributing it.

[0059] Furthermore, the central control unit issues adjustment commands with limited step size to the relevant generator sets to avoid excessively rapid adjustments. To optimize the output change step size when the generator sets pass through the vibration zone, this invention introduces a Lagrangian function to construct an optimization model:

[0060]

[0061] in, To optimize the model, ΔP i Let α be the output adjustment range of the i-th unit. i Its cost coefficient, ΔP i -ΔP 总 Let λ be the desired total output change, and let λ be the Lagrange multiplier used to introduce this constraint. By optimizing the Lagrange function, the output increment of each unit can be reasonably allocated while ensuring that the total output change meets the requirements, thereby achieving a smooth transition during the ride-through process.

[0062] This invention designs an event-triggered mechanism based on state changes, which triggers a control pulse when the mechanical state or vibration index exceeds a threshold. Let t k Real-time monitoring error e i (t), the triggering condition can be expressed as:

[0063] ||e i (t)||≥γ i ||e i (t k )||

[0064] Where, γ i It uses a dynamic threshold. Control signals are sent only at critical moments, reducing communication and control frequency and effectively preventing instability caused by frequent switching.

[0065] First, each generator unit's monitoring module collects and transmits status data. Edge nodes use formulas to predict vibration; if the prediction indicates potential oscillation risk, game negotiation is initiated. The central control module constructs a multi-unit game model based on the current status and prediction results, determines the output adjustment commands for each unit by solving algorithms including the Shapley value, and then issues them to the corresponding units for execution via an event-triggered mechanism. This process iterates continuously, adjusting in real time until the vibration zone is safely passed. Throughout the process, the digital twin continuously updates and feeds back simulation results, and federated learning continuously optimizes the prediction model.

[0066] Furthermore, this system is applicable to various complex power system scenarios, not limited to traditional hydropower plants. In new energy microgrids, the generating units may include wind turbines, photovoltaic arrays, and energy storage systems, with similar coordination methods as described above. In energy storage systems, energy storage units can be considered as adjustable generating units participating in a game. In distribution networks or microgrids, distributed collaborative control is achieved through locally deployed edge nodes. Through the above design, this invention achieves deep integration of generating unit control strategies, system architecture, and algorithm models, significantly enhancing the feasibility of cross-domain applications and system stability.

[0067] Example 2, an embodiment of the present invention, provides an intelligent cooperative vibration zone crossing system for generator units based on dynamic game theory and adaptive prediction, comprising:

[0068] S1. Deploy sensors on each generator set to collect the unit's operating status data and send the data to the edge node and the central control unit.

[0069] Sensors are deployed on each generator set to collect key operating parameters of the unit in real time, such as power, speed, and vibration acceleration; the collected data is sent to the local edge unit and the central control unit through the communication unit.

[0070] S2. The edge node preprocesses the locally collected data and uses the locally deployed adaptive prediction model to predict whether the generator set will enter the vibration zone within a predetermined time window.

[0071] Each edge node preprocesses the operating data of the local unit; using the prediction model deployed locally, it determines whether the unit has a tendency to enter the vibration zone in the future; the prediction model integrates deep learning and time series modeling technology, taking into account both nonlinear and linear characteristics.

[0072] S3. Each edge node trains an adaptive prediction model locally and only uploads model parameters to the central control unit. The central control unit then merges the model parameters uploaded by each node to generate a unified model, which is then distributed to each edge node for iterative training.

[0073] All edge nodes train prediction models independently on their own; they only upload model parameters, not raw data, to ensure data security; the central control unit aggregates and merges the parameters to form a unified and optimized prediction model, which is then distributed to each node for continuous iterative updates.

[0074] S4. The central control unit collects the prediction results of all edge nodes and constructs a dynamic game model.

[0075] The central control unit collects the prediction results of all edge nodes; treats all units as participants and constructs a collaborative control model; and determines the responsibility and degree of adjustment of each unit in the collaborative traverse based on its contribution to the overall stability of the system.

[0076] S5. When the trigger control conditions are met, the central control unit sends a step-limited adjustment command to the relevant generator set to avoid adjusting too quickly.

[0077] Based on the current system status and the capabilities of each unit, a coordinated control strategy is generated. This strategy enables reasonable power adjustment and allocation, avoiding excessive or rapid adjustment of any single unit. The control strategy balances overall output targets with the controllability of individual units.

[0078] Determine whether the conditions for triggering control are met (e.g., predicting an imminent crossing of the vibration zone); if the conditions are met, send small-step adjustment commands to the relevant units; control signals are issued only when necessary to avoid frequent adjustments that could lead to system instability.

[0079] S6. Simulate the response of each generator set using a digital twin model that runs synchronously with the physical system, and adjust the control strategy and parameters based on the simulation deviation.

[0080] The synchronously running digital twin model simulates the control response of each unit; the simulation results are used to verify whether the current strategy is safe and feasible; if deviations are found, parameters can be fine-tuned or the strategy can be updated in real time to improve control accuracy.

[0081] S7. After executing the control operation, continue to collect feedback data, and update the adaptive prediction model in conjunction with the digital twin model through the federated learning mechanism to achieve closed-loop optimization of parameters.

[0082] After control is implemented, the system continuously collects feedback data; compares the actual results with the predicted results, and updates the model weights; the federated learning mechanism and the digital twin model are optimized simultaneously to achieve closed-loop regulation.

[0083] The adaptive prediction model is trained independently on its respective edge nodes and then fused and optimized by the central control unit. The dynamic game model is constructed based on dynamic game theory, using the vibration stability and power regulation cost of each generator unit as the payoff function to build the game relationship, and allocating the control responsibility of each unit in the regulation process based on the Shapley value.

[0084] This embodiment also provides an electronic device applicable to the intelligent cooperative traversal method of unit vibration zone based on dynamic game theory and adaptive prediction, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent cooperative traversal method of unit vibration zone based on dynamic game theory and adaptive prediction proposed in the above embodiment.

[0085] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the intelligent collaborative traversal method for unit vibration zones based on dynamic game theory and adaptive prediction as proposed in the above embodiment.

[0086] The storage medium proposed in this embodiment and the intelligent collaborative crossing method for unit vibration zone based on dynamic game theory and adaptive prediction proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0087] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0088] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A unit vibration zone intelligent cooperative crossing system based on dynamic game theory and adaptive prediction, characterized in that: This includes a unit monitoring unit, an edge computing unit, a communication unit, a central control unit, a digital twin simulation unit, and a federated learning training unit; Each generator set is equipped with multiple sensors to collect real-time status data and send it to the local edge computing unit and the central control unit. The edge computing unit includes edge computing nodes deployed next to each generator set for performing local data preprocessing and predictive computation; The communication unit includes a two-way communication network built using industrial communication protocols, connecting each edge module and the central control unit to realize the real-time transmission of monitoring data, prediction results and control commands; The central control unit includes aggregating data and forecasts from each unit and executing dynamic game theory algorithms; The digital twin simulation unit is used to build a simulation model for each generator set and run it in real time; The federated learning training unit includes training the prediction model using a federated learning approach.

2. The intelligent cooperative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction as described in claim 1, characterized in that: The edge computing unit is deployed in a one-to-one correspondence with the unit monitoring unit, and is responsible for running the adaptive prediction model locally to accurately predict future vibration trends; The communication network is used to connect the edge computing unit and the central control unit to enable real-time exchange of data and instructions; The central control unit is used to establish a dynamic game model based on the predicted data of each unit, solve the Nash equilibrium, and generate coordinated control commands for each unit.

3. The intelligent cooperative traversal system for unit vibration zones based on dynamic game theory and adaptive prediction as described in claim 2, characterized in that: The central control unit includes a game analysis module, a Shapley calculation module, and an optimization solution module.

4. The intelligent cooperative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction as described in claim 3, characterized in that: The game analysis module is used to construct a game utility function with cooperative terms based on the predicted output and vibration index of each unit. The Shapley calculation module is used to calculate the Shapley value for each participating unit. Represented as, in, Let N be the set of all participating units, S be the subset excluding the i-th unit, i be the i-th unit, f(S) be the collaboration benefit that the set of units S can achieve, and f(S∪{i}) be the collaboration benefit that the set of all participating units can achieve. The optimization solution module is used to solve the Nash equilibrium of the game and determine the optimal output adjustment value for all units.

5. The intelligent cooperative vibration zone crossing system for generator units based on dynamic game theory and adaptive prediction as described in claim 4, characterized in that: The adaptive prediction model integrates the advantages of long short-term memory networks and autoregressive integral moving average models, comprehensively analyzes the characteristics of nonlinear and linear time series, and uses the Shapley value of each unit as the revenue distribution mechanism in the dynamic game model.

6. The intelligent cooperative crossing system for unit vibration zones based on dynamic game theory and adaptive prediction as described in claim 5, characterized in that: The autoregressive integral moving average model uses differencing, AR, and MA components to linearly model the time series, and is expressed as follows: D d Y t =φ1Y t-1 +…+φ p Y t-p +e t +θ1ε t-1 +…+θ q e t-q Where, Δ d Y t For a stationary sequence after d differencing, φ1...φ p The coefficients of the autoregressive term are θ1...θ2. q ε is the coefficient of the moving average term. t ...ε t-q Y is the random error term. t-1 Let be the value of the t-1 time series, p be the order of the autoregressive term, and q be the order of the moving average term.

7. The intelligent cooperative vibration zone crossing system for generator units based on dynamic game theory and adaptive prediction as described in claim 6, characterized in that: The central control unit adopts an event-triggered small-step control strategy, setting a trigger threshold e for each unit i based on real-time error monitoring. i (t), when ||e i (t)||≥γ i ||e i (t k When ||, the i-th unit is triggered to perform a small-amplitude output pulse adjustment.

8. A method for intelligent cooperative crossing of generator vibration zones based on dynamic game theory and adaptive prediction, employing the intelligent cooperative crossing system for generator vibration zones based on dynamic game theory and adaptive prediction as described in any one of claims 1 to 7, characterized in that, include: Sensors are deployed on each generator set to collect the unit's operating status data and send the data to edge nodes and the central control unit; Edge nodes preprocess locally collected data and use locally deployed adaptive prediction models to predict whether the generator set will enter the vibration zone within a predetermined time window. Each edge node trains an adaptive prediction model locally and only uploads model parameters to the central control unit. The central control unit then merges the model parameters uploaded by each node to generate a unified model, which is then distributed to each edge node for iterative training. The central control unit collects the prediction results of all edge nodes and constructs a dynamic game model; When the trigger control conditions are met, the central control unit sends a step-limited adjustment command to the relevant generator set to avoid adjusting too quickly. The response of each generator unit is simulated using a digital twin model that runs synchronously with the physical system, and the control strategy and parameters are adjusted based on the simulation deviation. After executing control operations, feedback data continues to be collected, and the adaptive prediction model is jointly updated through a federated learning mechanism and a digital twin model to achieve closed-loop optimization of parameters.

9. The intelligent cooperative crossing method for unit vibration zones based on dynamic game theory and adaptive prediction as described in claim 8, characterized in that: The adaptive prediction models are trained independently on their respective edge nodes and then fused and optimized by the central control unit.

10. The intelligent cooperative crossing method for unit vibration zones based on dynamic game theory and adaptive prediction as described in claim 9, characterized in that: The dynamic game model is constructed based on dynamic game theory. It uses the vibration stability and power regulation cost of each generator unit as the payoff function to build the game relationship, and allocates the control responsibility of each unit in the regulation process based on the Shapley value.

Images