The Cortana Intelligence Suite has added a new on-premises template for forecasting energy demands to help utilities and energy companies allocate resources and make strategic decisions.
The Energy Demand Forecasting Template With SQL Server R Services allows a simple deployment of a machine learning-based solution for energy demand forecasting within on-premises infrastructure. It takes advantage and extends the capability of SQL Server already in deployment.
Summary
Demand forecasting is an important problem in various domains including energy, retail, services, etc. Accurate demand forecasting helps companies conduct better production planning, resource allocation, and make other important business decisions. In the energy sector, demand forecasting is critical for reducing energy storage cost and balancing supply and demand. This template demonstrates how to use SQL Server R Services to build an end-to-end, on-prem solution for electricity demand forecasting.
Description
This template demonstrates how to use SQL Server R Services to build an end-to-end, on-prem solution for electricity demand forecasting. For a cloud-based solution using Cortana Intelligence Suite (CIS), please see CIS Solution Template: Demand Forecasting for Energy.The solution template includes a real time data simulator, feature engineering, model retraining, forecasting, and visualization.
In this template with SQL Server R Services, we show two versions of implementation:
In this template with SQL Server R Services, we show two versions of implementation:
- Model Development with Microsoft R Server in R IDE. Run the code in R IDE (e.g., RStudio, R Tools for Visual Studio) with data in SQL Server, and execute the computation in SQL Server.
- Model Operationalization In SQL. Deploy the modeling steps to SQL Stored Procedures, which can be run within SQL environment (such as SQL Server Management Studio) or called by applications to make predictions. A powershell script is provided to deploy the template automatically.
Below is the directory structure for this template:
- SQLR: SQL stored procedures for data simulation, data preprocessing, feature engineering, model training and scoring. The stored procedures are run on a SQL server. A PowerShell script is provided to automatically deploy the stored procedures on to a SQL server.
- Data: Sample demand and temperature data for model training and testing.
- R: R development code (Microsoft R Server). It runs in R IDE, with computation being done in-database (by setting compute context to SQL Server).