Electric vehicle and renewable energy source integrated energy management measurement and control system and method

Abstract

The invention relates to the technical field of control systems, and particularly discloses an electric vehicle and renewable energy source integrated energy management measurement and control system and method, and the system comprises an acquisition module, an optimization control module, a control execution module, a constraint module, and a fault tolerance module. The acquisition module acquires photovoltaic, wind power and electric vehicle load data through a distributed sensor network, and the control execution module issues an instruction through a multi-priority message queue and performs closed-loop feedback; the constraint module decouples space-time constraints through causal derivation, and updates strategy network parameters in combination with federated learning; and the fault-tolerant module performs dynamic fault tolerance based on the Bayesian network and the generative adversarial network. The method covers multi-source data processing, hierarchical optimization instruction generation, closed-loop verification and anti-interference training processes, solves the problem of optimization path deviation caused by difficult solution of high-dimensional mixed integer programming and time-space correlation interference, and realizes dynamic balance of energy supply and demand and low-carbon operation.

Description

Technical Field

[0001] The present invention relates to the technical field of control systems, and particularly to an energy management measurement and control system and method for integrating electric vehicles and renewable energy. Background Art

[0002] The energy management measurement and control system for integrating electric vehicles and renewable energy realizes the efficient regulation of distributed energy, energy storage units, and flexible loads through multi-level collaborative optimization technology. Based on a real-time acquisition module and a wide-area communication network, this system dynamically obtains the output characteristics of intermittent power sources such as photovoltaic and wind power and the load demands of electric vehicle clusters, and uses stochastic optimization theory to establish a mixed-integer programming model including time-coupling constraints. By using the model predictive control (MPC) algorithm to roll-correct the renewable energy prediction error, combined with the fast power response characteristics of the bi-directional current conversion device of electric vehicles, a multi-objective optimization function including electricity price signals and carbon emission factors is constructed to generate energy storage charge and discharge scheduling strategies and electric vehicle orderly charging instructions on a minute-level time scale. This process needs to satisfy the node voltage constraints of the distribution network and the transformer capacity limit. At the same time, the deep reinforcement learning algorithm is used to extract features from historical operation data, gradually improving the system's adaptability to the uncertainties on both the source and load sides, and finally achieving the goals of dynamic balance of energy supply and demand, optimal economy, and low-carbon operation.

[0003] When the existing electric vehicle control and regulation system deals with the strong randomness and multi-time-scale coupling characteristics on both the source and load sides, the high-dimensional non-linear constraints of the mixed-integer programming model make it difficult for the solver to converge to the feasible region within a limited time window. Especially when the fluctuations of renewable energy output and the sudden changes in electric vehicle charging demands form spatio-temporal correlation interference, the traditional rolling optimization framework is prone to falling into local optimal solutions, resulting in the mismatch between the control instruction sequence and the real-time working conditions, thereby weakening the dynamic regulation robustness of the system. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides an energy management measurement and control system and method for integrating electric vehicles and renewable energy, which are used to solve the problem of deviation of the rolling optimization path caused by the difficulty in solving the high-dimensional mixed-integer programming model and spatio-temporal correlation interference in the existing control and regulation system.

[0005] To solve the above technical problems, the specific technical solutions of the present invention are as follows: In a first aspect, an energy management measurement and control system for integrating electric vehicles and renewable energy includes: An acquisition module, configured to collect photovoltaic output data, wind power output data, and electric vehicle load demand data through a distributed sensor network, filter outlier points in the data by using an anomaly detection algorithm, perform time series compensation on multi-source data based on a time series alignment algorithm, and output synchronized time series data to an optimization control module; An optimization control module, configured to receive the synchronous timing data output by the acquisition module, and generate optimization instructions by layering according to long-term, medium-term, and short-term time scales: The long-term optimization layer generates an uncertainty scenario tree of renewable energy output and load demand based on the synchronous timing data, dynamically corrects the energy storage capacity threshold in combination with the energy storage life loss model, and inputs the corrected threshold into the medium-term optimization layer; The medium-term optimization layer embeds a robust optimization model in the model predictive control framework, generates rolling optimization instructions according to the corrected energy storage capacity threshold and the real-time electricity price signal, and transmits the instructions to the short-term optimization layer; The short-term optimization layer uses a distributed solver to decompose the mixed-integer programming problem in the rolling optimization instructions, and generates charge and discharge instructions based on the priority of the distribution network node voltage constraint; A control execution module, configured to receive the charge and discharge instructions generated by the short-term optimization layer, issue instructions to the electric vehicle bidirectional converter device through a multi-priority message queue, dynamically adjust the execution threshold based on the stability control theory, and feedback the execution result to the constraint module; A constraint module, configured to update the policy network parameters in the optimization control module through a dual experience replay mechanism, decouple the spatio-temporal correlation constraints of the charge and discharge instructions in combination with causal inference technology, and input the decoupled constraints into the distributed solver of the short-term optimization layer; A fault tolerance module, configured to simulate extreme condition data through an adversarial sample generation network, inject it into the policy network training data set of the optimization control module, analyze the root cause of the instruction conflict of the control execution module based on a Bayesian network, and dynamically adjust the service priority of the medium-term optimization layer.

[0006] Furthermore, for the energy management measurement and control system integrating electric vehicles and renewable energy of the present invention, the long-term optimization layer generates an uncertainty scenario tree of renewable energy output and load demand through the Monte Carlo sampling method, and dynamically corrects the charge and discharge capacity threshold of the energy storage unit based on the energy storage life loss model; The medium-term optimization layer receives the scenario tree generated by the long-term optimization layer, dynamically allocates multi-objective optimization weights in combination with the entropy weight method, generates rolling optimization instructions including electricity price signals and carbon emission factors, and transmits the instructions to the short-term optimization layer; The short-term optimization layer decomposes the mixed-integer programming problem into parallel subtasks through a topological sorting algorithm based on the priority of the distribution network node voltage constraint according to the rolling optimization instructions transmitted by the medium-term optimization layer, and allocates the subtasks to the distributed solver for parallel computing.

[0007] Furthermore, for the energy management measurement and control system integrating electric vehicles and renewable energy of the present invention, the constraint module includes: A federated learning unit is used to encrypt the electric vehicle load data on the edge side through differential privacy technology, upload the encrypted local model parameters to the blockchain node for distributed evidence storage, and generate global model parameters through federated aggregation, and send them to the policy network of the optimization control module; A constraint decomposition unit receives the global model parameters generated by the federated learning unit, smooths the boundary of the feasible region of the mixed-integer programming problem based on the hyperplane projection algorithm, generates dynamic constraint conditions and inputs them into the distributed solver of the short-term optimization layer.

[0008] Further, for the energy management measurement and control system integrating electric vehicles and renewable energy of the present invention, the fault tolerance module includes: A fault tolerance training unit is used to train the policy network of the optimization control module in stages through a curriculum learning strategy, inject the extreme working condition data generated by the generative adversarial network into the training data set of the double experience replay mechanism of the optimization control module, so as to improve the anti-interference ability of the policy network; A service orchestration unit dynamically adjusts the service priority of the optimization level according to the resource availability evaluation result. When the resource availability is lower than the preset threshold, it shields the scenario tree generation service of the long-term optimization layer and sends a downgrade instruction to the medium-term optimization layer to switch to a simplified rolling optimization model.

[0009] Further, for the energy management measurement and control system integrating electric vehicles and renewable energy of the present invention, the acquisition module includes: A spectrum analysis unit is used to perform spectrum analysis on the photovoltaic output data, identify the main frequency characteristics of the periodic noise components, and dynamically match the basis function type and decomposition layer number of the wavelet denoising algorithm according to the main frequency characteristics to generate denoised photovoltaic output data; A time series alignment unit receives the denoised photovoltaic output data and the original wind power output data generated by the spectrum analysis unit, calibrates the high-frequency collected electric vehicle load data and the low-frequency updated renewable energy prediction data through a timestamp-matched cache queue, and generates synchronized time series data and outputs it to the long-term optimization layer and the medium-term optimization layer of the optimization control module.

[0010] Further, for the energy management measurement and control system integrating electric vehicles and renewable energy of the present invention, the closed-loop verification unit of the control execution module models the dependency relationship of the charge and discharge instructions through a directed acyclic graph. When an instruction conflict is detected, it transmits the characteristic parameters of the conflict instruction to the Bayesian network inference unit of the fault tolerance module to locate the root cause of the conflict and generate an adjustment instruction; The instruction issuing unit retransmits the invalid instructions according to the verification failure result, and feeds back the verification log including the conflict type and the retransmission times to the policy distillation unit of the constraint module, which is used to update the training priority of the policy network in the optimization control module.

[0011] Further, for the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the medium-term optimization layer receives the energy storage capacity threshold generated by the long-term optimization layer through a rolling time domain window, combines real-time electricity price signals to generate charging and discharging power instructions at a minute-level time scale, and transmits the instructions to the short-term optimization layer; The short-term optimization layer calculates the electric vehicle cluster scheduling scheme that satisfies the distribution network node voltage constraints through the distributed solver according to the received charging and discharging power instructions, marks the instructions exceeding the voltage safety threshold, and inputs the marked out-of-limit instructions into the constraint decomposition unit of the constraint module to adjust the dynamic constraint conditions.

[0012] Further, for the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the policy learning unit of the constraint module migrates the constraint parsing rules of the complex policy network in the optimization control module to a lightweight model through knowledge transfer technology, and deploys the migrated model parameters to the edge-side controller; The federated learning unit of the constraint module detects parameter conflicts of the lightweight model according to the parameter version information stored in the blockchain, triggers a parameter rollback operation to restore historical valid parameters, and synchronizes the rollback instructions to all edge nodes to update the local model parameters.

[0013] Further, for the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the consistency unit of the fault tolerance module smooths the parameter update sequence of the policy network in the optimization control module through a parameter smoothing algorithm, and inputs the processed parameters into the causal inference unit of the constraint module, which is used to decouple the spatio-temporal correlation constraints in the charging and discharging instructions; The fault tolerance training unit of the fault tolerance module dynamically adjusts the difficulty gradient of the curriculum learning strategy according to the simulation data of historical extreme working conditions, and feeds back the gradient adjustment signal to the generative adversarial network to optimize the generation parameters of adversarial samples.

[0014] In a second aspect, an energy management measurement and control method for integrating electric vehicles and renewable energy provided by the present invention is applied to the energy management measurement and control system for integrating electric vehicles and renewable energy, and includes: Collect photovoltaic output data, wind power output data and electric vehicle load demand data through the acquisition module, filter outliers using the isolation forest algorithm and compensate the timing deviation based on the dynamic time warping algorithm to generate synchronous timing data; By optimizing the long-term optimization layer of the control module, an uncertainty scenario tree of renewable energy output and load demand is generated based on the synchronous timing data, and the energy storage capacity threshold is dynamically corrected in combination with the energy storage life loss model; Through the medium-term optimization layer of the optimization control module, the corrected energy storage capacity threshold of the long-term optimization layer is received, a robust optimization layer is embedded in the model predictive control framework, and a rolling optimization instruction is generated in combination with the real-time electricity price signal; Through the short-term optimization layer of the optimization control module, the distributed solver is used to decompose the mixed integer programming problem in the rolling optimization instruction, and the charging and discharging instructions are generated by preferentially processing the distribution network node voltage constraints; The charging and discharging instructions are sent to the electric vehicle bidirectional converter device through the control execution module, the execution threshold is dynamically adjusted based on the Lyapunov stability theory, and the execution result is fed back to the constraint module; The policy network parameters of the optimization control module are updated through the double experience replay mechanism of the constraint module, the spatio-temporal correlation constraints in the charging and discharging instructions are decoupled by using causal reasoning technology, and the decoupled constraint conditions are input into the distributed solver of the short-term optimization layer; The generation adversarial network of the fault tolerance module simulates extreme working conditions to generate adversarial samples and injects them into the training data set of the policy network. The root cause of the instruction conflict is inferred based on the Bayesian network, and the service priority of the medium-term optimization layer is dynamically adjusted.

[0015] Advantages of the present invention: The advantages of the present invention lie in significantly reducing the solution complexity of the high-dimensional mixed integer programming model through the multi-time scale hierarchical optimization architecture and the dynamic constraint decomposition technology. The long-term optimization layer compresses the solution space dimension based on the uncertainty scenario tree, the medium-term optimization layer uses robust optimization and dynamic weight allocation to generate anti-interference instructions, and the short-term optimization layer parallelly processes the voltage constraint problem through the distributed solver to improve the solution efficiency; combining causal reasoning and federated learning mechanisms to decouple spatio-temporal correlation constraints, suppressing path deviation caused by new energy output fluctuations and load mutations, the closed-loop verification module dynamically corrects instruction conflicts through directed acyclic graphs and Bayesian networks, the fault tolerance module adopts a curriculum learning strategy and adversarial sample training to enhance the system's anti-interference ability, and the time series alignment and noise suppression technology of the acquisition module ensure the consistency of multi-source data, ultimately achieving dynamic balance of energy supply and demand, robust execution of optimization instructions, and efficient cross-layer resource coordination. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, other drawings can be obtained according to the drawings without creative efforts.

[0017] Figure 1 This is a flowchart of the energy management measurement and control method for integrating electric vehicles and renewable energy provided by an embodiment of the present invention. Detailed implementation manners

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. The technical solutions provided by each embodiment of the present invention will be described in detail below with reference to the drawings.

[0019] To better understand the objectives of the present invention, the present invention will be further described in detail below.

[0020] In a first aspect, the energy management measurement and control system for integrating electric vehicles and renewable energy provided by the present invention includes: A collection module, configured to collect photovoltaic output data, wind power output data, and electric vehicle load demand data through a distributed sensor network, filter outliers in the data using the Isolation Forest algorithm, perform time series alignment on multi-source data based on the Dynamic Time Warping algorithm, and output the processed multi-source data to the optimization control module; An optimization control module, configured to receive the multi-source data output by the collection module and generate optimization instructions at different time scales: The long-term optimization layer generates an uncertainty scenario tree for renewable energy output and load demand based on the multi-source data, corrects the energy storage capacity threshold based on the energy storage life loss model, and outputs it to the medium-term optimization layer; The medium-term optimization layer generates rolling optimization instructions through a model predictive control framework embedded with a robust optimization layer, combines the energy storage capacity threshold and real-time electricity price signals, and outputs them to the short-term optimization layer; The short-term optimization layer uses a distributed alternating direction method of multipliers solver to decompose the mixed integer programming problem in the rolling optimization instructions, and preferentially processes the distribution network node voltage constraints to generate charge and discharge instructions; A control execution module, configured to receive the charge and discharge instructions generated by the short-term optimization layer, send instructions to the electric vehicle bidirectional converter through a multi-priority message queue, dynamically adjust the instruction execution threshold based on the Lyapunov stability theory, and feedback the execution result to the constraint module; A constraint module, configured to update the policy network parameters in the optimization control module through a dual experience replay mechanism, decouple the spatio-temporal correlation constraints in the charge and discharge instructions using causal inference technology, and input the constraint decomposition result into the distributed solver of the short-term optimization layer; A fault-tolerant module, which is used to simulate extreme working conditions through a generative adversarial network and inject them into the training data set of the policy network, infer the root cause of instruction conflicts in the control execution module based on a Bayesian network, and dynamically degrade the service priority of the medium-term optimization layer of the optimization control module.

[0021] The energy management measurement and control system integrating electric vehicles and renewable energy provided by the present invention collects photovoltaic output data, wind power output data and electric vehicle load demand data in real time through a distributed sensor network, uses the isolation forest algorithm to detect and filter outliers in the collected data, and compensates and aligns the time series deviation of multi-source data based on the dynamic time warping algorithm to form synchronous time series data and transmit it to the optimization control module. The optimization control module processes data on a time scale. The long-term optimization layer uses Monte Carlo sampling to generate an uncertainty scenario tree of renewable energy output and load demand, dynamically corrects the capacity threshold of the energy storage unit in combination with the energy storage life loss model, and outputs it to the medium-term optimization layer; the medium-term optimization layer embeds a robust optimization layer in the model predictive control framework, generates rolling optimization instructions based on the real-time electricity price signal and the corrected energy storage capacity threshold, and transmits them to the short-term optimization layer; the short-term optimization layer uses a distributed alternating direction multiplier method solver to decompose the mixed integer programming problem, and preferentially processes the distribution network node voltage constraint to generate charge and discharge instructions.

[0022] After receiving the charge and discharge instructions output by the short-term optimization layer, the control execution module distributes the instructions to the electric vehicle bidirectional converter device through a multi-priority message queue, and at the same time dynamically adjusts the instruction execution threshold based on the Lyapunov stability theory to achieve the stability of the charge and discharge process. The feedback data generated during the execution process is transmitted to the constraint module, which updates the parameters of the policy network in the optimization control module through a double experience replay mechanism, decouples the spatio-temporal correlation constraints in the charge and discharge instructions using causal inference technology, and inputs the decomposed dynamic constraint conditions into the distributed solver of the short-term optimization layer to optimize subsequent instruction generation. The fault-tolerant module simulates extreme working condition data through a generative adversarial network, injects it into the training data set of the policy network to enhance the model robustness, and at the same time infers the root cause of instruction conflicts in the control execution module based on a Bayesian network, and dynamically adjusts the service priority of the medium-term optimization layer to cope with sudden disturbances.

[0023] The acquisition module further includes an edge - side fast Fourier transform unit, which is used to identify the periodic noise components of the photovoltaic output data, and dynamically adjust the decomposition level of the wavelet denoising algorithm based on the noise spectrum characteristics to improve the data pre - processing accuracy. The time - series alignment unit calibrates the electric vehicle load data collected at high frequency and the renewable energy prediction data updated at low frequency through a cache queue to achieve the synchronization of multi - source data. The medium - term optimization layer of the optimization control module combines the energy storage capacity threshold output by the long - term optimization layer and the real - time electricity price signal within a rolling time - domain window to generate charge - discharge power instructions at a minute - level time scale. The short - term optimization layer calculates the electric vehicle cluster scheduling scheme that meets the node voltage constraints according to this instruction, and marks the out - of - limit instructions and inputs them to the constraint decomposition unit for dynamic constraint adjustment.

[0024] The constraint module migrates the knowledge of the complex policy network to a lightweight model through policy distillation technology and deploys it to the edge - side controller to reduce the computational load. The federated learning unit uses differential privacy technology to protect the edge data, realizes the traceability of model parameters through blockchain evidence storage, and triggers a rollback mechanism to synchronize all edge nodes when parameter conflicts are detected. The consistency unit of the fault - tolerance module uses an exponentially weighted moving average algorithm to smooth the parameter mutations of the policy network, realizes the stability of the optimization process, and adjusts the difficulty gradient of the curriculum learning strategy according to the simulation results of historical extreme working conditions to optimize the adversarial sample generation logic to improve the anti - interference ability of the system.

[0025] Specifically, for the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the long - term optimization layer generates an uncertainty scenario tree of renewable energy output and load demand through Monte Carlo sampling, and dynamically adjusts the charge - discharge capacity threshold of the energy storage unit based on the energy storage life loss model; The medium - term optimization layer inputs the scenario tree into a multi - objective weight allocation model dynamically calculated by the entropy weight method, generates a rolling optimization instruction including an electricity price signal and a carbon emission factor, and transmits it to the short - term optimization layer; The short - term optimization layer, according to the rolling optimization instruction, preferentially processes the node voltage constraints of the distribution network through topological sorting, and distributes the decomposed sub - problems to the distributed alternating direction multiplier method solver for parallel calculation.

[0026] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the long-term optimization layer simulates the uncertainty distributions of renewable energy output and load demand through the Monte Carlo sampling method, and generates an uncertainty scenario tree containing the power fluctuation characteristics of multiple time sections. The long-term optimization layer dynamically adjusts the charge and discharge capacity thresholds of the energy storage unit based on the mapping relationship between the charge and discharge cycle times and the capacity attenuation rate in the energy storage life loss model, avoids the impact of overcharging and over-discharging on the life of the energy storage device, and outputs the corrected capacity thresholds to the medium-term optimization layer. The medium-term optimization layer receives the capacity thresholds and real-time electricity price signals, calculates the dynamic weights of the economic objective and the low-carbon objective in the rolling optimization through the entropy weight method, constructs a multi-objective optimization function containing the electricity price sensitivity coefficient and the carbon emission factor, combines with the robust optimization layer to generate the rolling optimization instruction of the adversarial prediction error, and transmits it to the short-term optimization layer as the constraint boundary.

[0027] After receiving the rolling optimization instruction output by the medium-term optimization layer, the short-term optimization layer identifies the voltage over-limit risk nodes based on the distribution network topology structure, and preferentially processes the charge and discharge power distribution problem in the voltage constraint sensitive area through the topological sorting algorithm. The short-term optimization layer decomposes the mixed integer programming problem into multiple sub-problems, uses the distributed alternating direction multiplier method solver to perform parallel calculations on the sub-problems, synchronizes the optimization results of each sub-problem through the consistency coordination mechanism, and generates the charge and discharge instructions of the electric vehicle cluster that meet the node voltage constraints. The distributed solver marks the over-limit instruction as a dynamic constraint adjustment signal and feeds it back to the constraint decomposition unit to update the feasible region boundary in the short-term optimization model. A closed-loop optimization link is formed between the optimization levels through parameter transfer and constraint inheritance, realizing the step-by-step refinement and dynamic coordination of multi-time scale instructions.

[0028] Specifically, for the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the constraint module includes: a federated learning unit, which is used to protect the electric vehicle load data on the edge side through differential privacy technology, upload the encrypted local model parameters to the blockchain node for storage, and send the globally aggregated model parameters to the policy network in the optimization control module; A constraint decomposition unit, according to the globally aggregated model parameters updated by the federated learning unit, smooths the feasible region boundary through the hyperplane projection algorithm, generates dynamic constraint conditions and inputs them to the distributed alternating direction multiplier method solver of the short-term optimization layer.

[0029] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the federated learning unit adds noise and perturbs the electric vehicle load data collected at the edge side through differential privacy technology to generate a local data set that meets the privacy protection requirements. The local data set is input into the edge side controller to train the policy network sub-model, generate encrypted model parameters and attach time stamp and device identification information, and upload them to the distributed ledger node through the blockchain smart contract for evidence storage. The federated learning unit aggregates the encrypted model parameters of multiple edge nodes, uses the homomorphic encryption algorithm for parameter fusion and weight update, generates global model parameters and distributes them to the policy network of the optimization control module to achieve collaborative model optimization across edge nodes.

[0030] After receiving the updated global model parameters from the federated learning unit, the constraint decomposition unit analyzes the spatio-temporal correlation constraint conditions output by the policy network, and smooths the boundary of the feasible region of the mixed integer programming problem based on the hyperplane projection algorithm. The constraint decomposition unit identifies the coupling relationship between the voltage constraint and the power balance constraint according to the current distribution network topology and real-time load distribution, decomposes the complex constraints into independent sub-constraint sets and generates dynamic adjustment coefficients, and inputs them into the distributed alternating direction multiplier method solver of the short-term optimization layer. The distributed solver adjusts the iteration step size and penalty factor according to the dynamic constraint conditions, synchronously updates the constraint boundaries of each sub-problem during the parallel calculation process, reduces the dimension of the optimization model and improves the solution efficiency.

[0031] The federated learning unit and the constraint decomposition unit form a closed-loop feedback mechanism through parameter transfer. The update of the global model parameters triggers the constraint decomposition unit to re-evaluate the boundary of the feasible region, and the adjustment result of the dynamic constraint conditions further affects the training data distribution of the policy network. The model version information stored in the blockchain provides historical reference parameters for constraint decomposition. When parameter conflicts are detected, the federated learning unit calls the smart contract to verify the model version consistency and triggers a parameter rollback operation to maintain the stability of the optimization process. After receiving the rollback instruction, the edge node extracts the historical valid parameters from the blockchain distributed ledger to replace the abnormal parameters to ensure the reliability of the input data of the constraint decomposition unit.

[0032] Specifically, in the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the fault tolerance module includes: A fault tolerance training unit for gradually improving the anti-interference ability of the policy network in the optimization control module through a curriculum learning strategy, and injecting the extreme working condition data simulated by the generative adversarial network into the training data set of the double experience replay mechanism; A service orchestration unit that dynamically shields the services of the long-term optimization layer according to the resource availability evaluation result, and transmits a degradation instruction to the medium-term optimization layer to adjust the generation logic of the rolling optimization instruction.

[0033] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the fault-tolerant training unit constructs a training scenario library in stages based on the curriculum learning strategy. In the initial stage, historical normal operating condition data is loaded to train the policy network, and extreme power fluctuations and load mutation scenarios simulated by the generative adversarial network are gradually introduced to improve the adaptability of the policy network to abnormal operating conditions. The generator of the generative adversarial network reconstructs extreme event data such as grid voltage dips and sudden drops in new energy output through a spatio-temporal feature extraction module. The discriminator combines the historical real data distribution characteristics in the double experience replay mechanism, dynamically adjusts the credibility threshold of the generated data, and injects the qualified data into the policy network training data set to enhance the robustness of the model.

[0034] The service orchestration unit monitors the computing resource utilization rate and communication link status in real time, evaluates the computing load of the long-term optimization layer and the real-time requirements of the medium-term optimization layer through the analytic hierarchy process. When the resource availability is lower than the preset threshold, it dynamically shields the scenario tree generation service of the long-term optimization layer and sends a downgrade instruction to the medium-term optimization layer synchronously. After receiving the downgrade instruction, the medium-term optimization layer switches to a simplified model predictive control framework, adopts a fixed-weight multi-objective optimization model to replace the dynamic weight allocation mechanism, shortens the generation cycle of the rolling optimization instruction, and maintains the instruction issuance frequency at the minute-level time scale.

[0035] The fault-tolerant training unit and the service orchestration unit achieve collaborative operation through a data bus. The extreme operating condition data simulated by the generative adversarial network triggers changes in the resource availability evaluation index, and the service orchestration unit dynamically adjusts the stage switching speed of the curriculum learning strategy according to the complexity of the operating conditions. After the anti-interference ability of the policy network is improved, the service orchestration unit gradually restores some service functions of the long-term optimization layer, and fuses the simplified model parameters into the complete optimization framework through a progressive weight migration method to achieve a smooth transition of the service level. The blockchain node records the operation logs during service downgrade and recovery, provides state migration data with time series marks for policy network training, and optimizes the data sampling strategy of the double experience replay mechanism.

[0036] Specifically, in the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the acquisition module includes: An edge-side fast Fourier transform unit for identifying the periodic noise components of photovoltaic output data and dynamically adjusting the decomposition layer number of the wavelet denoising algorithm according to the noise spectrum characteristics to preprocess the photovoltaic output data; A time series alignment unit that calibrates the electric vehicle load data collected at high frequency and the renewable energy prediction data updated at low frequency through a cache queue, and outputs synchronized time series data to the long-term optimization layer and the medium-term optimization layer of the optimization control module.

[0037] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the edge-side fast Fourier transform unit performs spectrum analysis on the photovoltaic output data, and extracts periodic noise characteristics through fundamental frequency detection and harmonic component identification. According to the amplitude and distribution characteristics of the main frequency components in the noise spectrum, the type of basis function of the wavelet denoising algorithm is dynamically matched, and the decomposition layer is adaptively adjusted to separate the noise frequency band, and the low-frequency trend component of the effective signal of the photovoltaic output is retained. The denoised photovoltaic output data is merged with the original wind power output data and input into the time series alignment unit for multi-source data fusion.

[0038] For the electric vehicle load data collected at high frequency and the renewable energy prediction data updated at low frequency, the time series alignment unit establishes a cache queue mechanism based on timestamps. The high-frequency data is sliced according to a fixed time window and temporarily stored in the queue. When the low-frequency data arrives, it triggers the matching and interpolation calculation of the corresponding time segment in the queue, fills the missing time point data through the linear weighted compensation algorithm, and generates a synchronized time series data set. The synchronized data is resampled according to the time scale requirements of the long-term optimization layer and the medium-term optimization layer. The long-term optimization layer receives the trend data at the hourly granularity for scenario tree construction, and the medium-term optimization layer obtains the detailed data at the minute granularity to support rolling optimization.

[0039] The edge-side fast Fourier transform unit and the time series alignment unit form a series processing link. The denoised photovoltaic data output by the Fourier transform unit carries noise marking information. The time series alignment unit performs weighted downweighting processing on the abnormal periods in the high-frequency load data according to the marking information to suppress the interference of noise propagation on the optimization control module. The feedback signal of the optimization control module dynamically adjusts the spectrum analysis window length of the Fourier transform unit. When it is detected that the new energy output fluctuation intensifies, the analysis window is shortened to improve the timeliness of noise recognition, and the interpolation calculation frequency of the time series alignment unit is synchronously optimized to maintain the time alignment accuracy of multi-source data.

[0040] Specifically, in the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the closed-loop verification unit of the control execution module describes the dependency relationship of the charge and discharge instructions through a directed acyclic graph, and triggers the Bayesian network inference unit of the fault tolerance module to locate the root cause of the conflict when detecting an instruction conflict; The instruction issuing unit retransmits the failed instructions according to the cyclic redundancy check result, and feeds back the check log to the policy distillation unit of the constraint module to optimize the policy network parameters.

[0041] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the closed-loop verification unit establishes a dependency relationship model of charge and discharge instructions based on a directed acyclic graph, maps each charge and discharge instruction to a graph node, and maps the timing constraints and resource competition relationships between instructions to directed edges. The closed-loop verification unit monitors the execution status of the nodes in the graph in real time. When it detects resource preemption or time window overlap between nodes, it marks it as an instruction conflict event and triggers the Bayesian network inference unit of the fault tolerance module. The Bayesian network receives the priority labels of the conflicting instructions, the current resource occupancy rate, and historical conflict data, and determines through probability inference that the root cause of the conflict is abnormal resource allocation strategy or overly tight time scale coupling, and generates an instruction priority adjustment plan or a resource release instruction.

[0042] The instruction issuing unit embeds a cyclic redundancy check code when transmitting charge and discharge instructions. After the receiving end fails the verification, it locates the physical link node of the invalid instruction based on the instruction timestamp and device identification information, and preferentially retransmits the high-priority voltage overlimit correction instruction according to the level division of the multi-priority message queue. The verification log records the content, failure type, and retransmission times of the invalid instruction, and transmits it to the policy distillation unit of the constraint module after attaching the timestamp and environmental parameters. The policy distillation unit analyzes the conflict pattern features in the log, extracts the spatio-temporal distribution law of high-frequency invalid instructions, adjusts the sampling weights of historical data in the double experience replay mechanism, and optimizes the training data distribution of the lightweight policy network.

[0043] The closed-loop verification unit and the policy distillation unit form a dynamic optimization mechanism through data interaction. The conflict root cause analysis result output by the Bayesian network is encoded as a constraint adjustment signal and input into the parameter update process of the policy distillation unit to guide the lightweight model to enhance its generalization ability for specific conflict scenarios. After the optimized policy network parameters are deployed to the edge controller, new charge and discharge instruction dependency relationship data is generated, and the structural features of the directed acyclic graph are updated in reverse to reduce the triggering frequency of similar conflict events. The fault tolerance module synchronously receives the update log of the policy network and dynamically adjusts the extreme working condition simulation parameters of the generative adversarial network to make the adversarial samples cover the instruction generation mode optimized by the policy network.

[0044] Specifically, in the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the medium-term optimization layer of the optimization control module receives the energy storage capacity threshold generated by the long-term optimization layer through a rolling time domain window, and generates a charge and discharge power instruction at a minute-level time scale in combination with the real-time electricity price signal; The short-term optimization layer calculates an electric vehicle cluster scheduling scheme that satisfies the node voltage constraint through the distributed alternating direction multiplier method solver according to the charge and discharge power instruction, and marks the overlimit instructions and inputs them into the constraint decomposition unit of the constraint module.

[0045] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the medium-term optimization layer receives the energy storage capacity threshold output by the long-term optimization layer based on a rolling time domain window, and constructs a dynamic optimization objective function in combination with the real-time collected electricity price signal. The rolling time domain window dynamically adjusts the window length according to the statistical characteristics of the new energy output prediction error, and embeds a robust optimization layer within the model predictive control framework to generate charge and discharge power commands against prediction fluctuations. The robust optimization layer maps the real-time electricity price signal to an economic weight coefficient for power regulation by introducing slack variables and uncertainty boundary constraints, generates rolling optimization commands on a minute-level time scale, and outputs them to the short-term optimization layer.

[0046] After receiving the rolling optimization command, the short-term optimization layer identifies the voltage over-limit risk in the high-load area based on the distribution network topology structure and the node voltage sensitivity analysis result. The distributed alternating direction multiplier method solver decomposes the mixed integer programming problem into multiple sub-optimization tasks, coordinates the power distribution schemes of each sub-problem through parallel computing, and generates charge and discharge scheduling commands for the electric vehicle cluster that satisfy the voltage constraints. The solver marks the commands that exceed the voltage safety threshold, and inputs them to the constraint decomposition unit of the constraint module after attaching the time stamp and node location information.

[0047] The constraint decomposition unit analyzes the spatio-temporal correlation constraint conditions according to the marked over-limit commands, adjusts the relaxation factor of the feasible region boundary through the hyperplane projection algorithm, and generates a dynamic constraint adjustment coefficient. The adjusted constraint conditions are fed back to the distributed solver of the short-term optimization layer to update the constraint boundaries and penalty term parameters of the sub-optimization tasks, forming a closed-loop optimization mechanism. The constraint module synchronously receives the spatio-temporal distribution characteristics of the over-limit commands, updates the training data priority of the policy network through a double experience replay mechanism, and optimizes the generation logic of subsequent rolling optimization commands.

[0048] The medium-term optimization layer and the short-term optimization layer form a cross-time scale cooperation mechanism through command flow and constraint feedback. The optimization result of the rolling time domain window restricts the feasible region range of the short-term optimization, and the over-limit commands of the short-term optimization layer trigger the dynamic correction of the scenario tree parameters by the long-term optimization layer. The parallel computing results of the distributed solver are used to reversely update the robust optimization parameters within the rolling time domain window, improve the adaptability of the medium-term optimization layer to real-time operating condition fluctuations, and maintain the dynamic balance of the multi-level optimization process.

[0049] Specifically, in the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the policy learning unit of the constraint module transfers the knowledge of the complex policy network to the lightweight model through policy distillation technology, and deploys the distilled model parameters to the edge-side controller; The federated learning unit rolls back the update operations that conflict with the lightweight model according to the parameter version information stored in the blockchain, and synchronizes the rollback commands to all edge nodes to update the local models.

[0050] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the policy learning unit constructs a teacher-student model architecture through policy distillation technology, and transfers the spatio-temporal correlation constraint knowledge implicit in the complex policy network to the lightweight model. The teacher-student model uses the attention mechanism to match the output of the feature layer, and through the joint optimization of the soft label loss function and the feature distillation loss function, extracts the key parameters of the charge and discharge instruction generation mode in the complex policy network, and generates a lightweight model suitable for edge computing resources. After the distilled model parameters are quantized, compressed and format-converted, they are deployed to the real-time inference engine of the edge-side controller for rapid generation of local charge and discharge instructions.

[0051] The federated learning unit records the model parameter version information of each edge node through the blockchain smart contract, including the parameter hash value, update timestamp and device identifier. When it is detected that there is a gradient conflict between the newly uploaded local parameters and the global parameters of the lightweight model, the federated learning unit calls the smart contract to verify the consistency of the parameter version and triggers a rollback operation based on the version tree. The rollback instruction carries the blockchain index information of the historical valid parameters and is synchronized to all edge nodes through the peer-to-peer communication protocol to replace the conflicting parameters and reload the lightweight model to maintain the global consistency of the distributed optimization process.

[0052] The policy learning unit and the federated learning unit form a collaborative optimization mechanism through the data bus. The deployment parameters of the lightweight model are reversely input into the training process of the complex policy network, and the adaptability of the original model to the edge computing scenario is enhanced through adversarial training. The parameter update log stored in the blockchain is parsed into time series features and input into the loss function of policy distillation to dynamically adjust the knowledge transfer weight and optimize the feature alignment accuracy of the teacher-student model. After receiving the rollback instruction, the edge node synchronously updates the constraint condition parsing rule of the local model, and fuses the historical valid parameters and real-time data through incremental learning to reduce the interference of parameter conflicts on the optimization control module.

[0053] Specifically, in the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the consistency unit of the fault tolerance module smooths the parameter mutation of the policy network in the optimization control module through the exponentially weighted moving average algorithm, and inputs the smoothed parameters into the causal inference unit of the constraint module to decouple the spatio-temporal constraints; The fault tolerance training unit adjusts the difficulty gradient of the curriculum learning strategy according to the simulation results of historical extreme working conditions, and feeds back the adjustment signal to the generative adversarial network to update the adversarial sample generation logic.

[0054] In the energy management measurement and control system integrating electric vehicles and renewable energy according to the present invention, the consistency unit of the fault tolerance module smooths the parameter update sequence of the policy network in the optimization control module through the exponentially weighted moving average algorithm, suppressing the violent oscillation of parameters caused by the fluctuation of new energy output or load mutation. The smoothed parameters carry the stability characteristics of the historical training stage and are input to the causal reasoning unit of the constraint module to analyze the coupling relationship between the voltage regulation target and the charge-discharge power distribution in the spatio-temporal constraints, generating a decoupled set of independent constraint conditions. The causal reasoning unit adjusts the projection direction of the constraint decomposition unit based on the decoupling result, reducing the dimensional conflict of the feasible region boundary and improving the convergence efficiency of the distributed solver in the short-term optimization layer.

[0055] The fault tolerance training unit reconstructs the composite fault scenario of the sudden drop of new energy output and the sudden increase of electric vehicle load according to the historical extreme working condition data stored in the blockchain, and gradually increases the training difficulty gradient through the curriculum learning strategy. In the initial stage, the data of a single fault mode is loaded to train the policy network, and the spatio-temporal correlation interference factors are gradually superimposed. The generated adversarial network dynamically adjusts the simulation intensity and complexity of the extreme working conditions according to the anti-interference ability of the current policy network. The adjustment signal is fed back to the spatio-temporal feature extraction module of the generated adversarial network to optimize the noise injection ratio of the generator and the anomaly detection threshold of the discriminator, generating adversarial samples covering multi-dimensional interference features.

[0056] The consistency unit and the fault tolerance training unit form a dynamic balance mechanism through parameter interaction. The smoothed parameters output by the exponentially weighted moving average algorithm are used as the feature input of the generated adversarial network to match the generation range of the adversarial samples with the adaptation boundary of the policy network. The adjustment of the difficulty gradient of the curriculum learning strategy triggers the consistency unit to recalculate the parameter smoothing coefficient and optimize the decoupling accuracy of the causal reasoning unit. The fault tolerance module synchronously records the generation characteristics of the adversarial samples and the response data of the policy network, and updates the fault mode library stored in the blockchain through the federated learning unit, providing incremental data support for subsequent parameter smoothing and adversarial training.

[0057] In a second aspect, please refer to Figure 1 The energy management measurement and control method for integrating electric vehicles and renewable energy provided by the present invention is applied to the energy management measurement and control system for integrating electric vehicles and renewable energy as described above, and includes: Step S101, collecting photovoltaic output data, wind power output data and electric vehicle load demand data, filtering outliers by using the isolation forest algorithm and compensating the time series deviation based on the dynamic time warping algorithm, and outputting the processed multi-source data; Step S102, generating optimization instructions according to the processed multi-source data on different time scales: The long-term optimization generates an uncertainty scenario tree of renewable energy output and load demand, and corrects the energy storage capacity threshold based on the energy storage life loss model; The medium-term optimization generates rolling optimization instructions through a model predictive control framework with a robust optimization layer embedded, in combination with the energy storage capacity threshold and real-time electricity price signals; The short-term optimization uses a distributed alternating direction method of multipliers solver to decompose the mixed-integer programming problem in the rolling optimization instructions, and preferentially processes the distribution network node voltage constraints to generate charge and discharge instructions; In step S103, the charge and discharge instructions are sent to the electric vehicle bidirectional converter device through a multi-priority message queue, the instruction execution threshold is dynamically adjusted based on the Lyapunov stability theory, and the execution result is fed back to the policy network parameter update process; In step S104, the policy network parameters in the optimization control module are updated through a double experience replay mechanism, the spatio-temporal correlation constraints in the charge and discharge instructions are decoupled using causal inference technology, and the constraint decomposition result is input into the distributed solver for short-term optimization; In step S105, adversarial samples are generated by a generative adversarial network to simulate extreme working conditions and injected into the training dataset of the policy network, the root cause of instruction conflicts is inferred based on a Bayesian network, and the service priority of the medium-term optimization layer is dynamically degraded.

[0058] Explanation of the technical feature terms of the present invention is as follows: Acquisition module: Distributed sensor network: Composed of multiple sensors distributed in photovoltaic power plants, wind farms, and electric vehicle charging stations, used to collect photovoltaic power output, wind power output, and electric vehicle load data in real time.

[0059] Isolation forest algorithm: An unsupervised machine learning algorithm used to detect and filter outliers in data to improve data quality.

[0060] Dynamic time warping algorithm: A time series alignment technique used to compensate for the time deviation between multi-source data with different acquisition frequencies (such as high-frequency load data and low-frequency prediction data) to generate synchronized time series data.

[0061] Optimization control module: Long-term optimization layer: Generates an uncertainty scenario tree for renewable energy output and load demand based on Monte Carlo sampling, simulates the power fluctuations of future multi-time sections, and dynamically corrects the energy storage capacity threshold in combination with the energy storage life loss model to avoid overcharging and over-discharging.

[0062] Medium-term optimization layer: Embeds a robust optimization model in a model predictive control (MPC) framework, generates rolling optimization instructions in combination with real-time electricity price signals, and dynamically allocates the weights of economic and low-carbon targets through the entropy weight method to improve the anti-interference ability of the instructions.

[0063] Short-term optimization layer: Using a distributed Alternating Direction Method of Multipliers (ADMM) solver, decompose the mixed-integer programming problem into parallel subtasks, prioritize the processing of distribution network node voltage constraints, and generate charge and discharge instructions.

[0064] Control execution module: Multi-priority message queue: Classify and issue instructions to the bidirectional converter of electric vehicles according to the urgency of the instructions (such as the highest priority for voltage violation correction instructions), and implement the priority execution of key instructions.

[0065] Lyapunov stability theory: By dynamically adjusting the instruction execution threshold (such as the upper limit of charge and discharge power), maintain the stability of the power grid operation and avoid voltage or frequency oscillations.

[0066] Constraint module: Dual experience replay mechanism: Sample and train the policy network from historical data and real-time feedback, balance the weights of new and old experiences, and avoid model overfitting.

[0067] Causal inference technology: Decouple the spatio-temporal correlation constraints in the charge and discharge instructions (such as the coupling relationship between voltage regulation and power distribution) through the hyperplane projection algorithm, and generate the boundary of the dynamic feasible region.

[0068] Fault tolerance module: Generative Adversarial Network (GAN): Consisting of a generator and a discriminator, the generator simulates extreme working condition data (such as a sudden drop in new energy output), and the discriminator verifies the authenticity of the data to enhance the anti-interference ability of the policy network.

[0069] Bayesian network: Based on probabilistic reasoning, locate the root cause of instruction conflicts (such as resource competition or timing conflicts), and generate a priority adjustment plan or resource release instructions.

[0070] Monte Carlo sampling: Generate multiple groups of possible new energy output and load demand scenarios through random sampling, and construct an uncertainty scenario tree for the solution space compression of long-term optimization.

[0071] Federated learning: On the premise of protecting the data privacy of the edge side (through differential privacy encryption), aggregate the local model parameters of multiple edge nodes to generate a global optimization model.

[0072] Blockchain evidence storage: Store the encrypted model parameters and operation logs in a distributed ledger, achieve data immutability, and provide a traceability basis for parameter version conflicts.

[0073] Policy distillation technology: Transfer the knowledge of complex policy networks to lightweight models, reduce the computational load of edge-side controllers, and retain the key constraint parsing ability at the same time.

[0074] Course learning strategy: Train the strategy network in stages, gradually increase the complexity of adversarial samples (such as from single faults to compound faults), and enhance the generalization ability of the model.

[0075] The synchronous timing data of the acquisition module of the present invention → the long-term optimization layer scenario tree → the medium-term optimization layer rolling instruction → the short-term optimization layer charge and discharge instruction → the control execution module issues and feedbacks → the constraint module dynamically adjusts → the fault tolerance module diagnoses conflicts, forming a closed-loop optimization link.

[0076] Cross-layer collaboration: The long-term optimization layer provides energy storage capacity threshold constraints for medium-term optimization, the medium-term optimization layer generates instruction boundary constraints for short-term optimization, and the short-term optimization layer feeds back out-of-limit instructions to correct the long-term scenario tree parameters, realizing multi-time scale dynamic coordination.

[0077] Anti-interference mechanism: The fault tolerance module improves the robustness of the strategy network through adversarial sample training, the Bayesian network diagnoses conflicts in real time, and the service orchestration unit dynamically degrades non-critical services to ensure the stability of the system under extreme working conditions.

[0078] Long-term optimization layer model: Monte Carlo sampling model: Function: Generate an uncertainty scenario tree of the output of renewable energy (photovoltaic, wind power) and the load demand of electric vehicles through random sampling, and simulate the power fluctuation distribution of multiple time sections.

[0079] Technical effect: Provide a multi-dimensional solution space for long-term optimization and reduce the complexity of high-dimensional mixed integer programming problems.

[0080] Medium-term optimization layer model: Model predictive control (MPC) model embedded with robust optimization: Function: Combine real-time electricity price signals and energy storage capacity thresholds within a rolling time domain window to generate rolling optimization instructions against prediction errors.

[0081] Technical effect: By introducing slack variables and uncertainty boundary constraints, improve the robustness of instructions to new energy output fluctuations.

[0082] Short-term optimization layer model: Distributed alternating direction multiplier method (ADMM) solver Function: Decompose the mixed integer programming problem into parallel subtasks, prioritize the processing of distribution network node voltage constraints, and generate charge and discharge instructions.

[0083] Technical effect: Identify voltage out-of-limit risk nodes through topological sorting, coordinate the optimization results of each sub-problem, and improve the solution efficiency.

[0084] Constraint module model: Federated learning model: Function: Generate a global model by aggregating the local model parameters of edge nodes while protecting data privacy (differential privacy encryption).

[0085] Technical effect: Achieve traceability of parameter versions through blockchain evidence storage, trigger parameter rollback after detecting conflicts, and maintain optimization consistency.

[0086] Causal inference model: Function: Analyze the spatio-temporal coupling relationship in charge and discharge instructions (such as the correlation between voltage regulation and power distribution), and generate a decoupled independent constraint set.

[0087] Technical effect: Suppress the deviation of the optimization path caused by new energy fluctuations and load mutations, and improve the stability of instruction execution.

[0088] Fault tolerance module model: Generative adversarial network (GAN) model: Function: The generator simulates extreme working condition data (such as grid voltage dips, sudden drops in new energy output), and the discriminator verifies the authenticity of the data.

[0089] Technical effect: Inject adversarial samples into the training dataset of the policy network to enhance the model's adaptability to abnormal working conditions.

[0090] Bayesian network model: Function: Based on probabilistic reasoning, locate the root cause of instruction conflicts (such as resource competition, timing overlap), and generate instructions for priority adjustment or resource release.

[0091] Technical effect: Collaborate with the closed-loop verification module to dynamically correct the instruction dependency relationship and reduce the conflict trigger frequency.

[0092] Acquisition module model: Dynamic time warping (DTW) algorithm: Function: Calibrate the electric vehicle load data collected at high frequency and the renewable energy prediction data updated at low frequency to generate synchronous time series data.

[0093] Technical effect: Eliminate the time series deviation of multi-source data through timestamp matching and interpolation compensation, and provide consistent input for hierarchical optimization.

[0094] Wavelet denoising model: Function: Based on spectrum analysis, identify the periodic noise components in the photovoltaic output data, dynamically adjust the wavelet basis function and decomposition level, and retain the effective signal.

[0095] Technical effect: Suppress the interference of noise on optimal control and improve the accuracy of data preprocessing.

[0096] Policy network model: Dual experience replay mechanism: Function: Balance sampling from historical data and real-time feedback, update the parameters of the policy network, and avoid model overfitting.

[0097] Technical effect: Combine the policy distillation technology to transfer complex policy knowledge to a lightweight model, reducing the computational load on the edge side.

[0098] Service orchestration model: Analytic Hierarchy Process (AHP) resource evaluation model: Function: Dynamically evaluate the utilization rate of computing resources and the status of communication links, and generate service priority adjustment signals.

[0099] Technical effect: When the resource availability is lower than the threshold, mask non-critical services in the long-term optimization layer, switch the medium-term optimization to a simplified model, and ensure real-time performance.

[0100] The specific implementation manner of the present invention relates to an energy management measurement and control system and method for integrating electric vehicles and renewable energy. Aiming at the problems of instruction mismatch and path deviation existing in the prior art when dealing with high-dimensional mixed integer programming solutions and spatio-temporal correlation interference, a hierarchical optimization architecture and dynamic constraint decomposition technology are adopted to achieve efficient regulation. The system includes a collection module, an optimization control module, a control execution module, a constraint module, and a fault tolerance module. The specific implementation manner is as follows: The collection module collects the output data of the photovoltaic power station, the output data of the wind farm, and the load demand data of the electric vehicle charging station in real time through a distributed sensor network. The isolation forest algorithm is used to detect and filter outliers in the original data, and abnormal data points are removed. For the periodic noise existing in the photovoltaic output data, spectral analysis is performed through the fast Fourier transform. After identifying the main frequency characteristics of the noise, the type of basis function and the decomposition layer number of the wavelet denoising algorithm are dynamically matched, and the decomposition layer number is set to 3-5 layers to effectively separate the noise frequency band and retain the low-frequency trend component. For the high-frequency collected electric vehicle load data and the low-frequency updated renewable energy prediction data, time series alignment is performed through the dynamic time warping algorithm, and the time series deviation of the data is calibrated using a cache queue with timestamp matching to generate synchronous time series data and output it to the long-term optimization layer and the medium-term optimization layer of the optimization control module.

[0101] After the long-term optimization layer of the optimization control module receives the synchronous timing data, it uses the Monte Carlo sampling method to generate an uncertainty scenario tree for the renewable energy output and load demand, simulates the new energy output fluctuations and load demand changes per hour within the next 24 hours, and generates 100-200 groups of scenario samples. Combining with the energy storage life loss model, according to the mapping relationship between the charge-discharge cycle times and the capacity attenuation rate, dynamically corrects the charge-discharge capacity threshold of the energy storage unit, and the correction range is ±5% of the nominal capacity, to avoid the impact of overcharging and over-discharging on the life of the energy storage device, and transmits the corrected capacity threshold to the medium-term optimization layer. The medium-term optimization layer embeds a robust optimization model in the model predictive control framework, sets the rolling time domain window length to 15-30 minutes, combines the real-time electricity price signal and the corrected energy storage capacity threshold, dynamically allocates the weights of the economic objective and the low-carbon objective through the entropy weight method, generates a rolling optimization instruction including the electricity price sensitivity coefficient and the carbon emission factor, and transmits it to the short-term optimization layer. The short-term optimization layer uses a distributed alternating direction multiplier method solver to decompose the mixed integer programming problem into 8-12 parallel subtasks, identifies the voltage violation risk nodes based on the distribution network topology structure, preferentially processes the charge-discharge power distribution problem in the sensitive area through topological sorting, generates an electric vehicle cluster scheduling instruction that satisfies the node voltage constraint, marks the instructions that exceed the voltage safety threshold by ±5%, and feeds back the violation instructions to the constraint module.

[0102] After the control execution module receives the charge-discharge instructions generated by the short-term optimization layer, it distributes the instructions to the electric vehicle bidirectional converter device in levels through a multi-priority message queue. The priority is divided according to the voltage constraint urgency, and the voltage violation correction instruction is set as the highest priority. Dynamically adjusts the instruction execution threshold based on the Lyapunov stability theory, and sets the charge-discharge power adjustment step size to 2%-5% of the rated power to ensure the stability of the charge-discharge process. The execution result is transmitted to the constraint module through a closed-loop feedback link, including the instruction execution success rate, voltage fluctuation range, and power deviation data. The constraint module updates the policy network parameters using a double experience replay mechanism, samples from the historical data pool and real-time feedback data according to a 7:3 ratio, combines causal inference technology to analyze the spatio-temporal coupling relationship in the charge-discharge instructions, decomposes the complex constraints into independent sub-constraint sets through the hyperplane projection algorithm, dynamically adjusts the relaxation factor of the feasible region boundary to 0.1-0.3, generates constraint conditions and inputs them into the distributed solver of the short-term optimization layer to optimize the subsequent instruction generation logic.

[0103] The fault-tolerant module simulates extreme operating conditions such as grid voltage dips, sudden drops in new energy output, and rapid load surges through a generative adversarial network. The generator takes historical fault data as input and reconstructs adversarial samples with a voltage dip amplitude of 10%-20% of the rated value and a sudden output drop rate of 5%-10% per minute. The discriminator verifies the sample credibility based on the real data distribution characteristics in the dual experience replay mechanism and injects the adversarial samples that meet the threshold conditions into the training dataset of the policy network at a ratio of 20%-30%. The Bayesian network receives the instruction conflict feature parameters transmitted by the closed-loop verification module, including priority tags, resource occupancy rates, and historical conflict frequencies, locates the root cause of the conflict as abnormal resource allocation or overlapping time windows through probabilistic inference, generates an instruction priority adjustment plan or a resource release instruction, dynamically reduces the service priority of the medium-term optimization layer to 70%-80% of the original priority, triggers the service orchestration unit to shield the scenario tree generation service of the long-term optimization layer, switches the medium-term optimization layer to a simplified model predictive control framework, replaces the dynamic weight allocation with a fixed-weight multi-objective optimization model, and shortens the rolling optimization instruction generation period to 50% of the original period.

[0104] During the system operation, the federated learning unit adds Laplace noise (with a noise scale set to 0.1-0.5) to the edge-side electric vehicle load data through differential privacy technology. The encrypted local model parameters are attached with timestamps and device identifiers and uploaded to the blockchain node for distributed storage. The edge node parameters are aggregated every 1-2 hours, and the global model parameters are fused and generated using the homomorphic encryption algorithm and then sent to the policy network. When a conflict is detected between the parameter hash value and the historical version, the smart contract is triggered to roll back to the nearest valid version, and all edge node local model parameters are updated synchronously, with the deviation tolerance range set to ±2%. The policy distillation unit migrates the constraint parsing rules in the complex policy network to the lightweight model, compresses the model parameter quantity to 30%-40% of the original model, deploys it to the edge-side controller for real-time inference, and controls the response delay within 200 milliseconds.

[0105] Through the above implementation methods, the system realizes the dynamic generation and closed-loop correction of multi-time-scale instructions. The long-term optimization layer compresses the solution space dimension, the medium-term optimization layer improves the instruction anti-interference ability, the short-term optimization layer efficiently solves the voltage constraint problem, and combines the constraint decomposition and fault-tolerant mechanism to suppress the path deviation, ultimately achieving the goals of energy supply-demand balance, optimal economy, and low-carbon operation.

[0106] The present invention effectively solves the problem of difficult solution of high-dimensional mixed-integer programming models through a multi-time-scale hierarchical optimization architecture and dynamic constraint decomposition technology. The long-term optimization layer of the optimization control module generates an uncertainty scenario tree based on Monte Carlo sampling, and combines the energy storage life loss model to compress the dimension of the solution space; the medium-term optimization layer embeds a robust optimization layer in the model predictive control framework, and dynamically allocates multi-objective weights through the entropy weight method to generate rolling instructions, reducing the complexity of the real-time optimization problem; the short-term optimization layer uses a distributed alternating direction multiplier method solver to decompose the mixed-integer programming into parallel sub-problems, and gives priority to processing voltage violation constraints by topological sorting, realizing the efficient solution of high-dimensional models.

[0107] Aiming at the path deviation problem caused by spatio-temporal correlation interference, the constraint module decouples spatio-temporal constraints through causal inference technology and updates the parameters of the policy network in combination with the federated learning mechanism. The causal inference unit analyzes the spatio-temporal coupling relationship in the charge and discharge instructions, and generates dynamic constraint conditions by smoothing the boundary of the feasible region through the hyperplane projection algorithm; the federated learning unit aggregates the encrypted parameters of the edge nodes to build a global model, uses the blockchain evidence storage mechanism to trace parameter version conflicts, and rolls back abnormal update operations to eliminate the impact of spatio-temporal interference on the optimization path.

[0108] The system maintains the stability of the rolling optimization path through a closed-loop feedback mechanism. The control execution module checks the instruction dependency relationship based on a directed acyclic graph, triggers a Bayesian network to locate the root cause of the conflict, and retransmits the failed instructions; the fault-tolerant training unit of the fault-tolerant module uses a curriculum learning strategy to inject adversarial samples in stages, and dynamically adjusts the extreme working condition simulation logic of the generative adversarial network; the acquisition module compensates the time series deviation of multi-source data through the dynamic time warping algorithm, providing accurate input for hierarchical optimization. Each module collaborates to realize the closed-loop generation and dynamic correction of multi-objective optimization instructions, suppressing the risk of path deviation.

Claims

1. An energy management and control system integrating electric vehicles and renewable energy, characterized in that: include: The acquisition module is used to collect photovoltaic output data, wind power output data and electric vehicle load demand data, process the collected data, and output synchronous time series data to the optimization control module; The optimization control module is used to receive synchronous time series data and generate optimization instructions in layers according to long-term, medium-term and short-term time scales: The long-term optimization layer generates an uncertainty scenario tree of renewable energy output and load demand based on the synchronized time series data, dynamically corrects the energy storage capacity threshold in combination with the energy storage life loss model, and inputs the corrected threshold into the medium-term optimization layer; The medium-term optimization layer embeds a robust optimization model in the model predictive control framework, generates rolling optimization instructions according to the modified energy storage capacity threshold and the real-time electricity price signal, and transmits the instructions to the short-term optimization layer; The short-term optimization layer uses a distributed solver to decompose the mixed integer programming problem in the rolling optimization instruction, and generates charging and discharging instructions based on the voltage constraint priority of the distribution network node; A control execution module, used for receiving the charge and discharge instructions generated by the short-term optimization layer, issuing instructions to the electric vehicle bidirectional converter through a multi-priority message queue, dynamically adjusting the execution threshold based on stability control theory, and feeding back the execution result to the constraint module; A constraint module, used to update the policy network parameters in the optimization control module through a dual experience playback mechanism, decouple the spatiotemporal correlation constraints of the charge and discharge instructions in combination with causal reasoning technology, and input the decoupled constraints into the distributed solver of the short-term optimization layer; The fault-tolerant module is used to simulate extreme operating condition data through adversarial sample generation network, inject the policy network training data set of the optimization control module, analyze the root cause of the instruction conflict of the control execution module based on the Bayesian network, and dynamically adjust the service priority of the mid-term optimization layer.

2. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1 is characterized in that: The long-term optimization layer generates an uncertainty scenario tree of renewable energy output and load demand through a Monte Carlo sampling method, and dynamically corrects the charge and discharge capacity threshold of the energy storage unit based on the energy storage life loss model; The mid-term optimization layer receives the scenario tree generated by the long-term optimization layer, dynamically allocates multi-objective optimization weights in combination with the entropy weight method, generates a rolling optimization instruction including an electricity price signal and a carbon emission factor, and transmits the instruction to the short-term optimization layer; The short-term optimization layer decomposes the mixed integer programming problem into parallel subtasks through a topological sorting algorithm according to the rolling optimization instructions transmitted by the medium-term optimization layer and based on the voltage constraint priority of the distribution network nodes, and assigns the subtasks to the distributed solver to perform parallel computing.

3. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1 is characterized in that: The constraint module includes: A federated learning unit, which is used to encrypt the electric vehicle load data on the edge side through differential privacy technology, upload the encrypted local model parameters to the blockchain node for distributed evidence storage, and generate global model parameters through federated aggregation, and send them to the policy network of the optimization control module; The constraint decomposition unit receives the global model parameters generated by the federated learning unit, smoothes the feasible domain boundary of the mixed integer programming problem based on the hyperplane projection algorithm, generates dynamic constraint conditions and inputs them into the distributed solver of the short-term optimization layer.

4. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1, characterized in that: The fault-tolerant module comprises: A fault-tolerant training unit, used for training the policy network of the optimization control module in stages through a curriculum learning strategy, and injecting the extreme operating condition data generated by the generative adversarial network into the training data set of the dual experience playback mechanism of the optimization control module to enhance the anti-interference ability of the policy network; The service orchestration unit dynamically adjusts the optimization layer service priority according to the resource availability evaluation result. When the resource availability is lower than the preset threshold, the scene tree generation service of the long-term optimization layer is blocked, and a downgrade instruction is sent to the medium-term optimization layer to switch to the simplified rolling optimization model.

5. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1, characterized in that: The acquisition module comprises: A spectrum analysis unit is used to perform spectrum analysis on photovoltaic output data, identify the main frequency characteristics of periodic noise components, and dynamically match the basis function type and decomposition layer number of the wavelet denoising algorithm according to the main frequency characteristics to generate denoised photovoltaic output data; The timing alignment unit receives the denoised photovoltaic output data and the original wind power output data generated by the spectrum analysis unit, calibrates the high-frequency collected electric vehicle load data and the low-frequency updated renewable energy forecast data through a cache queue matched with a timestamp, generates synchronized timing data and outputs it to the long-term optimization layer and the medium-term optimization layer of the optimization control module.

6. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1, characterized in that: The closed-loop verification unit of the control execution module models the dependency relationship of the charge and discharge instructions through a directed acyclic graph. When an instruction conflict is detected, the characteristic parameters of the conflicting instructions are transmitted to the Bayesian network reasoning unit of the fault-tolerant module to locate the source of the conflict and generate an adjustment instruction. The instruction issuing unit retransmits the invalid instruction according to the verification failure result, and feeds back the verification log including the conflict type and the number of retransmissions to the strategy distillation unit of the constraint module for updating the training priority of the strategy network in the optimization control module.

7. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1, characterized in that: The mid-term optimization layer receives the energy storage capacity threshold generated by the long-term optimization layer through a rolling time domain window, generates a charge and discharge power instruction on a minute time scale in combination with the real-time electricity price signal, and transmits the instruction to the short-term optimization layer; The short-term optimization layer calculates the electric vehicle cluster scheduling plan that meets the distribution network node voltage constraints through the distributed solver based on the received charging and discharging power instructions, marks the instructions that exceed the voltage safety threshold, and inputs the marked out-of-limit instructions into the constraint decomposition unit of the constraint module to adjust the dynamic constraint conditions.

8. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1, characterized in that: The strategy learning unit of the constraint module migrates the constraint parsing rules of the complex strategy network in the optimization control module to the lightweight model through the knowledge migration technology, and deploys the migrated model parameters to the edge side controller; The federated learning unit of the constraint module detects the parameter conflict of the lightweight model according to the parameter version information stored in the blockchain, triggers the parameter rollback operation to restore the historical valid parameters, and synchronizes the rollback instruction to all edge nodes to update the local model parameters.

9. The energy management and control system for integrating electric vehicles and renewable energy according to claim 1, characterized in that: The consistency unit of the fault-tolerant module smoothes the parameter update sequence of the strategy network in the optimization control module through a parameter smoothing algorithm, and inputs the processed parameters into the causal reasoning unit of the constraint module for decoupling the spatiotemporal correlation constraints in the charge and discharge instructions; The fault-tolerant training unit of the fault-tolerant module dynamically adjusts the difficulty gradient of the course learning strategy according to the simulation data of historical extreme working conditions, and feeds back the gradient adjustment signal to the generative adversarial network to optimize the generation parameters of the adversarial samples.

10. An energy management and control method for integrating electric vehicles and renewable energy, applied to an energy management and control system for integrating electric vehicles and renewable energy as claimed in any one of claims 1 to 9, characterized in that: include: The acquisition module collects photovoltaic output data, wind power output data and electric vehicle load demand data, uses the isolation forest algorithm to filter outliers and compensates for timing deviations based on the dynamic time warping algorithm to generate synchronized timing data; By optimizing the long-term optimization layer of the control module, an uncertainty scenario tree of renewable energy output and load demand is generated based on the synchronized time series data, and the energy storage capacity threshold is dynamically corrected in combination with the energy storage life loss model; The mid-term optimization layer of the optimization control module receives the energy storage capacity threshold value corrected by the long-term optimization layer, embeds the robust optimization layer in the model predictive control framework, and generates rolling optimization instructions in combination with the real-time electricity price signal; Through the short-term optimization layer of the optimization control module, a distributed solver is used to decompose the mixed integer programming problem in the rolling optimization instruction, and the voltage constraints of the distribution network nodes are processed first to generate charging and discharging instructions; The charging and discharging instructions are issued to the bidirectional converter of the electric vehicle through the control execution module, the execution threshold is dynamically adjusted based on the Lyapunov stability theory, and the execution result is fed back to the constraint module; The policy network parameters of the optimization control module are updated through the dual experience playback mechanism of the constraint module, the spatiotemporal correlation constraints in the charge and discharge instructions are decoupled using causal reasoning technology, and the decoupled constraints are input into the distributed solver of the short-term optimization layer; The generative adversarial network of the fault-tolerant module simulates extreme working conditions to generate adversarial samples and injects them into the training data set of the policy network. Based on the Bayesian network inference of the root cause of the instruction conflict, the service priority of the mid-term optimization layer is dynamically adjusted.

Images