Data Mining and Machine Learning in Building Energy Analysis

2Data Acquisition for Building Energy Analysis

2.1. Introduction

As described in Chapter 1, the energy system in buildings is complex and contains many uncertainties. Therefore, accurate consumption data, especially the time series for each variable, are difficult to obtain. How to collect sufficient clean data is always an important concern in statistical analysis and model development. There are three common approaches to solve this problem. The first approach is to survey or do questionnaires from customers and collect data from building plan or energy providers. Data collected in these ways are always at the rough level, i.e. monthly or yearly consumption, or at the building level or even section level. The second approach is to do measurements in real buildings. This approach can provide energy performance in details, such as the hourly profile of a sensor. The third approach is to simulate these data numerically with some devoted software. Depending on the quality of these pieces of simulation software, this can provide us with clean consumption datasets the same as real consumption measurements.

This chapter will discuss how to collect real consumption data and how to perform simulation, providing the basic datasets that will be used in later chapters for developing models.

2.2. Surveys or questionnaires

Surveys or questionnaires are usually launched by utility companies, building management companies, energy analysis companies or government organizations. For instance, ...