Building energy consumption forecasting procedure using ambient temperature, enthalpy, bias corrected weather forecast and outlier corrected sensor data

Patent metadata

Field	Value
Publication number	US-9568519-B2
Application number	US-201414278603-A
Country	US
Kind code	B2
Filing date	May 15, 2014
Priority date	May 15, 2014
Publication date	Feb 14, 2017
Grant date	Feb 14, 2017

Abstract

A procedure for forecasting building energy consumption by evaluating performance of variable base degree and variable based enthalpy models. Dynamic weights are computed for the variable base degree and variable based enthalpy models and used in making future energy prediction based on weather forecast data. The weather forecast data may be corrected for bias. The variable base degree and variable based enthalpy models may be calibrated based on outlier removed historic energy consumption data and historic ambient air temperature data.

First claim

We claim:

1. A method of predicting energy consumption in a building, comprising:
receiving historic ambient air data;
receiving historic energy consumption data associated with a building;
calibrating, by one or more hardware processors, a variable base degree model based on the historic ambient air data and the historic energy consumption data;
calibrating, by said one or more hardware processors, a variable based enthalpy model based on the historic ambient air data and the historic energy consumption data;
receiving weather forecast data;
running, by said one or more hardware processors, the variable base degree model with the weather forecast data to produce a first energy consumption prediction;
running, by said one or more hardware processors, the variable based enthalpy model with the weather forecast data to produce a second energy consumption prediction;
computing, by said one or more hardware processors, a first weight associated with the variable base degree model dynamically based on performance of the variable base degree model and performance of the variable based enthalpy model during a predefined time period;
computing, by said one or more hardware processors, a second weight associated with the variable based enthalpy model dynamically based on performance of the variable based enthalpy model and the variable base degree model during the predefined time period; and
combining, by said one or more hardware processors, the first energy consumption prediction and the second energy consumption prediction as a function of the first weight and the second weight.

2. The method of claim 1 , further comprising:
removing outlier data associated with ambient air data from the historic ambient air data; and
removing outlier data associated with energy consumption data from the historic energy consumption data.

3. The method of claim 1 , further comprising:
correcting bias in the weather forecast data.

4. The method of claim 3 wherein the correcting bias in the weather forecast data comprises:
computing a regression model that formulates a relationship between previously forecasted weather data of a period of time and actual observed weather data over the same period of time; and
correcting the bias in the weather forecast data based on the relationship.

5. The method of claim 4 , further comprising accounting for time dependent bias.

6. The method of claim 1 , wherein the first weight and the second weight are computed as:

w
k

=

1

σ
^

k
2

Σ
k

⁢

1

σ
^

k
2

with

σ
^

k
2

=

1

n
t

⁢

∑

t
=
1

n
t

⁢

⁢

(

Y
t

-

Y
^

k
,
t

)

2

where
{circumflex over (σ)} k represents sigma value associated with model k;
n t represents the total number of time periods being considered;
Y t represents actual consumption data at time period t;
Ŷ k,t represents predicted energy consumption by model k at time period t.

7. The method of claim 6 , wherein the combining comprises:
Ŷ new =Σ k w k Ŷ k

where
Ŷ new represents a dynamically weighted combined result of the first energy consumption prediction and the second energy consumption prediction;
k represents model k;
Ŷ k represents energy consumption predicted by k-th model; and
w k represents weight associated with k-th model.

8. A computer readable storage medium storing a program of instructions executable by a machine to perform a method of predicting energy consumption in a building, the method comprising:
receiving historic ambient air data;
receiving historic energy consumption data associated with a building;
calibrating a variable base degree model based on the historic ambient air data and the historic energy consumption data;
calibrating a variable based enthalpy model based on the historic ambient air data and the historic energy consumption data;
receiving weather forecast data;
correcting bias in the weather forecast data;
running the variable base degree model with the weather forecast data to produce a first energy consumption prediction;
running the variable based enthalpy model with the weather forecast data to produce a second energy consumption prediction;
computing a first weight associated with the variable base degree model dynamically based on performance of the variable base degree model and performance of the variable based enthalpy model during a predefined time period;
computing a second weight associated with the variable based enthalpy model dynamically based on the performance of the variable based enthalpy model and the performance of the variable base degree model during the predefined time period; and
combining the first energy consumption prediction and the second energy consumption prediction as a function of the first weight and the second weight.

9. The computer readable storage medium of claim 8 , further comprising:
removing outlier data associated with ambient air data from the historic ambient air data; and
removing outlier data associated with energy consumption data from the historic energy consumption data.

10. The computer readable storage medium of claim 8 , wherein the correcting bias in the weather forecast data comprises:
computing a regression model that formulates a relationship between previously forecasted weather data of a period of time and actual observed weather data over the same period of time; and

Assignees

• IBM

Inventors

• Hwang Youngdeok
• Lee Young Min
• Zhu Yada

Classifications

• G06Q10/04 Primary

  Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" (market predictions or forecasting for commercial activities G06Q30/0202) · CPC title

• G01N25/20

  by investigating the development of heat, i.e. calorimetry, e.g. by measuring specific heat, by measuring thermal conductivity (calorimeters per se G01K) · CPC title

• G01R21/133 Primary

  by using digital technique · CPC title

• G06Q50/06

  Energy or water supply · CPC title