BEV BMS Battery Consistency Bias Alarm Prediction

What does this Amazon Web Services Solution do?

The solution provides a tool chain for alert prediction including data store tool, data analysis tool, model process tool and data visualization tool, allows Original Equipment Manufacturer (OEM) speed up BMS Alert prediction machine learning developing process, and makes it easier to securely manage terabytes of BEV BMS and connected vehicle data for store and pipeline.

Amazon Web Services Solution overview

This solution is used for Battery Electric Vehicle(BEV) Battery Management System(BMS) alert prediction in automotive industry. The solution offers smooth workflow in Amazon Web Services Cloud that makes it easier to securely manage terabytes of data for battery and connected vehicle data to store and scale parallel algorithm training workloads to hundreds of cores in short time. Leveraging the Amazon SageMaker, the solution offers data engineer and scientist one-stop platform for data pre-processing, feature engineering, algorithm selection/training/fine-tuning and model deployment.

Architecture Description

With Amazon S3, Amazon Lambda, Amazon SageMaker, Amazon DynamoDB and open-source Apache Superset, this solution offers machine learning developers and data scientists an entire tool chain for BMS Alert Prediction, including data store tool, data analysis tool, model process tool and data visualization tool.

Workflow in details:

1. OEM or BEV battery vendors upload battery data (captured with IoT, etc.) to S3 bucket.

2. SageMaker notebook instance obtain dataset from S3 bucket, which could be used for model training.

3. After model building, model training and model deployment, SageMaker outputs a runtime endpoint, which can be used to provide inference service.

Scenario 1:

    4. Connected vehicles upload battery data (batch data) to S3 bucket.

    5. S3 bucket create event invoke lambda function.

    6. Lambda function invokes the deployed SageMaker endpoint, perform inference on the uploaded batch data.

    7. Batch inference results are written into DymanoDB.

    8. Batch inference results are written into S3 for superset visualization.

Scenario 2:

    9. User invokes the prediction service via HTTP POST request.

    10. API Gateway routes the request to lambda function.

    11. Lambda function invokes the deployed SageMaker endpoint to perform inference with the post data as input.

    12. API single inference results are written into Dymanodb.

    13. API single inference results are written into S3 for superset visualization.

14. Superset is hosted in Fargate. Users can access inference results using SQL query.

15. Athena queries inference events from Glue Data Catalog.

16. The table defined in Glue Data Catalog describes the schema of inference events which are stored in S3 bucket.

BEV BMS Battery Consistency Bias Alarm Prediction

Version 1.0.0
Last updated: 12/2020
Author: Amazon Web Services 

Estimated deployment time: 10 min

Source code 

Features

User interaction

The solution decouples solution deployment and the machine learning deployment. You can run Sagemaker notebook script to get familiar with each phase of machine learning pipeline, including data loading, data visualization, feature engineering, data split, model training, model deployment and test, which facilitates the customize process.

End-to-End user experience

Leveraging Amazon API Gateway, Amazon Lambda and Amazon SageMaker runtime endpoint, this solution offers end-to-end deployment for battery consistency bias alarm prediction user scenarios. Two scenarios are supported: 1) users can upload batch data to S3 to trigger automatic batch inference. 2) users can send HTTP requests to inference via Restful API call.

Data visualization

The solution integrates open-source Apache Superset to support SQL query over inference events. You can filter, extract and visualize your interested inference result using SQL commands intuitively.
Explore all Amazon Web Services Solutions

Browse our portfolio of Amazon Web Services -built solutions to common architectural problems.

Learn more 
Find a Partner

Find Amazon Web Services certified consulting and technology partners to help you get started.

Learn more 
Start building in the console

Sign-up and start exploring our services.

Get started 
Close
Hot Contact Us

Hotline Contact Us

1010 0766
Beijing Region
Operated By Sinnet
1010 0966
Ningxia Region
Operated By NWCD