Energy management method for mobile energy storage vehicles based on intelligent algorithm

Abstract

The present invention relates to the technical field of energy management for mobile energy storage vehicles, and discloses an energy management method for mobile energy storage vehicles based on an intelligent algorithm. The steps include: collecting multi-dimensional energy status data such as battery pack charge and discharge rate, ambient temperature, and grid load fluctuation; constructing a heterogeneous layered architecture of edge computing nodes, cloud collaboration platforms, and on-board control terminals, dividing energy partitions according to grid peaks and valleys, and configuring low-latency response, global optimization, and local closed-loop control modes; fusing multi-source heterogeneous energy data streams and eliminating outliers; constructing a dynamic optimization decision model through a battery degradation model and a grid supply and demand balance equation, and using a multi-objective iterative solver and a parallel gradient descent algorithm to jointly solve multi-parameter collaborative constraint relationships; based on constraint boundary trigger conditions, generating a three-level adaptive response instruction sequence for local battery overload alarms, regional grid frequency regulation warnings, and global energy scheduling imbalance predictions. This method improves the accuracy, coordination, and robustness of energy management.

Description

Technical Field

[0001] The present invention relates to the technical field of energy management of mobile energy storage vehicles, and in particular to an energy management method for mobile energy storage vehicles based on an intelligent algorithm. Background Art

[0002] With the rapid development of the new energy industry, mobile energy storage vehicles (ESS) are increasingly being used as flexible energy storage and regulation units in scenarios such as peak load shaving and valley filling, and emergency power supply. However, current energy management of ESSs faces numerous challenges, and more efficient management methods are urgently needed to improve their operational performance and grid coordination.

[0003] From the perspective of energy data management, traditional methods can only process energy data in a single dimension, making it difficult to comprehensively consider the coupled effects of multiple factors, such as battery pack charge and discharge rates, ambient temperature distribution, and grid load fluctuations. For example, ignoring the significant impact of ambient temperature on battery performance can lead to reduced charge and discharge efficiency and even increased battery life loss at extreme temperatures. Failure to fully analyze the characteristics of grid load fluctuations makes it impossible to accurately grasp energy scheduling needs during peak and valley periods, making it difficult to achieve efficient interaction between mobile energy storage vehicles and the grid.

[0004] In terms of system architecture deployment, existing energy management systems mostly use a centralized control architecture, which suffers from high response latency and insufficient global optimization capabilities. This centralized architecture suffers from poor coordination between edge devices and cloud platforms, making it impossible to flexibly partition energy for peak and valley periods of grid load fluctuations. For example, during peak load periods, the lack of an effective hierarchical scheduling mechanism can prevent mobile energy storage vehicles from responding to grid frequency regulation needs in a timely manner, impacting grid stability.

[0005] When it comes to data processing and optimization decision-making, traditional algorithms struggle to integrate and dynamically optimize heterogeneous data from multiple sources. Multi-source data (such as current time series data, load frequency data, and temperature sampling data) suffers from issues like sampling interval deviation and noise interference. Traditional data cleaning and fusion methods struggle to ensure temporal and spatial consistency, resulting in insufficient input data accuracy for optimization decision-making models. Furthermore, existing optimization models are often single-objective (e.g., focusing solely on energy efficiency or thermal management) and are unable to simultaneously address the collaborative constraints of multiple variables, such as charge and discharge efficiency, temperature compensation, and load distribution, making it difficult to achieve global optimal energy management.

[0006] In terms of response mechanisms, traditional threshold-triggered alarms lack adaptive regulation capabilities and are unable to adjust response strategies in a timely manner based on dynamic changes in system operating conditions. For example, a single current threshold alarm cannot distinguish between instantaneous current fluctuations and actual overload conditions, potentially leading to false or missed alarms. Furthermore, the system lacks multi-scale analysis and early prediction capabilities for grid frequency regulation warnings and global energy scheduling imbalance predictions, making it difficult to implement effective control measures before system instability occurs.

[0007] Furthermore, with the widespread deployment of mobile energy storage vehicles, the complexity of their interactions with the power grid has increased significantly, and traditional energy management methods are struggling in terms of scalability and robustness. Building a scalable, heterogeneous, layered architecture that enables collaborative operation between edge computing nodes, cloud platforms, and onboard terminals, as well as designing efficient optimization algorithms to accommodate large-scale data processing and real-time decision-making, are pressing technical challenges.

[0008] Existing energy management methods for mobile energy storage vehicles have significant shortcomings in multi-dimensional data processing, layered architecture coordination, dynamic optimization decision-making, and adaptive response mechanisms. These shortcomings make it difficult to meet the smart grid's demand for efficient, reliable, and flexible energy management of mobile energy storage systems. Therefore, it is urgent to propose an energy management method for mobile energy storage vehicles based on intelligent algorithms. By integrating multiple technologies and innovating architectures, this method could improve the energy utilization efficiency, grid coordination, and system reliability of mobile energy storage vehicles. Summary of the Invention

[0009] The purpose of the present invention is to provide an energy management method for a mobile energy storage vehicle based on an intelligent algorithm to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention provides the following technical solution: an energy management method for a mobile energy storage vehicle based on an intelligent algorithm, the method comprising:

[0011] Collect multi-dimensional energy status data of the mobile energy storage vehicle, including the real-time charge and discharge rate of the battery pack, the ambient temperature distribution characteristics, and the grid load fluctuation characteristics;

[0012] Based on the dynamic response requirements of the energy scheduling unit, a heterogeneous layered deployment architecture of edge computing nodes, cloud collaboration platforms, and vehicle control terminals is generated;

[0013] Synchronously acquiring the real-time energy interaction data stream of each node under the heterogeneous layered deployment architecture, and performing multi-source heterogeneous data fusion and outlier elimination processing on the real-time energy interaction data stream;

[0014] A dynamic optimization decision model is constructed by combining the battery degradation model with the grid supply and demand balance equation, and the collaborative constraint relationship among the charge and discharge efficiency matrix, the temperature compensation coefficient tensor, and the load distribution vector in the dynamic optimization decision model is solved jointly;

[0015] Based on the boundary trigger conditions of the collaborative constraint relationship, a three-level adaptive response instruction sequence is generated, including local battery overload alarm, regional power grid frequency regulation warning and global energy scheduling imbalance prediction.

[0016] Preferably, the heterogeneous layered deployment architecture is characterized by a distributed computing hierarchy; the generation of the heterogeneous layered deployment architecture of the edge computing nodes, the cloud collaboration platform and the vehicle-mounted control terminal includes:

[0017] Based on the peak and valley period division strategy in the power grid load fluctuation characteristics, define the energy partitioning rules for the core dispatching area, buffer dispatching area and backup dispatching area;

[0018] Based on the energy partitioning rules, a low-latency response mode is configured for the edge computing node, a global optimization mode is configured for the cloud collaboration platform, and a local closed-loop control mode is configured for the vehicle-mounted control terminal.

[0019] Preferably, the dynamic optimization decision model is implemented by a multi-objective iterative solver; the simultaneous solution of the collaborative constraint relationship between the charge and discharge efficiency matrix, the temperature compensation coefficient tensor and the load distribution vector in the dynamic optimization decision model includes:

[0020] Loading the charge and discharge efficiency matrix into the energy conservation equations of the multi-objective iterative solver, and updating the capacity decay factor based on the battery health state model;

[0021] The temperature compensation coefficient tensor is embedded in the thermodynamic diffusion equation, and the anisotropic heat conductivity coefficient is calculated using Fourier's law.

[0022] The load distribution vector is discretized into time series task queues, and the task constraint matrix is constructed in combination with the dynamic priority scheduling strategy;

[0023] A parallel gradient descent algorithm is used to synchronously update the optimization residuals of the energy conservation equations, the thermodynamic diffusion equations and the task constraint matrix until a steady-state convergence condition is reached.

[0024] Preferably, the multi-objective iterative solver includes an energy efficiency optimization module, a thermal management module and a task scheduling module; the energy efficiency optimization module, the thermal management module and the task scheduling module are coupled across modules via a state sharing interface;

[0025] The energy efficiency optimization module is configured with a battery life prediction model, the thermal management module is configured with a phase change material thermal buffer model, and the task scheduling module is configured with a real-time preemptive scheduling algorithm;

[0026] The state sharing interface transmits the charge and discharge efficiency correction amount, the temperature gradient correction amount and the task priority correction amount to the adjacent functional modules in each iteration.

[0027] Preferably, the parallel gradient descent algorithm is implemented using a staged learning rate adjustment strategy; the use of the parallel gradient descent algorithm to synchronously update the optimization residuals of the energy conservation equations, the thermodynamic diffusion equation, and the task constraint matrix includes:

[0028] The charge and discharge efficiency matrix, temperature compensation coefficient tensor and load distribution vector of the current iteration step are used as initial optimization inputs;

[0029] The energy residual norm is calculated through the energy efficiency optimization module, the thermal balance residual norm is calculated through the thermal management module, and the scheduling delay residual norm is calculated through the task scheduling module;

[0030] Adaptively adjust the learning rate attenuation coefficient of the next iteration step based on the dynamic weighted geometric mean of the three types of residual norms;

[0031] When the dynamic weighted geometric mean is lower than the preset optimization threshold, the iteration is terminated and the optimized decision parameters are output.

[0032] Preferably, the adjustment strategy of the learning rate attenuation coefficient is implemented based on an adaptive momentum mechanism; the adaptive adjustment of the learning rate attenuation coefficient of the next iteration step includes:

[0033] If the residual norm convergence rate of the current iteration step is higher than the rate threshold, the learning rate amplification factor is multiplied by the baseline learning rate as a new parameter;

[0034] If the residual norm convergence rate is in a flat range, the learning rate attenuation factor is multiplied by the baseline learning rate as a new parameter;

[0035] If the residual norm shows periodic fluctuations, the random restart strategy is enabled to reset the optimization direction and step size range.

[0036] Preferably, the three-level adaptive response instruction sequence is generated by an event-driven state machine; the generation of the three-level adaptive response instruction sequence including local battery overload alarm, regional power grid frequency regulation warning and global energy scheduling imbalance prediction includes:

[0037] For local battery overload alarms, a primary trigger condition based on current mutation is set, and a sliding time window statistical method is used to detect abnormal instantaneous current peaks.

[0038] Set secondary trigger conditions based on frequency offset for regional power grid frequency regulation warnings, and identify power grid harmonic distortion areas through multi-scale frequency domain analysis algorithms;

[0039] For the prediction of global energy scheduling imbalance, the final trigger condition based on load deviation exceeding the limit is set, and the critical point of system instability is determined by calculating the energy distribution entropy value.

[0040] Preferably, the event-driven state machine includes nested conditional jump logic; the detection of abnormal instantaneous current peak value using a sliding time window statistical method includes:

[0041] Extract the mean, range and kurtosis characteristics of current monitoring data within a fixed time window;

[0042] The current peak probability distribution function is fitted by the generalized Pareto distribution model;

[0043] When the observed current value exceeds the 99.9% confidence interval upper limit of the probability distribution function, a local battery overload alarm instruction is triggered.

[0044] Preferably, the multi-scale frequency domain analysis algorithm is implemented using wavelet transform decomposition technology; the identifying of the power grid harmonic distortion area using the multi-scale frequency domain analysis algorithm includes:

[0045] The power grid frequency fluctuation signal is mapped into a multi-resolution wavelet coefficient matrix, and the high-frequency abnormal components in the matrix are extracted;

[0046] Based on the cluster analysis algorithm, clusters of high-amplitude coefficients with temporal and spatial correlations are merged to generate markers for potential harmonic interference areas.

[0047] Calculate the energy proportion and frequency offset product index of each marked area, and select the target interval that meets the preset frequency modulation level.

[0048] Preferably, the multi-source heterogeneous data fusion and outlier elimination processing are implemented by adopting a robust data cleaning strategy; the multi-source heterogeneous data fusion and outlier elimination processing on the real-time energy interaction data stream includes:

[0049] Establish a data mapping relationship between the current timing sequence of the edge computing node, the load frequency sequence of the cloud platform, and the temperature sampling sequence of the vehicle terminal;

[0050] The high-frequency noise components in the data stream of each node are separated by the empirical mode decomposition algorithm, and the low-frequency effective data are cross-modally aligned based on the robust principal component analysis algorithm;

[0051] A time series interpolation algorithm is used to compensate for the sampling interval deviation among multi-source data streams and generate a spatiotemporally consistent fusion energy management tensor.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] In terms of multidimensional energy data management, by collecting multidimensional data such as the real-time charge and discharge rates of battery packs, ambient temperature distribution characteristics, and grid load fluctuations, and employing robust data cleaning strategies (such as empirical mode decomposition, robust principal component analysis, and time series interpolation algorithms) to fuse and eliminate outliers from multi-source heterogeneous data, this method effectively overcomes the limitations of traditional single-source data processing models. This method not only removes high-frequency noise and compensates for sampling interval deviations, but also generates a spatiotemporally consistent fused energy management tensor, providing a high-precision data foundation for subsequent optimization decisions and improving the accuracy of energy status assessment.

[0054] The innovative design of the heterogeneous layered deployment architecture is an important breakthrough of the present invention. By defining the energy partitioning rules of the core dispatching area, buffer dispatching area and backup dispatching area, and configuring the low-latency response mode, global optimization mode and local closed-loop control mode for the edge computing nodes, cloud collaboration platform and on-board control terminal respectively, a heterogeneous layered architecture at the distributed computing level is constructed. This architecture enables the system to dynamically adjust the dispatching strategy according to the peak and valley periods of the power grid load: the edge nodes can quickly respond to local energy demands (such as instantaneous overload alarms), the cloud platform performs global energy optimization scheduling (such as cross-regional load distribution), and the on-board terminal ensures the real-time performance of local control, which significantly improves the response speed and global optimization capabilities of the system and enhances the flexibility of the coordinated dispatching of mobile energy storage vehicles and the power grid.

[0055] The construction and solution of the dynamic optimization decision model realizes multi-objective collaborative optimization. By combining the battery degradation model with the grid supply and demand balance equation, a dynamic optimization decision model including the charge and discharge efficiency matrix, the temperature compensation coefficient tensor and the load distribution vector is constructed, and solved using a multi-objective iterative solver (integrated energy efficiency optimization, thermal management and task scheduling modules) and a parallel gradient descent algorithm. This method can not only synchronously update the optimization residuals of energy conservation, thermodynamic diffusion and task constraints, but also adjust the learning rate through an adaptive momentum mechanism, thereby realizing multi-objective collaboration of energy efficiency improvement, thermal management optimization and load balancing distribution. For example, the battery life in the energy efficiency optimization module is

[0056] The life prediction model can extend the battery cycle life, the phase change material thermal buffer model of the thermal management module can reduce the damage to the battery caused by temperature fluctuations, and the real-time preemptive algorithm of the task scheduling module can improve the load distribution efficiency, thereby improving the energy utilization efficiency and system operation reliability as a whole.

[0057] The three-level adaptive response instruction sequence generation mechanism significantly enhances the system's robustness and safety. Through an event-driven state machine and nested conditional jump logic, primary, secondary, and final trigger conditions are set for local battery overload, regional grid frequency regulation, and global energy dispatch imbalance. Technologies such as sliding time window statistics, multi-scale frequency domain analysis (wavelet transform), and energy distribution entropy calculation are used to achieve precise detection. For example, current peak detection based on the generalized Pareto distribution accurately identifies overload anomalies, wavelet transform combined with cluster analysis locates areas of harmonic interference, and energy distribution entropy predicts the critical point of system instability. This hierarchical response mechanism not only triggers alarms promptly but also dynamically adjusts response strategies based on the severity of the anomaly, effectively reducing false alarms and missed alarms, thereby enhancing the system's adaptability and fault warning capabilities under complex operating conditions.

[0058] Furthermore, multi-module coupling and cross-layer collaborative design (such as state-sharing interfaces for transferring corrections and staged learning rate adjustment for parallel gradient descent algorithms) further enhance system integration and optimization efficiency. Through real-time data sharing and coordinated parameter updates, each functional module forms a closed-loop optimization feedback mechanism, making energy management more intelligent and efficient. This provides technical support for the large-scale application of mobile energy storage vehicles in smart grids, with significant economic value and social significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a working principle diagram of the energy management method for mobile energy storage vehicles based on intelligent algorithms according to the present invention;

[0060] Figure 2 Flowchart for solving the dynamic optimization decision model;

[0061] Figure 3 This is the structural diagram of the multi-objective iterative solver;

[0062] Figure 4 A diagram for generating three-level adaptive response instructions;

[0063] Figure 5 This is a processing diagram for multi-source heterogeneous data fusion. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0065] See also Figure 1-Figure 5 The present invention relates to a mobile energy storage vehicle energy management method based on an intelligent algorithm, and the specific implementation steps are as follows:

[0066] The mobile energy storage vehicle's battery pack's real-time charge and discharge rates, ambient temperature distribution characteristics, and grid load fluctuation characteristics are collected in real time through onboard sensors and grid access equipment. The battery pack's charge and discharge rates are acquired through current sensors and power metering modules, while ambient temperature distribution characteristics are collected through a network of temperature sensors located in key locations such as the battery compartment and motor compartment. Grid load fluctuation characteristics are synchronized with grid load curves and peak and valley time period divisions in real time through the grid data interface.

[0067] Based on the dynamic response requirements of the energy dispatch unit, a heterogeneous layered deployment architecture consisting of edge computing nodes, a cloud collaboration platform, and an onboard control terminal is constructed. This architecture implements functional division at different levels through distributed computing layers to meet the dual requirements of real-time performance and global optimization.

[0068] The system simultaneously acquires real-time energy interaction data streams from each node within a heterogeneous, layered deployment architecture, including current time series data from edge computing nodes, load frequency data from the cloud platform, and temperature sampling data from the vehicle control terminal. Multi-source heterogeneous data fusion and outlier removal are performed on these data streams. By establishing cross-node data mappings, separating high-frequency noise, aligning low-frequency valid data across modalities, and compensating for sampling interval deviations, a spatiotemporally consistent fused energy management tensor is generated.

[0069] A dynamic optimization decision-making model is constructed by combining the battery degradation model with the grid supply and demand balance equation. The collaborative constraints of the charge and discharge efficiency matrix, the temperature compensation coefficient tensor, and the load distribution vector in the model are solved jointly. Specifically, a multi-objective iterative solver is used to solve the model, comprehensively considering energy conservation, thermodynamic diffusion, and task scheduling constraints to optimize the charging and discharging strategies and energy distribution of energy storage vehicles.

[0070] Based on the boundary trigger conditions of collaborative constraints, a three-level adaptive response instruction sequence is generated, including local battery overload warnings, regional grid frequency regulation warnings, and global energy dispatch imbalance predictions. Instruction generation is achieved through an event-driven state machine, with differentiated trigger conditions and detection methods set for different levels of warnings and alerts.

[0071] Example 1:

[0072] When implementing a heterogeneous, layered deployment architecture, it's first necessary to define energy zoning rules for core, buffer, and backup dispatch areas based on the peak-valley period division strategy inherent in grid load fluctuations. Grid load fluctuations are determined through statistical analysis of historical load data. Peak-valley period division can be determined using a time-of-use electricity pricing mechanism or load forecasting algorithm. For example, frequency domain analysis of the grid load curve using a Fourier transform can identify daily peak load periods (such as daytime industrial peaks), valley periods (such as nighttime low residential load periods), and flat periods. The core dispatch area corresponds to high-load areas or critical power supply nodes during peak and valley periods, prioritizing energy supply stability. Examples include central business districts and industrial parks in cities. These areas place extremely high demands on energy storage vehicles for real-time response during peak load periods, and any interruption in energy supply can result in significant economic losses or service disruptions. The buffer dispatch area, located between the core dispatch area and the backup dispatch area, serves as an energy transfer and regulation hub. It can be located in a moderately dense, geographically central area, such as a city suburb or near a transportation hub. Its function is to quickly replenish energy when energy demand surges in the core dispatch area, or to temporarily store excess energy before the backup dispatch area is activated, thereby mitigating drastic fluctuations in energy supply and demand. As a system redundancy reserve, the backup dispatch area is typically located in areas far from core load centers but with good transportation and energy access, such as suburban energy storage stations or areas rich in renewable energy. It is activated only when both the core and buffer dispatch areas are unable to meet energy demand, in response to extreme load events or large-scale equipment failures.

[0073] Based on the above energy partitioning rules, differentiated working modes are configured for edge computing nodes, cloud collaboration platforms, and on-board control terminals. Edge computing nodes are deployed in physical locations close to energy storage vehicles or grid access points, such as on-board edge servers or edge computing devices at distributed energy sites. Their hardware architecture usually includes high-performance processors, real-time operating systems, and high-speed communication modules. To meet the low-latency response requirements of the core scheduling area, edge computing nodes are configured with millisecond-level real-time operating systems (such as QNX or VxWorks), and their data processing processes are optimized: when a sudden change signal of the battery pack's charge and discharge rate is received (such as the current change rate exceeds a preset threshold), the edge computing node must complete data collection, anomaly detection, and control instruction generation within 10 milliseconds. The specific implementation method is: periodic data sampling is triggered by a hardware timer with a sampling period set to 1 millisecond. After each sample, digital filtering (such as finite impulse response filtering) is immediately performed to remove high-frequency noise. Subsequently, an anomaly detection algorithm based on threshold comparison is executed. If an anomaly is detected, a predefined control strategy library (such as a proportional-integral-derivative control algorithm) is immediately invoked to generate instructions for adjusting the charge and discharge power. These instructions are sent to the on-board actuators via a hard real-time communication interface (such as the CAN bus or industrial Ethernet), ensuring that the instruction transmission delay is less than 5 milliseconds. For edge computing nodes in the buffer and backup scheduling areas, the response time can be appropriately relaxed to the 100 millisecond level to reduce hardware costs and computing resource consumption. Their data processing flow can also include data caching and batch processing. For example, data can be collected every 100 milliseconds and averaged over a sliding window to reduce over-response to sudden noise.

[0074] The cloud-based collaborative platform, the core of global energy optimization, is deployed in a cloud computing data center and features large-scale data storage, distributed computing, and cross-regional coordination capabilities. Its software architecture adopts a microservices design pattern and includes multiple independent service modules, including energy forecasting, optimized scheduling, and security monitoring. To meet the global optimization needs of the core dispatching area, the cloud-based collaborative platform integrates real-time data from multiple edge computing nodes (such as the status information of hundreds to thousands of mobile energy storage vehicles). Using machine learning models (such as long-short-term memory networks), it predicts grid load trends and energy status of energy storage vehicles for the next 15 minutes to one hour. It then solves the global energy optimization problem using a mixed-integer programming algorithm, with objective functions such as minimizing energy transmission costs and maximizing grid peak-valley smoothing. For example, 30 minutes before the peak load arrives, the cloud-based collaborative platform, based on the predicted results, sends centralized charging or discharging commands to mobile energy storage vehicles surrounding the core dispatching area. By optimizing the charge and discharge power distribution among the vehicles, the platform reduces grid load fluctuations by over 20%. (This is a logical description only and does not include specific experimental data.) For the buffer dispatch area, the cloud-based collaborative platform's optimization cycle can be extended to one hour, focusing on inter-regional energy balance. For example, this platform coordinates energy exchange between energy storage vehicles within the buffer dispatch area and distributed photovoltaic and wind farms to achieve regional energy self-sufficiency. The cloud-based collaborative function for the backup dispatch area is primarily used for long-term energy planning. For example, it adjusts the charging schedule of energy storage vehicles in the backup dispatch area based on seasonal energy demand changes to ensure sufficient energy reserves in extreme situations.

[0075] The onboard control terminal (OCT) is the core unit for implementing local energy closed-loop control in energy storage vehicles. It is typically integrated into the vehicle's central control system and includes a microcontroller, sensor interface, actuator driver circuitry, and a human-machine interface. For local closed-loop control, the OCT must collect real-time data on battery pack voltage, current, and temperature. Using a PID controller or model predictive control algorithm, it adjusts actuators in the battery management system (BMS), such as the charge and discharge switches and cooling fan speed, to maintain the battery pack operating within a safe range. The specific control process is as follows: A temperature sensor network (such as the DS18B20 digital temperature sensor) collects battery cell temperatures in real time and transmits the data to the OCT every 50 milliseconds. The OCT first uses a Kalman filter to fuse the multi-sensor data to eliminate measurement noise, then calculates the average battery pack temperature and temperature gradient. If the average temperature exceeds a preset threshold of 25°C, the cooling fans are activated and the speed of each fan is adjusted based on the temperature gradient to ensure a uniform temperature distribution across the battery pack. If the temperature continues to rise to 35°C, a power reduction charging strategy is triggered, reducing the charge and discharge power by 30% to minimize heat generation. At the same time, the on-board control terminal must also maintain communication with the edge computing node and the cloud collaboration platform, receive upper-level scheduling instructions and feedback local status information. For example, when receiving an emergency discharge instruction from the edge computing node, it must switch to discharge mode within 50 milliseconds and report the discharge power and remaining battery capacity in real time.

[0076] In the coordinated operation of energy partitioning rules and node operating modes, data exchange between various layers is achieved through standardized communication protocols (such as OPC UA or MQTT). Data packets uploaded by edge computing nodes to the cloud collaboration platform contain fields such as timestamp, energy storage vehicle ID, location information, charge and discharge status, and battery health status. The upload frequency ranges from once per second (in the core scheduling area) to once per minute (in the backup scheduling area). Command packets issued by the cloud collaboration platform to edge computing nodes and onboard control terminals contain fields such as command type, target device ID, parameter values, and execution time. For urgent commands in the core scheduling area, a reliable transmission protocol (such as TCP) is used to ensure command loss, with transmission latency limited to less than 100 milliseconds. For non-urgent commands, UDP can be used to improve transmission efficiency. Furthermore, to address abnormal situations such as communication interruptions, each node has local caching and offline operation capabilities. For example, edge computing nodes can store the last hour of historical data and preset control strategies. If connection with the cloud is lost, they automatically switch to a locally predefined energy scheduling mode to maintain basic system operation.

[0077] The implementation of the entire heterogeneous layered deployment architecture also requires consideration of the compatibility between hardware selection and software optimization. Edge computing nodes can utilize ARM-based embedded processors (such as the Nvidia Jetson series), which combine high performance with low power consumption, making them suitable for deployment in vehicles or distributed scenarios. The cloud collaboration platform is built on an x86 server cluster and utilizes containerization technologies (such as Docker) to achieve elastic service scalability. The in-vehicle control terminal uses an automotive-grade microcontroller (such as Infineon's AURIX) to meet reliability requirements in harsh environments such as high temperature and vibration. At the software level, the real-time operating system of the edge computing node requires kernel pruning to remove unnecessary services to reduce memory usage and interrupt latency. The optimization algorithms of the cloud collaboration platform must utilize parallel computing technologies (such as the Spark distributed computing framework) to shorten the time required for large-scale data processing. The control algorithms of the in-vehicle control terminal must use code optimization and fixed-point processing to reduce computational complexity and ensure real-time operation within limited hardware resources.

[0078] By defining the aforementioned energy partitioning rules and differentially configuring heterogeneous nodes, this architecture enables precise scheduling of mobile energy storage vehicles across different timescales and spatial ranges: edge computing nodes address local real-time control issues, cloud-based collaborative platforms handle global optimization and cross-regional coordination, and onboard control terminals ensure local system stability. These three elements collaborate to form a layered, distributed energy management system, effectively improving the responsiveness and operational efficiency of mobile energy storage vehicles in complex power grid environments.

[0079] Example 2:

[0080] The simultaneous solution process of the dynamic optimization decision model is implemented through a multi-objective iterative solver. Its core lies in the coupled solution of multi-dimensional constraints such as charging and discharging efficiency, temperature field distribution, and load distribution to generate optimal decision parameters that take into account energy efficiency, thermal safety, and scheduling efficiency. The following is a detailed description of the implementation method with reference to specific scenario examples:

[0081] Consider a city's power grid entering its peak load period (11:00 AM - 3:00 PM) during the summer midday hours. Multiple mobile energy storage vehicles in the area must coordinate and participate in grid frequency regulation. In this scenario, a dynamic optimization decision-making model must solve for the collaborative constraints between the charge-discharge efficiency matrix, the temperature compensation coefficient tensor, and the load distribution vector based on real-time data collected from the battery pack's charge and discharge rates (e.g., a vehicle's current charging power is 50 kW), ambient temperature distribution (battery compartment average temperature 28°C, with local hotspots reaching 32°C), and grid load fluctuations (current regional load exceeding 110% of rated capacity).

[0082] Charge and discharge efficiency matrix loading and capacity attenuation update:

[0083] The charge-discharge efficiency matrix stores energy conversion efficiency parameters at different charge and discharge powers. For example, when the charging power is ≤30kW, the efficiency is 95%. Within the 30-60kW range, the efficiency drops linearly to 88%. Above 60kW, the efficiency enters a high-loss range, dropping to 80%. These matrix parameters are pre-calibrated through battery charge-discharge experiments and stored in the energy efficiency optimization module's database. During the solution process, the current charge and discharge power of each vehicle is first mapped to a specific entry in the matrix. For example, 50kW charging power corresponds to an efficiency value of 88%. Simultaneously, the capacity decay factor is updated based on the battery state of health (SOH) model. This model calculates the capacity decay rate based on data such as the cumulative number of charge and discharge cycles and operating temperature history. For example, if a battery pack has completed 500 cycles and its current SOH is 85%, the capacity decay factor is set to 0.85. This factor is used to modify the available capacity parameter in the energy conservation equations, enabling the model to reflect the impact of battery aging on charge and discharge efficiency.

[0084] Temperature compensation coefficient tensor embedding and heat conduction calculation:

[0085] The temperature compensation coefficient tensor describes the impact of temperature on battery performance. It includes correction factors for charge and discharge efficiency and internal resistance variation for different temperature ranges. For example, when the battery temperature is between 25-30°C, the charge and discharge efficiency correction factor is 1.0 (no compensation). For every 1°C increase in temperature, the charge efficiency correction factor decreases by 0.5%, and the discharge efficiency correction factor decreases by 0.3%. After embedding this tensor into the thermodynamic diffusion equation, the anisotropic thermal conductivity coefficient is calculated using Fourier's law. Taking the temperature distribution within the battery compartment as an example, assuming a temperature gradient of 2°C / m along the battery pack's length (x-axis), 1.5°C / m along its width (y-axis), and 3°C / m along its height (z-axis), the anisotropic thermal conductivity coefficients are 0.2 W / (m·K) along the x, y, and z axes, respectively. 0.15 W / (m·K) and 0.25 W / (m·K) along the z, y, and z axes, respectively. By calculating the heat flux density in different directions, the hot spot area of the battery pack can be determined (such as the local hot spot of 32°C mentioned above), and then the temperature compensation coefficient can be adjusted. For example, a -5% charge and discharge power penalty can be imposed on the battery cells in the hot spot area to suppress further temperature increase.

[0086] Load distribution vector discretization and task constraint matrix construction:

[0087] The load distribution vector corresponds to the energy dispatching task issued by the grid dispatching center. For example, within one hour, a certain area requires energy storage vehicles to provide a total discharge capacity of 200MWh. This vector is discretized into a time series task queue, with a time slice of 15 minutes, and decomposed into four task units (each unit needs to provide 50MWh). Each task unit contains parameters such as the start time, target power, and duration. For example, the first time slice (11:00-11:15) requires a vehicle to discharge at a power of 60kW. The task constraint matrix is constructed in combination with the dynamic priority scheduling strategy. The priority is determined by factors such as the urgency of the task, the remaining capacity of the energy storage vehicle, and the location distance. For example, energy storage vehicles within 5 kilometers of the load center and with a remaining capacity of ≥80% are given high priority and must perform the discharge task first; vehicles more than 10 kilometers away or with a remaining capacity of <30% have a lower priority and can be used as backup forces. The task constraint matrix also needs to consider the physical constraints of the energy storage vehicle, such as the maximum charging and discharging power (the maximum discharge power of a certain vehicle is 70kW), the upper and lower limits of SOC (20%-90%), etc., to ensure that the assigned tasks are within the vehicle's capabilities.

[0088] Iterative optimization process of parallel gradient descent algorithm:

[0089] Taking the initial optimization input as an example, assume that the charge and discharge efficiency matrix for the current iteration is [0.95, 0.88, 0.80] (corresponding to the low, medium, and high power ranges), the temperature compensation coefficient tensor is [1.0, 0.995, 0.99] (corresponding to 25°C, 26°C, and 27°C), and the load distribution vector is [60kW, 55kW, 50kW, 45kW] (discharge power for four time slices). The energy efficiency optimization module calculates the energy residual norm, which is the square root of the sum of the squares of the difference between the actual energy output and the task demand. Assume that the demand for the first time slice is 50MWh (the energy equivalent of 60kW power for 15 minutes is 15kWh × 4 = 60kWh = 0.06MWh, which is clearly insufficient, but this is just a logical example). The residual norm reflects this gap and drives the model to adjust power distribution. The thermal management module calculates the residual norm of thermal balance. If the deviation between the currently calculated battery temperature field and the preset safe temperature range (20-35°C) exceeds a threshold (for example, if the average temperature is 28°C, close to the upper limit), a temperature correction signal is generated. The task scheduling module calculates the residual norm of scheduling delay. If a vehicle's task execution is delayed by more than 5 minutes due to a path planning issue, the priority of its subsequent tasks needs to be adjusted.

[0090] The dynamically weighted geometric mean of the three residual norms is used to adaptively adjust the learning rate decay coefficient. For example, if the energy efficiency residual accounts for 60%, the thermal residual accounts for 30%, and the scheduling residual accounts for 10%, and the current residual norm converges rapidly (e.g., the residual decreases by >10% between two consecutive iterations), the learning rate amplification mechanism is triggered. The baseline learning rate (e.g., 0.1) is multiplied by the amplification factor (e.g., 1.2) to obtain a new learning rate (0.12) to accelerate the optimization process. If the convergence rate is slow (decline by <5%), the learning rate (0.09) is reduced by a decay factor (e.g., 0.9) to avoid optimization overshoot. If the residual norm exhibits periodic fluctuations (e.g., the residual decreases then increases for three consecutive iterations), indicating a local optimum, a random restart strategy is activated, randomly adjusting the initial charge and discharge power by ±10% and the temperature compensation coefficient by ±5% to re-search for the optimal solution.

[0091] During the iteration process, each module exchanges information in real time through a state-sharing interface. For example, if the energy efficiency optimization module discovers that a vehicle's efficiency drops sharply during high-power discharge, it transmits a charge / discharge efficiency correction (e.g., -8%) to the thermal management module. The thermal management module then adjusts the vehicle's temperature threshold (e.g., lowering the upper safety limit from 35°C to 32°C) and sends a power limit signal to the task scheduling module. The task scheduling module then reallocates tasks for that vehicle, shifting some of the load to other, more efficient vehicles. This cross-module coupling mechanism allows the model to comprehensively consider multi-dimensional constraints, avoiding system imbalances caused by single-objective optimization.

[0092] After several iterations (e.g., 10-20), when the dynamic weighted geometric mean falls below a preset optimization threshold (e.g., 0.05), the iterations are terminated and the optimized decision parameters are output. For example, the charge and discharge efficiency matrix is adjusted to [0.95, 0.89, 0.82] (efficiency in the mid-power range is improved by 1%), the temperature compensation coefficient tensor is updated to [1.0, 0.996, 0.992] (efficiency loss is reduced with every 1°C temperature increase), and the load distribution vector is optimized to [65kW, 60kW, 55kW, 40kW] (power is increased in the first two time slices to cope with peak loads and reduced in the last two time slices to prevent battery overload). These parameters are transmitted to the onboard control terminal via the edge computing node, guiding the energy storage vehicle in performing specific charging and discharging operations.

[0093] The entire solution process must be completed within minutes to adapt to rapid changes in grid load. Through the collaborative operation of multi-objective iterative solvers, the dynamic optimization decision-making model can find a balance between energy efficiency, thermal safety, and task scheduling. This ensures that mobile energy storage vehicles, when participating in grid frequency regulation, meet load demands while avoiding battery overheating or excessive loss, achieving dynamic optimization of energy management.

[0094] Example 3:

[0095] The multi-objective iterative solver, as the core execution unit of the dynamic optimization decision model, consists of an energy efficiency optimization module, a thermal management module, and a task scheduling module. Each module is coupled across modules through a state sharing interface, forming a closed-loop system for multi-dimensional constraint collaborative optimization. The following describes the implementation method in detail based on the specific operation logic:

[0096] 1. Module function definition and basic configuration

[0097] The energy efficiency optimization module integrates a battery life prediction model, which uses a linear regression algorithm to construct a life prediction equation based on parameters such as the number of battery charge and discharge cycles, operating temperature, and depth of charge and discharge (DOD). For example, the model input parameters include: the current cycle number , average operating temperature , average charge and discharge depth , the output is the remaining cycle life of the battery , its simplified expression is:

[0098]

[0099] in, The module calculates the impact of different charge and discharge strategies on battery life based on the model, such as deep discharge (DOD>80%) or high temperature environment ( ) will significantly reduce , thereby introducing the life loss cost term into the optimization objective to suppress the short-sighted high-power charging and discharging behavior.

[0100] The thermal management module has a built-in phase change material thermal buffer model, which simulates the heat absorption and release characteristics of phase change materials during the melting / solidification process. Taking paraffin-based phase change materials as an example, their phase change temperature range is 28-32°C and the latent heat is 200kJ / kg. The model input parameters include: phase change material mass , current temperature , heat flux , the output is the phase change progress of the material and temperature change rate When the battery pack temperature rises to the lower limit of the phase change temperature (28°C), the phase change material begins to absorb heat and melt. ( Latent heat (latent heat) can offset some of the battery's heat generation, slowing the rate of temperature rise. When the temperature drops below 28°C, the material solidifies, releasing heat and maintaining a stable battery pack temperature. This model calculates temperature evolution trends at different charge and discharge powers, providing thermal constraints for energy efficiency optimization.

[0101] The task scheduling module uses a real-time preemptive scheduling algorithm that assigns a priority to each energy scheduling task. , priority is determined by task urgency , delay tolerance , Energy storage vehicle location distance For example, emergency frequency regulation tasks (such as grid frequency deviation exceeding ±0.5Hz) Set to the highest level (level 10), delay tolerance Set to 5 minutes; non-urgent valley charging tasks is level 1, Set to 1 hour. When the algorithm receives a new task, it immediately compares its priority with the priority of the currently executing task. If the new task has a higher priority, the current task is paused and the new task is switched to ensure the real-time performance of the high-priority task.

[0102] 2. Cross-module coupling mechanism and state transfer

[0103] The state sharing interface serves as a channel for data exchange between modules, passing three types of corrections at each iteration:

[0104] Charge and discharge efficiency correction : Calculated by the energy efficiency optimization module, it reflects the impact of the current charge and discharge strategy on the battery life. For example, when a car is discharged at 70kW power, the battery life prediction model shows that the remaining cycle life is reduced by 100 times, and the module generates This value (indicating a virtual 5% reduction in efficiency to suppress that power level) is then passed to the thermal management module and the task scheduling module. The thermal management module uses this correction as a basis for adjusting temperature boundary conditions, for example, lowering the vehicle's upper temperature safety limit from 35°C to 32°C. The task scheduling module then lowers the priority of tasks at that power level to avoid frequent use of high-loss operating conditions.

[0105] Temperature gradient correction : Calculated by the thermal management module based on the temperature field distribution in the battery compartment. For example, if the temperature gradient in the x-axis direction of a vehicle's battery pack reaches 3°C / m (exceeding the safety threshold of 2°C / m), the module generates The correction value (dimensionless, representing the degree of gradient overshoot) is passed to the energy efficiency optimization module and the task scheduling module. The energy efficiency optimization module adjusts the charge and discharge efficiency matrix based on this correction value, imposing a power penalty on areas with high temperature gradients (for example, reducing the charge and discharge power of battery cells in the corresponding area by 10%). The task scheduling module then prioritizes low-power tasks (such as 20kW charging) for the vehicle to reduce heat generation and balance the temperature field.

[0106] Task priority modifier :Calculated by the task scheduling module based on the matching degree between the task execution status and the vehicle capacity. For example, if the remaining capacity of a vehicle is only 25% (below the safety limit of 30%), and the currently assigned discharge task consumes 15% of the capacity, the module generates The priority is reduced by 3 levels and then passed to the energy efficiency optimization module and the thermal management module. The energy efficiency optimization module limits the vehicle's discharge power limit (for example, from 60kW to 40kW) and triggers a charging recommendation. The thermal management module adjusts its temperature compensation coefficient, allowing a slightly higher operating temperature (for example, raising the upper limit to 38°C) to prioritize the remaining tasks and avoid interruptions due to insufficient capacity.

[0107] 3. Specific process of module collaborative optimization

[0108] Take a mobile energy storage vehicle cluster participating in power grid peak shaving as an example:

[0109] Initial parameter input: The cloud-based collaborative platform issues a task queue requiring the cluster to provide a total discharge capacity of 150 MWh between 2:00 PM and 3:00 PM, while ensuring that the battery life loss is less than the equivalent of five cycles and that the average battery pack temperature does not exceed 35°C. The energy efficiency optimization module calculates the initial charge and discharge power allocation plan (e.g., six vehicles each providing 25 MWh at 50 kW for one hour) based on the SOH of each vehicle (e.g., average SOH = 88%) and the number of historical cycles (average N = 450).

[0110] Cross-module interactive iteration:

[0111] After receiving the power allocation plan, the thermal management module calculates the temperature evolution of each vehicle's battery pack. Assuming that the battery compartment of vehicle A has poor ventilation, the calculated average temperature will reach 37°C (exceeding the threshold by 2°C) after 1 hour, generating And pass it to other modules. After the energy efficiency optimization module responds, the charging and discharging efficiency of car A is corrected to (original value 88%), the task scheduling module reduces its task priority from level 5 to level 3, triggering task reallocation (for example, vehicle B takes over vehicle A’s 5MWh task, and the power is increased to 60kW).

[0112] After the task scheduling module reallocates tasks, it needs to verify whether the remaining capacity of each vehicle meets the new task requirements. Assuming that the remaining capacity of vehicle B is 75%, after taking on the 5MWh task, the capacity will drop to 75%-(5 / battery capacity) (assuming the battery capacity is 50kWh, it will drop to 65%), meeting the ≥30% constraint. However, the energy efficiency optimization module calculated that when vehicle B discharges at 60kW power, the cycle life loss will increase by 1.2 times (exceeding the upper limit of 0.8 times allowed for a single task), generating The data is then passed to the thermal management module and the task scheduling module. The thermal management module then requires vehicle B to discharge no more than 55 kW, and the task scheduling module extends its task execution time to 1.1 hours (55 kW power, 55 × 1.1 = 60.5 kWh ≈ 6.05 MWh, slightly exceeding the required value, rounded off to 1.09 hours).

[0113] After three iterations, the dynamic weighted geometric mean of the three types of residual norms (energy efficiency residual, thermal residual, and scheduling residual) is lower than the preset threshold. Each module outputs the final parameters: the power of vehicle A is adjusted to 40kW (bearing 20MWh, taking 0.5 hours + subsequent charging replenishment), the power of vehicle B is 55kW (bearing 30MWh, taking 1.09 hours), and the power of the remaining vehicles is maintained at 50kW. The total mission completion time is controlled within 1 hour, while meeting the life and temperature constraints.

[0114] State feedback and closed-loop control: The optimized decision parameters are sent to the on-board control terminal through the edge computing node, and each vehicle performs real-time charging and discharging operations. The on-board sensors continuously collect data such as voltage, current, and temperature, and feed them back to the multi-objective iterative solver, forming a closed loop of "data acquisition-optimization calculation-instruction execution-state feedback". For example, when vehicle B is performing a 55kW discharge, real-time temperature monitoring shows an average temperature of 34°C and a temperature gradient of 1.8°C / m, both within the safe range and requiring no further adjustment. During the 40kW charging process of vehicle A, the temperature drops to 29°C. Based on the feedback data, the energy efficiency optimization module gradually restores its charging and discharging efficiency to 88% of the original value, and redistributes the reasonable load in subsequent tasks.

[0115] 4. Key Points for Hardware and Algorithm Adaptation

[0116] The real-time performance of the multi-objective iterative solver depends on hardware acceleration and algorithm optimization:

[0117] The battery life prediction model of the energy efficiency optimization module uses a table lookup method to accelerate the process. The life loss parameters under different working conditions are stored in a two-dimensional table (the horizontal axis is DOD and the vertical axis is temperature). During operation, linear interpolation is used to quickly obtain results, avoiding complex and time-consuming calculations.

[0118] The phase change material model of the thermal management module adopts a simplified lumped parameter method, treating the battery pack and phase change material as a single thermal node and ignoring the spatial distribution details. The calculation speed is increased by more than 50%, meeting millisecond-level iteration requirements.

[0119] The real-time preemptive algorithm of the task scheduling module is implemented through the hardware interrupt mechanism. When a high-priority task arrives, the CPU immediately responds to the interrupt, saves the current task context (register status, memory pointer, etc.) and switches to the new task. The interrupt response time is controlled within 10 microseconds to ensure the real-time performance of task scheduling.

[0120] Through the above-mentioned modular design and collaborative mechanism, the multi-objective iterative solver can achieve dynamic optimization of the energy management of mobile energy storage vehicles under complex constraints, taking into account the grid scheduling requirements, battery health status and thermal safety boundaries, and providing core technical support for the efficient and reliable operation of mobile energy storage systems.

[0121] Example 4:

[0122] In the actual operation scenario of a certain city's power grid, the phased learning rate adjustment strategy of the parallel gradient descent algorithm achieves dynamic optimization of the energy management of mobile energy storage vehicles through an adaptive momentum mechanism. When the grid load reaches its peak in the summer afternoon, multiple mobile energy storage vehicles in the area need to cooperate in peak load regulation. In the initial stage, the system uses the charge and discharge efficiency matrix, temperature compensation coefficient tensor, and load distribution vector of each vehicle as optimization input. For example, the current charging power of a vehicle is 50kW, which corresponds to an efficiency value of 88% in the charge and discharge efficiency matrix; the average temperature of the battery compartment is 28°C, and the correction factor corresponding to this temperature in the temperature compensation coefficient tensor is 0.995; in the load distribution vector, the vehicle's task during the 14:00-14:15 period is to deliver 60kW of electricity to the grid.

[0123] The energy efficiency optimization module first calculates the energy residual norm, which is the deviation between the actual energy output and the task requirements. If the vehicle's actual discharge during this period is only 55kW (due to decreased efficiency caused by increased battery internal resistance), the energy residual is 5kW. The thermal management module simultaneously calculates the thermal balance residual norm, obtaining the temperature value of each monitoring point through a network of temperature sensors arranged in the battery compartment. Assuming that the temperature at a certain monitoring point is 32°C, exceeding the safety threshold of 30°C, a thermal residual signal is generated. The task scheduling module calculates the scheduling delay residual norm. If the vehicle fails to arrive at the designated area on time due to traffic congestion, resulting in a 3-minute delay in task execution, the scheduling residual is triggered.

[0124] Based on the dynamically weighted geometric mean of the three types of residual norms, the system adaptively adjusts the learning rate decay coefficient for the next iteration. If, in the current iteration, the energy residual accounts for 60%, the thermal residual accounts for 30%, and the scheduling residual accounts for 10%, and the residual norm convergence rate exceeds the preset threshold (for example, the residual decrease between two consecutive iterations is greater than 10%), the optimization process is progressing smoothly. At this point, the system multiplies the learning rate amplification factor by the baseline learning rate to obtain a new learning rate. For example, if the baseline learning rate is 0.1 and the amplification factor is 1.2, the new learning rate is adjusted to 0.12 to accelerate convergence.

[0125] After adjusting the learning rate, the system recalculates all optimization parameters. The energy efficiency optimization module adjusts the charge and discharge power distribution based on the new learning rate, increasing the vehicle's mission power to 65kW during the 2:15-2:30 PM period to compensate for the energy shortfall during the previous period. The thermal management module simultaneously adjusts the cooling strategy, increasing the fan speed to keep the battery compartment temperature within a safe range. The task scheduling module replans the vehicle's route to avoid congested roads and ensure that subsequent tasks are completed on time.

[0126] As iterations progress, if the residual norm convergence rate enters a plateau (decreasing by less than 5%), the optimization process enters the fine-tuning phase. At this point, the system multiplies the learning rate decay factor by the baseline learning rate to reduce the learning rate. For example, if the decay factor is 0.9 and the baseline learning rate is 0.1, the new learning rate is adjusted to 0.09 to improve optimization accuracy and prevent the algorithm from oscillating near the optimal solution.

[0127] In some cases, the iterative process may become trapped in a local optimum, manifesting as periodic fluctuations in the residual norm. For example, in three consecutive iterations, the residual norm first decreases and then increases. In this case, the system activates a random restart strategy, randomly adjusting the initial charge and discharge power by ±10% and the temperature compensation coefficient by ±5%. For example, the vehicle's charge and discharge power is randomly adjusted to 55kW and the temperature compensation coefficient to 0.99, and a new search for the optimal solution is conducted to escape the local optimum.

[0128] Throughout the optimization process, each module exchanges information in real time via a shared state interface. When the energy efficiency optimization module detects that increasing power will cause the battery temperature to rise, it immediately transmits a charge and discharge efficiency correction to the thermal management module. The thermal management module adjusts the temperature threshold accordingly and sends a power limit signal to the task scheduling module. The task scheduling module then reallocates tasks to ensure a balance between energy efficiency, thermal safety, and scheduling efficiency.

[0129] After multiple iterations, when the dynamic weighted geometric mean falls below a preset optimization threshold, the system determines that the iterative process has converged, terminates the iteration, and outputs the optimized decision parameters. For example, the charge and discharge efficiency matrix is adjusted to [0.95, 0.89, 0.82], the temperature compensation coefficient tensor is updated to [1.0, 0.996, 0.992], and the load distribution vector is optimized to [65kW, 60kW, 55kW, 40kW]. These parameters are transmitted to the on-board control terminal via the edge computing node, guiding the energy storage vehicle to perform specific charging and discharging operations.

[0130] In actual operation, the system continuously repeats this optimization process based on real-time data. For example, when grid load fluctuates, the system recalculates the energy residual norm; when ambient temperature changes, the thermal management module updates the thermal balance residual norm. Through this continuous optimization, the system can adapt to various complex scenarios, ensuring that mobile energy storage vehicles, when participating in grid regulation, both meet load demands and safeguard the safety and lifespan of their batteries.

[0131] The entire optimization process requires no human intervention and is fully automated. From data acquisition to decision output, the entire cycle is controlled within seconds to minutes, meeting the requirements of real-time grid dispatch. By utilizing a phased learning rate adjustment strategy based on a parallel gradient descent algorithm, the system significantly improves convergence speed while maintaining optimization accuracy, providing efficient and reliable decision support for energy management in mobile energy storage vehicles.

[0132] Example 5:

[0133] The three-level adaptive response instruction sequence is generated by an event-driven state machine, which implements hierarchical monitoring of mobile energy storage vehicles and power grid systems based on multi-dimensional trigger conditions. The following describes the implementation method in detail based on specific scenarios:

[0134] 1. Local battery overload alarm (primary trigger condition)

[0135] A mobile energy storage vehicle performs a rapid discharge task during peak grid load periods. Onboard current sensors collect real-time battery pack output current data. The system extracts the mean, range, and kurtosis characteristics of the current monitoring data within a fixed time window (e.g., 100 milliseconds). For example, if five current samples collected at a certain moment are 58A, 62A, 65A, 59A, and 61A, the calculated mean is 61A, the range is 7A (the difference between the maximum value of 65A and the minimum value of 58A), and the kurtosis value is 3.2 (reflecting the degree of current spikes).

[0136] The current peak probability distribution function is fitted using a generalized Pareto distribution model, which describes the probability distribution of extreme current events exceeding a certain threshold. Assuming the current threshold in historical data is set at 60A, the model determines the distribution parameters using maximum likelihood estimation and calculates the upper limit of the 99.9% confidence interval to be 68A. If the measured current value reaches 70A (exceeding the upper limit of the confidence interval) at a certain moment, the system determines a sudden current change and triggers a local battery overload alarm.

[0137] After the alarm is triggered, the event-driven state machine executes nested conditional jump logic: first, a power reduction command is sent to the vehicle control terminal, reducing the discharge power from 60kW to 40kW to reduce the current. Simultaneously, a sliding time window extension mechanism is activated, extending the monitoring window from 100 milliseconds to 500 milliseconds to continuously monitor current fluctuation trends. If subsequent monitoring shows that the current drops below 60A and remains stable, the current power is maintained. If the current continues to exceed the threshold, a further command is triggered to force the battery cooling system to start, increasing the fan speed to the highest level. The abnormal status is reported to the edge computing node, requesting an adjustment to the vehicle's task allocation.

[0138] 2. Regional power grid frequency regulation warning (secondary trigger conditions)

[0139] In an industrial park power grid, the system collects grid frequency fluctuation signals (e.g., rated frequency 50 Hz) using monitoring equipment deployed at substations. A multi-scale frequency domain analysis algorithm uses wavelet transform decomposition technology to map the frequency signal into a multi-resolution wavelet coefficient matrix. For example, a three-layer wavelet decomposition of a 10-second frequency signal yields a low-frequency approximate component (reflecting the fundamental frequency trend) and a high-frequency detail component (reflecting harmonic content).

[0140] Extracting high-frequency anomalies from the matrix revealed a significant increase in the 250Hz frequency band (e.g., a jump from 0.5 to 3.0) during a certain period (2:20 PM to 2:25 PM), indicating the presence of fifth-order harmonic interference (250Hz = 5 × 50Hz). Using a cluster analysis algorithm, clusters of high-amplitude coefficients with temporal and spatial correlations were merged to generate markers for potential harmonic interference areas (e.g., Area A in the industrial park). The energy contribution (the proportion of harmonic energy to total energy) and the frequency offset product (frequency deviation × duration) were calculated for each marked area. Assuming that Area A's energy contribution reached 15% and the frequency offset product was 0.3Hz·s (exceeding the preset frequency regulation threshold of 0.2Hz·s), the system would trigger a regional power grid frequency regulation warning.

[0141] After the warning is generated, the state machine executes a hierarchical response: first, it sends the harmonic interference area location information to the cloud collaboration platform, requesting that mobile energy storage vehicles in the area be mobilized to perform harmonic suppression tasks. Simultaneously, it issues instructions to the relevant on-board control terminals to adjust the inverter control parameters of the energy storage vehicles, such as increasing the PWM modulation frequency from 10kHz to 15kHz to suppress the generation of the fifth harmonic. If the harmonic interference persists for more than 10 minutes, the energy storage vehicles in the backup dispatch area are further activated and moved to the area to enhance the frequency regulation capability of the local power grid.

[0142] 3. Prediction of Global Energy Scheduling Imbalance (Final Trigger Condition)

[0143] In cross-regional grid interconnection scenarios, the system collects load data and energy status of energy storage vehicles in real time across each region. The energy distribution entropy measures the uniformity of the system's energy distribution. Higher entropy values indicate more disordered distribution and a greater risk of instability. For example, at a certain moment, the calculated energy distribution entropy for the entire network is 2.8 (with a preset threshold of 2.5), indicating a dispersed energy distribution. Some areas (such as the city center) are overloaded, while the energy idle rate of suburban energy storage stations exceeds 40%.

[0144] When the entropy value exceeds the threshold, the system determines that there is a risk of global energy scheduling imbalance and triggers a final warning. The event-driven state machine first performs a traceability analysis of the energy distribution entropy value to locate the main contributing areas of the entropy increase (for example, the load in the commercial area exceeds 120% of the rated capacity, and the average SOC of energy storage vehicles in the suburbs reaches 90%). It then generates a global energy scheduling reorganization instruction: batch charging instructions are issued to mobile energy storage vehicles at suburban energy storage stations to replenish energy using low-priced electricity during off-peak periods; at the same time, emergency discharge instructions are sent to energy storage vehicles around the commercial area, increasing the charging and discharging power to 90% of the rated value. The traffic management system is coordinated to create a green passage for energy storage vehicles, shortening their time to reach the load center.

[0145] To avoid system shock, the state machine sets up an early warning delay mechanism: if the entropy value falls back to a safe range within 15 minutes after the threshold is triggered, the reorganization command is canceled; if it continues to rise, the energy storage vehicles in the backup dispatch area will be activated for cross-regional support. For example, 10 energy storage vehicles will be dispatched from a backup site 50 kilometers away to the commercial area, and are expected to arrive within 2 hours and participate in the dispatch.

[0146] 4. State Machine Nesting Logic and Instruction Collaboration

[0147] The nested conditional jump logic of the event-driven state machine is implemented through multi-level state nodes. For example, the local battery overload alarm state node contains sub-nodes such as "current returns to normal", "cooling system start", and "task reallocation", which execute corresponding jumps based on different feedback signals:

[0148] If the current drops below the threshold within 5 minutes after power reduction, it will switch to the "normal monitoring" state;

[0149] If the temperature continues to rise after the cooling system is started, the system will switch to the "Emergency Stop" state, cut off the charge and discharge circuits, and report the fault;

[0150] If the current still exceeds the limit after task reallocation, the system will jump to the "Exit Scheduling" state, mark the vehicle as awaiting maintenance, and remove it from the task queue.

[0151] In the collaborative scenario of regional power grid frequency regulation warning and global imbalance prediction, when a regional harmonic warning is triggered, if the cloud-based collaborative platform fails to effectively suppress harmonics within 30 minutes (for example, the harmonic energy ratio rises to 20%), the state machine automatically increases the response level, triggering the global imbalance prediction process, incorporating the frequency regulation needs of the region into the energy scheduling optimization model of the entire network, and achieving a higher level of balance through cross-regional energy allocation.

[0152] 5. Hardware Deployment and Real-time Guarantee

[0153] Local current monitoring relies on high-precision Hall current sensors (accuracy ±0.5%) with a sampling frequency of 10kHz to ensure capture of instantaneous current peaks. Grid frequency analysis uses dedicated signal acquisition cards (such as the NIPXIe-5124), which support simultaneous acquisition of multiple signals and real-time wavelet transform. Energy distribution entropy calculations are performed by a cloud server cluster, using GPU acceleration technology to control computational latency to less than 2 seconds.

[0154] The event-driven state machine uses a real-time operating system (RTX) to implement microsecond-level task scheduling, ensuring real-time trigger condition detection and instruction generation. For example, the latency for current mutation detection is less than 50 microseconds, the latency for frequency offset calculation is less than 10 milliseconds, and the latency for global entropy calculation is less than 1 second, meeting the requirements of real-time grid monitoring and rapid response.

[0155] Through the above three-level response mechanism and state machine logic, the system can provide layered warnings and interventions from local equipment anomalies to global system risks, forming a complete closed loop covering "detection-warning-response-adjustment", effectively improving the reliability of mobile energy storage vehicle energy management and the stability of power grid operation.

[0156] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0157] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A mobile energy storage vehicle energy management method based on intelligent algorithm, characterized in that: The method comprises: Collect multi-dimensional energy status data of the mobile energy storage vehicle, including the real-time charge and discharge rate of the battery pack, the ambient temperature distribution characteristics, and the grid load fluctuation characteristics; Based on the dynamic response requirements of the energy scheduling unit, a heterogeneous layered deployment architecture of edge computing nodes, cloud collaboration platforms, and vehicle control terminals is generated; Synchronously acquiring the real-time energy interaction data stream of each node under the heterogeneous layered deployment architecture, and performing multi-source heterogeneous data fusion and outlier elimination processing on the real-time energy interaction data stream; A dynamic optimization decision model is constructed by combining the battery degradation model with the grid supply and demand balance equation, and the collaborative constraint relationship among the charge and discharge efficiency matrix, the temperature compensation coefficient tensor, and the load distribution vector in the dynamic optimization decision model is solved jointly; Based on the boundary trigger conditions of the collaborative constraint relationship, a three-level adaptive response instruction sequence is generated, including local battery overload warning, regional power grid frequency regulation warning, and global energy scheduling imbalance prediction; The three-level adaptive response instruction sequence is generated by an event-driven state machine; the generation of the three-level adaptive response instruction sequence including local battery overload alarm, regional power grid frequency regulation warning and global energy scheduling imbalance prediction includes: For local battery overload alarms, a primary trigger condition based on current mutation is set, and a sliding time window statistical method is used to detect abnormal instantaneous current peaks. Set secondary trigger conditions based on frequency offset for regional power grid frequency regulation warnings, and identify power grid harmonic distortion areas through multi-scale frequency domain analysis algorithms; For the prediction of global energy scheduling imbalance, the final trigger condition based on load deviation exceeding the limit is set, and the critical point of system instability is determined by calculating the energy distribution entropy value.

2. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 1, characterized in that: The heterogeneous layered deployment architecture is characterized by distributed computing layers; The generation of a heterogeneous layered deployment architecture for edge computing nodes, cloud collaboration platforms, and vehicle-mounted control terminals includes: Based on the peak and valley period division strategy in the power grid load fluctuation characteristics, define the energy partitioning rules for the core dispatching area, buffer dispatching area and backup dispatching area; Based on the energy partitioning rules, a low-latency response mode is configured for the edge computing node, a global optimization mode is configured for the cloud collaboration platform, and a local closed-loop control mode is configured for the vehicle-mounted control terminal.

3. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 2, characterized in that: The dynamic optimization decision model is implemented by a multi-objective iterative solver; the simultaneous solution of the collaborative constraint relationship between the charge and discharge efficiency matrix, the temperature compensation coefficient tensor and the load distribution vector in the dynamic optimization decision model includes: Loading the charge and discharge efficiency matrix into the energy conservation equations of the multi-objective iterative solver, and updating the capacity decay factor based on the battery health state model; The temperature compensation coefficient tensor is embedded in the thermodynamic diffusion equation, and the anisotropic heat conductivity coefficient is calculated using Fourier's law. The load distribution vector is discretized into time series task queues, and the task constraint matrix is constructed in combination with the dynamic priority scheduling strategy; A parallel gradient descent algorithm is used to synchronously update the optimization residuals of the energy conservation equations, the thermodynamic diffusion equations and the task constraint matrix until a steady-state convergence condition is reached.

4. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 3 is characterized in that: The multi-objective iterative solver includes an energy efficiency optimization module, a thermal management module and a task scheduling module; the energy efficiency optimization module, the thermal management module and the task scheduling module are coupled across modules via a state sharing interface; The energy efficiency optimization module is configured with a battery life prediction model, the thermal management module is configured with a phase change material thermal buffer model, and the task scheduling module is configured with a real-time preemptive scheduling algorithm; The state sharing interface transmits the charge and discharge efficiency correction amount, the temperature gradient correction amount and the task priority correction amount to the adjacent functional modules in each iteration.

5. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 4 is characterized in that: The parallel gradient descent algorithm is implemented using a staged learning rate adjustment strategy; the parallel gradient descent algorithm is used to synchronously update the optimization residuals of the energy conservation equations, the thermodynamic diffusion equations, and the task constraint matrix, including: The charge and discharge efficiency matrix, temperature compensation coefficient tensor and load distribution vector of the current iteration step are used as initial optimization inputs; The energy residual norm is calculated through the energy efficiency optimization module, the thermal balance residual norm is calculated through the thermal management module, and the scheduling delay residual norm is calculated through the task scheduling module; Adaptively adjust the learning rate attenuation coefficient of the next iteration step based on the dynamic weighted geometric mean of the three types of residual norms; When the dynamic weighted geometric mean is lower than the preset optimization threshold, the iteration is terminated and the optimized decision parameters are output.

6. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 5 is characterized in that: The adjustment strategy of the learning rate attenuation coefficient is implemented based on an adaptive momentum mechanism; the adaptive adjustment of the learning rate attenuation coefficient of the next iteration step includes: If the residual norm convergence rate of the current iteration step is higher than the rate threshold, the learning rate amplification factor is multiplied by the baseline learning rate as a new parameter; If the residual norm convergence rate is in a flat range, the learning rate attenuation factor is multiplied by the baseline learning rate as a new parameter; If the residual norm shows periodic fluctuations, the random restart strategy is enabled to reset the optimization direction and step size range.

7. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 1, characterized in that: The event-driven state machine includes nested conditional jump logic; The method of using a sliding time window statistical method to detect the abnormal instantaneous current peak value includes: Extract the mean, range and kurtosis characteristics of current monitoring data within a fixed time window; The current peak probability distribution function is fitted by the generalized Pareto distribution model; When the observed current value exceeds the 99.9% confidence interval upper limit of the probability distribution function, a local battery overload alarm instruction is triggered.

8. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 7 is characterized in that: The multi-scale frequency domain analysis algorithm is implemented using wavelet transform decomposition technology; The identifying of the power grid harmonic distortion area by a multi-scale frequency domain analysis algorithm includes: The power grid frequency fluctuation signal is mapped into a multi-resolution wavelet coefficient matrix, and the high-frequency abnormal components in the matrix are extracted; Based on the cluster analysis algorithm, clusters of high-amplitude coefficients with temporal and spatial correlations are merged to generate markers for potential harmonic interference areas. Calculate the energy proportion and frequency offset product index of each marked area, and select the target interval that meets the preset frequency modulation level.

9. The energy management method for mobile energy storage vehicles based on intelligent algorithms according to claim 1, characterized in that: The multi-source heterogeneous data fusion and outlier elimination processing are implemented by adopting a robust data cleaning strategy; the multi-source heterogeneous data fusion and outlier elimination processing of the real-time energy interaction data stream includes: Establish a data mapping relationship between the current timing sequence of the edge computing node, the load frequency sequence of the cloud platform, and the temperature sampling sequence of the vehicle terminal; The high-frequency noise components in the data stream of each node are separated by the empirical mode decomposition algorithm, and the low-frequency effective data are cross-modally aligned based on the robust principal component analysis algorithm; A time series interpolation algorithm is used to compensate for the sampling interval deviation among multi-source data streams and generate a spatiotemporally consistent fusion energy management tensor.

Images