1 Main points of modular smart battery research and development

In key fields such as new energy vehicles and energy storage power stations, modular smart batteries (SB) have become the core component for improving energy control efficiency due to their reconfigurability and redundant fault-tolerant characteristics. Compared with traditional series connected battery packs, modular intelligent batteries are composed of multiple independent batteries, each equipped with a dedicated half bridge circuit, which can be connected or bypassed in real-time according to the operating status. When the state of charge (SOC) of a battery is too low, the system can automatically switch to redundant batteries, ensuring power supply stability and extending overall service life. This characteristic makes it irreplaceable in high-end scenarios such as aerospace and uninterruptible power supply.

However, the development of controller algorithms for modular intelligent batteries has long been constrained by bottlenecks in the testing process. In the traditional research and development model, algorithm validation needs to be directly carried out on real battery packs. Once the algorithm has logical defects, it is highly likely to cause safety accidents such as battery overheating, short circuits, and even explosions. This not only causes expensive hardware losses, but also prolongs the research and development cycle by more than 30%. Although existing research has proposed simple battery simulator solutions, such tools can only simulate basic electrical characteristics and cannot integrate core control logic such as SOC balance and temperature field coupling, resulting in significant deviations between test results and actual application scenarios. The algorithm still needs to be repeatedly debugged when implemented.

The research team has proposed a virtual prototype testing platform to address the pain points in this industry. This platform links real controllers with virtual battery simulation systems, retaining the accuracy of real control scenarios while completely avoiding the safety risks of physical testing. This article will provide a deep analysis of the core technology, implementation path, and application value of the solution, providing a practical reference framework for industry research and development.

2 Technology Analysis: The Three Main Technical Pillars of Virtual Testing Platform

The core value of the virtual testing platform lies in “precise simulation+real-time interaction”. The research team has achieved full dimensional simulation of the real battery operating status by constructing battery electrical models, thermal coupling models, and intelligent control algorithms. These three technological pillars are both independent and deeply interconnected, collectively forming the technological barriers of the platform.

(1) Electrical Model: Balancing Precision and Real time Design

The precise simulation of battery electrical characteristics is the foundation of the platform. The research team abandoned complex but computationally intensive high-order models and chose the first-order Randall equivalent circuit model as the core. The model is based on the principle of “precision adapted real-time operation” and simulates the dynamic response of batteries through three key components: first, open circuit voltage (OCV), which has a non-linear correspondence with SOC, and establishes a dedicated database through experimental data fitting; The second is the Ohmic resistance (R ₀), which reflects the inherent resistance at the interface between the electrolyte and the electrode; The third is the RC parallel branch, which is used to simulate the voltage hysteresis effect during charge transfer and diffusion processes.

To meet the real-time computing requirements of digital signal processors (DSPs), the research team discretized the model, converting the differential equations in the continuous time domain into recursive formulas in the discrete time domain.

The advantage of this design is that the single cycle computation time is controlled within 10 milliseconds, fully meeting the requirements of real-time simulation. At the same time, the voltage simulation error is controlled within 2%, and the accuracy meets industrial grade application standards.

(2) Thermal coupling model

In real operating scenarios, there is a strong coupling relationship between the electrical characteristics of batteries and temperature: the Joule heating generated by the current passing through the resistor will cause the temperature to rise, and the temperature change will in turn affect the internal resistance and OCV, forming a closed-loop feedback of “electrical temperature electrical”. If this coupling relationship is ignored, the simulation results will show serious deviations, so the construction of a thermal model is crucial.

The research team adopts a lumped parameter thermal model, with the average temperature of the battery as the core monitoring indicator, and achieves real-time calculation of temperature changes through energy balance equations.

Through the thermal coupling model, the platform successfully simulated the real scenario of “sudden temperature rise → internal resistance increase → output voltage drop during high current discharge”, solving the core defect of traditional simulators that “only charge without heating”. The test data shows that the temperature simulation error is less than 1.5 ℃, providing accurate data support for subsequent temperature balance control.

(3) Control Algorithm: Dual Objective Optimization Strategy Based on K-Nearest Neighbors

The core goal of the controller algorithm is to achieve SOC and temperature balance of each battery while meeting the load voltage requirements, avoiding local overcharging, overdischarging or overheating. The research team adopted an improved K-nearest neighbor (KNN) algorithm and implemented intelligent control through the logic of “distance sorting dynamic group selection”. Compared with traditional PID control, the balance accuracy was improved by 40%.

The core execution process of the algorithm is divided into four steps: the first step is state sampling, which collects real-time SOC, temperature, and terminal voltage data of all batteries to determine the system target reference value (SOC_max, Tmin); The second step is distance calculation, constructing a weighted distance formula; The third step is to select battery groups, sort them in ascending order of distance, select the top K batteries to connect to the load, and ensure that the total output voltage matches the load requirements; The fourth step is signal output, generating switch control signals and sending them to each battery.

The flexibility of this algorithm lies in its ability to adapt to different application scenarios by adjusting the weight coefficients. For example, in the scenario of new energy vehicles, increasing the weight of k ₂ can prioritize ensuring temperature balance; In the scenario of energy storage power stations, increasing the weight priority to achieve SOC balance greatly improves the platform’s versatility.

3 Platform Architecture and Implementation

To achieve the integration of “virtual simulation+real control”, the research team designed a hardware system with a “master-slave” architecture, which combines virtual battery simulation with real controller operations to form a complete hardware in the loop testing loop. This architecture retains the computational logic of the real controller while avoiding safety risks through virtual batteries, making it a key design for the platform’s implementation.

(1) Hardware composition: Layered design ensures stability

The platform hardware is divided into three levels: the bottom layer is the DSP node layer, where each node simulates a battery and runs real-time electrical and thermal models with a sampling frequency of 100Hz; the middle layer is the main control layer, which consists of the same model DSP and Google Coral development board. The DSP is responsible for communication scheduling with the slave nodes, while Google Coral runs K-nearest neighbor control algorithm to achieve fast data processing; The top layer is the monitoring layer, which displays real-time status parameters such as SOC, temperature, and voltage of each battery through the PC upper computer, supporting data storage and curve analysis.

In terms of communication, the master-slave nodes adopt the TSCH (Time Slot Channel Hopping) protocol, which combines time slicing and frequency hopping strategies to ensure real-time communication (with a delay of less than 5 milliseconds) and enhance anti-interference capabilities, adapting to complex electromagnetic scenarios in industrial environments.

(2) Simulation process: Three step closed-loop implementation for real-time interaction

The operation process of the entire platform forms a closed loop, which is divided into three core steps: data collection, model calculation, and control execution. The specific steps are as follows:

Data interaction stage: The main controller sends initial control signals (switch status) to each slave DSP node. After receiving them, the slave nodes feed back their simulated SOC and temperature data to the main controller, completing the first round of data interaction;

Model operation stage: Based on the switch signal from the main controller, the nodes call the electrical and thermal models to calculate the SOC, voltage, and temperature data for the next cycle. At the same time, the main controller inputs the received state data into the K-nearest neighbor algorithm to generate new switch control signals;

Control execution phase: The main controller sends a new switch signal to the slave node, which adjusts the simulation state and feeds back data again, entering the next cycle.

The test results show that the entire closed-loop cycle is stable at 10 milliseconds, fully meeting the requirements of real-time simulation and accurately reproducing the dynamic response process of real battery packs.

(3) Performance verification: a dual breakthrough in security and efficiency

The research team validated the performance of the platform through a 12 minute continuous simulation test, with testing conditions set as follows: initial SOC distribution between 0.8-1.0pu (standard value), ambient temperature of 25 ℃, and load current of 100A. The test results presented three core advantages: firstly, fast balancing speed, achieving SOC deviation of less than 5% for all batteries within 6 seconds, and temperature deviation controlled within 2 ℃; Secondly, the safety risk is zero, and there are no abnormal states such as overcharging, overdischarging, overheating, etc. throughout the entire process; Thirdly, the research and development efficiency is high. Compared with real battery testing, the single algorithm verification time has been shortened from 2 hours to 12 minutes, and the efficiency has been improved by 90%.

These data fully demonstrate that the platform can completely replace traditional real battery testing for the initial development and validation of controller algorithms, significantly reducing research and development costs and cycles.

4 Industry application value

This virtual testing platform is not simply a laboratory technology, but a practical tool with clear industrial implementation scenarios. Based on the research and development needs of the new energy industry, its application value is mainly reflected in three core scenarios, while providing differentiated use value for different user groups.

For the BMS (Battery Management System) R&D team of new energy vehicle companies, the platform can be used for the initial validation of new vehicle controller algorithms. In the traditional mode, R&D personnel need to wait for the production of real battery samples before conducting testing. However, through this platform, algorithm debugging can be carried out during the battery design stage, achieving “parallel research and development of algorithms and hardware”, and the overall R&D cycle can be shortened by 40%. For example, in a pilot application of a certain car company, the platform was used for SOC balance algorithm testing of lithium iron phosphate battery packs, and the traditional mode of one month’s debugging workload was completed in just two weeks.

For energy storage system integrators, the platform can solve the testing problem of retired battery cascading utilization. There is a significant difference between the SOC and State of Health (SOH) of retired batteries, and traditional testing requires individual testing and grouping, which is costly. This platform can quickly simulate the combined operation status of different retired batteries, accurately screen the battery combination scheme with the highest matching degree, and reduce the testing cost of cascading utilization by more than 30%.

For universities and research institutions, the platform can serve as a teaching and research tool in the field of battery control. On the one hand, students can conduct secure algorithm debugging experiments through the platform to avoid the safety risks of real battery testing; On the other hand, researchers can quickly validate new control algorithms based on the platform, such as integrating reinforcement learning, fuzzy control and other algorithms into the platform, without the need to build complex hardware systems to complete preliminary validation.

In addition, the research team has reserved extension interfaces for the platform, which can be adapted to different types of batteries (such as ternary lithium batteries and sodium ion batteries) by modifying model parameters, further enhancing the platform’s versatility and lifecycle value.

5 Summary and Prospect

The modular intelligent battery virtual testing platform analyzed in this article successfully solves the three major pain points of “high safety risks, long development cycles, and high testing costs” in traditional research and development models through the deep integration of electrical thermal coupling models and intelligent control algorithms. The core innovation lies in achieving a balance between “precise simulation” and “real-time interaction”, ensuring the reliability of test results and adapting to the efficiency requirements of industrial level research and development.

From the perspective of industry development, with the increasingly widespread application of modular intelligent batteries in the field of new energy, such virtual testing tools will become standard in the research and development process. In the future, this technology can be upgraded in two directions: one is to introduce digital twin technology to build a virtual mapping of the entire lifecycle of batteries, realizing full chain simulation from research and development to operation and maintenance; The second is to combine artificial intelligence algorithms and train fault prediction models through massive simulation data to identify potential battery failure risks in advance.

For R&D personnel, they can focus on drawing on the design ideas of “model simplification and adaptation calculation” and “multi physics field coupling” of the platform, adjust model parameters and algorithm logic according to their own needs, and achieve localized implementation of the technology.