Cortana Intelligence Solution Template Playbook for demand forecasting of energy. Executive Summary. In the past few years, Internet of Things (Io. T), alternative energy sources, and big data have merged to create vast opportunities in the utility and energy domain. At the same time, the utility and the entire energy sector have seen consumption flattening out with consumers demanding better ways to control their use of energy. Hence, the utility and smart grid companies are in great need to innovate and renew themselves. Furthermore, many power and utility grids are becoming outdated and very costly to maintain and manage. During the last year, the team has been working on a number of engagements within the energy domain. During these engagements, we have encountered many cases in which the utilities or ISVs (Independent Software Vendors) have been looking into forecasting for future energy demand. These forecasts play an important role in their current and future business and have become the foundation for various use cases. These include short and long- term power load forecast, trading, load balancing, grid optimization etc. Big data and Advanced Analytics (AA) methods such as Machine Learning (ML) are the key enablers for producing accurate and reliable forecasts. In this playbook, we put together the business and analytical guidelines needed for a successful development and deployment of energy demand forecast solution. These proposed guidelines can help utilities, data scientists, and data engineers in establishing fully operationalized, cloud- based, demand- forecasting solutions. For companies who are just starting their big data and advanced analytics journey, such a solution can represent the initial seed in their long- term smart grid strategy. Overview. This document covers the business, data, and technical aspects of using Cortana Intelligence and in particular Azure Machine Learning (AML) for the implementation and deployment of Energy Forecasting Solutions. Free Sales Forecasting Template (Excel)If you do not have a system in place to manage your sales forecasting then please feel free to download this free sales.Microsoft PowerPoint template to display up-to-date weather information of your location. Click the Free Access button below! MS Access Templates give you a fast start to your Access project. You can use the Access database templates to build your own solution, or simply as examples of how. A unique collection of Microsoft. It\'s includes access database Templates, MS access Templates, ms access database. The document consists of three main parts: Business understanding. Data understanding. Technical implementation. The Business Understanding part outlines the business aspect one needs to understand and consider prior to making an investment decision. It explains how to qualify the business problem at hand to ensure that predictive analytics and machine learning are indeed effective and applicable. The document further explains the basics of machine learning and how it is used to address energy- forecasting problems. It outlines the prerequisites and the qualification criteria of a use case. Some sample use cases and business case scenarios are also provided. Data is the main ingredient for any machine learning solution. The Data Understanding part of this document covers some important aspects of the data. It outlines the kind of data that is needed for energy forecasting, data quality requirements, and what data sources typically exist.
If you do not have a mechanism in place to forecast your sales then please feel free to download this Free Sales Forecasting Template (Excel Spreadsheet). Free ms access forecasting template; Access Forecasting Templates. Workload forecasting for scheduling optimization, suited for call centers. Online templates and themes for Office. Find resumes, calendars, and budgets for Excel, Word and PowerPoint. The Excel forecasting templates are free to download. Find the perfect Excel sales forecast template and get your calculations right. We also explain how the raw data is used to prepare data features that actually drive the modeling part. The third part of the document covers the Technical Implementation aspect of a solution. Feature engineering and modeling are at the core of the data science process and are therefore being discussed in some detail. It covers the concept of web services, which are an important vehicle for cloud deployment of predictive analytics solutions. We also outline a typical architecture of an end- to- end operationalized solution. In addition, the document includes reference material that you can use to gain further understanding of the domain and technology. It is important to note that we do not intend to cover in this document the deeper data science process, its mathematical and technical aspects. These details can be found in Azure ML documentation and blogs. Target Audience. The target audience for this document is both business and technical personnel who would like to gain knowledge and understanding of Machine Learning based solutions and how these are being used specifically within the energy- forecasting domain. Data scientists can also benefit from reading this document to gain a better understanding of the high level process that drives the deployment of an energy forecasting solution. In this context it can also be used to establish a good baseline and starting point for more detailed and advanced material. Industry Trends. In the past few years, Io. T, alternative energy sources, and big data have merged to create vast opportunities in the utility and energy space. At the same time, the utility and the entire energy sectors have seen consumption flattening out with consumers demanding better ways to control their use of energy. Many utility and smart energy companies have been pioneering the smart grid by deploying a number of use cases that make use of the data generated by the grid. Many use cases revolve around the inherent characteristics of electricity production: it cannot be accumulated nor stored aside as inventory. So, what is produced must be used. Utilities that want to become more efficient need to forecast power consumption simply because that will give them greater ability to balance supply and demand, thus preventing energy wastage, reduce greenhouse gas emission, and control cost. When talking of costs, there is another important aspect, which is price. New abilities to trade power between utilities have brought in a great need to forecast future demand and future price of electricity. This can help companies determine their production volumes. When we use the word 'smart', we actually refer to a grid that can learn and then make predictions. It can anticipate seasonal changes in consumption as well as foresee temporary overload situations and automatically adjust for it. By remotely regulating consumption (with the help of these smart meters), localized overload situations can be handled. By first predicting and then acting, the grid makes itself smarter over time. For the rest of this document we will focus on a specific family of use cases that cover forecasting of future, short term, and long- term energy demand. We have been working in these areas for a few months and have gained some knowledge and skill that allow us to produce industry grade results. Other use cases will be covered as well in the document in the near future. Business Understanding. Business Goals. The Energy Demo goal is to demonstrate a typical predictive analytics and machine learning solution that can be deployed in a very short time frame. Specifically, our current focus is on enabling energy demand forecast solutions so that its business value can be quickly realized and leveraged upon. The information in this playbook can help the customer accomplishing the following goals: - Short time to value of machine learning based solution - Ability to expand a pilot use case to other use cases or to a broader scope based on their business need - Quickly gain Cortana Intelligence Suite product knowledge. With these goals in mind, this playbook aims at delivering the business and technical knowledge that will assist in achieving these goals. Power Load and Demand Forecasting. Within the energy sector, there could be many ways in which demand forecasting can help solve critical business problems. In fact, demand forecasting can be considered the foundation for many core use cases in the industry. In general, we consider two types of energy demand forecasts: short term and long term. Each one may serve a different purpose and utilize a different approach. The main difference between the two is the forecasting horizon, meaning the range of time into the future for which we would forecast. Short Term Load Forecasting. Within the context of energy demand, Short Term Load Forecasting (STLF) is defined as the aggregated load that is forecasted in the near future on various parts of the grid (or the grid as a whole). In this context, short term is defined to be time horizon within the range of 1 hour to 2. In some cases, a horizon of 4. Therefore, STLF is very common in an operational use case of the grid. Here are some examples of STLF driven use cases: - Supply and demand balancing - Power trading support - Market making (setting power price) - Grid operational optimization - Demand response - Peak demand forecasting - Demand side management - Load balancing and overload prevention - Long Term Load Forecasting - Fault and anomaly detection - Peak curtailment/leveling. Obtaining accurate temperature forecast for the next hour and up to 2. These models are less sensitive to seasonal patterns or long- term consumption trends. SLTF solutions are also likely to generate high volume of prediction calls (service requests) since they are being invoked on an hourly basis and in some cases even with higher frequency. It is also very common to see implantation where each individual substation or transformer is represented as a standalone model and therefore the volume of prediction requests are even greater. Long Term Load Forecasting. The goal of Long Term Load Forecasting (LTLF) is to forecast power demand with a time horizon ranging from 1 week to multiple months (and in some cases for a number of years). This range of horizon is mostly applicable for planning and investment use cases. For long- term scenarios, it is important to have high quality data that covers a span of multiple years (minimum 3 years). These models will typically extract seasonality patterns from the historical data and make use of external predicators such as weather and climate patterns. It is important to clarify that the longer the forecasting horizon is, the less accurate the forecast may be. It is therefore important to produce some confidence intervals along with the actual forecast that would allow humans to factor the possible variation into their planning process. Since the consumption scenario for LTLF is mostly planning, we can expect much lower prediction volumes (as compared to STLF). Free Templates for Office Online.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. Archives
January 2017
Categories |