# US 2014/0156322 A1 · Unified Framework for Electrical Load Forecasting

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2014/0156322 A1
- Document ID: US20140156322A1
- Status: Published application
- Kind: A1
- Application number: 13/931,583
- Filed: 2013-06-28
- Published: 2014-06-05
- Earliest claimed: 2012-08-10
- Assignees: Itron, Inc.
- Applicants: Itron, Inc.
- Inventors: Frank Anthony Monforte; Christine Ann Fordham
- Web page: https://patentgrep.com/patent/US20140156322A1

## CPC

- G06Q 50/06: Energy or water supply
- H02J 3/004: {Generation forecast, e.g. methods or systems for forecasting future energy generation}
- H02J 3/003: {Load forecast, e.g. methods or systems for forecasting future load demand}
- Y04S 10/50: Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- H02J: ELECTRIC POWER NETWORKS; CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- Y04S: SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS

## Abstract

Techniques for forming and using a unified framework for electrical load forecasting are described herein. In one example, data is received from supervisory control and data acquisition (SCADA) measurements available in an electrical grid. Data may be validated using ramp rate filter and load level filter tests. An augmented Savistky-Golay filter may be used to smooth the data, which may then be passed to autoregressive near-term forecast model(s). Forecasts are generated using near-term, ramp rate and short-term models. The near-term and ramp rate forecasts may be combined, such as with forecast horizon and ramp rate horizon parameters, to create a stitched forecast. The stitched forecast and short-term forecast are then blended using a blending period and blending weights that may be supplied exogenously. The final blended or unified near-term and short-term forecast may be exported to downstream generation scheduling and dispatching applications, which may calculate electrical production.

## Claims

### Claim 1

One or more computer-readable media storing computer-executable instructions which, when executed, cause one or more processors to perform acts comprising:
importing latest available load data from a supervisory control and data acquisition (SCADA) system associated with an electrical grid;
validating the data using a ramp rate filter test and a load level filter test;
smoothing the data using an augmented Savitzky-Golay filter using lift factors appropriate to a time of day;
generating a near-term forecast, a ramp rate forecast and a short-term forecast using exogenously supplied data;
passing the smoothed data to the near-term forecast;
combining the near-term forecast and the ramp rate forecast to create a stitched forecast; and
blending the stitched forecast and the short-term forecast to produce a unified forecast;
wherein the unified forecast is configured to regulate a spinning reserve of the electrical grid.

### Claim 2 (depends on 1)

The one or more computer-readable media of claim 1, additionally comprising:
updating the lift factors using additional load data; and
re-estimating the near-term forecast, the ramp rate forecast and the short-term forecast.

### Claim 3 (depends on 1)

The one or more computer-readable media of claim 1, wherein smoothing the data comprises:
smoothing the data using the augmented Savitzky-Golay filter using a moving-average window having a size that is a function of expected load changes.

### Claim 4 (depends on 1)

The one or more computer-readable media of claim 1, wherein smoothing the data comprises:
calculating the lift factors as a ratio of smoothed load from a centered Savitzky-Golay moving average to a right hand side constrained Savitzky-Golay moving average.

### Claim 5 (depends on 1)

The one or more computer-readable media of claim 1, wherein the combining comprises:
stitching a plurality of ramp rate forecasts to an end of the near-term forecast.

### Claim 6 (depends on 1)

The one or more computer-readable media of claim 1, wherein the combining comprises:
using forecast horizon parameters and ramp rate horizon parameters to create the stitched forecast.

### Claim 7 (depends on 1)

The one or more computer-readable media of claim 1, wherein the combining comprises:
stitching at least one ramp rate forecast to a horizon of the near-term forecast, wherein the stitching bridges a gap between the near-term forecast and short-term forecast, and wherein the near-term forecast is autoregressive and the short-term forecast is non-autoregressive.

### Claim 8 (depends on 1)

The one or more computer-readable media of claim 1, wherein the blending comprises:
reducing blending weights to zero over a blending period, wherein the blending weights are used in an equation that includes terms associated with the ramp rate forecast and short-term forecast.

### Claim 9 (depends on 1)

The one or more computer-readable media of claim 1, wherein the unified forecast comprises:
an autoregressive portion that utilizes recent load information and a non-autoregressive portion that utilizes calendar and weather information.

### Claim 10

A method of producing a unified near-term and short-term system load forecast, comprising:
under control of one or more processors configured with executable instructions:
receiving data of supervisory control and data acquisition (SCADA) measurements of an electrical grid;
generating near-term and short-term forecasts based on the received data;
stitching a ramp rate forecast to the near-term forecast to create a stitched forecast;
blending the stitched forecast to the short-term forecast to form the unified forecast; and
regulating a spinning reserve of the electrical grid using the unified forecast.

### Claim 11 (depends on 10)

The method as recited in claim 10, wherein the stitching is performed using relationships, comprising:
LL
_
t
+
h
RR
=
t
+
h
NT
,
∀
h
≤
H
LL
_
t
+
H
+
j
RR
=
t
+
H
NT
+
∑
j
=
1
J

t
+
H
+
j
RR
wherein, LLt+h+jRR is equal to a near-term forecast for forecast horizons less than or equal to a near-term forecast horizon indexed by (H), and a forecasted load resulting stitching the ramp rate forecast to an end of a near-term forecast horizon (t+H).

### Claim 12 (depends on 10)

The method as recited in claim 10, wherein the blending is performed using relationships, comprising:
LL_Blended
t
+
h
+
j
+
k
=
LL
_
t
+
h
+
j
+
k
RR
×
Weight
t
+
h
+
j
+
k
+
(
(
1
-
Weight
t
+
h
+
j
+
k
)
×
t
+
h
+
j
+
k
ST
)
where:
Weightt+h+j+k=1.0, when (t+h+j+k)≦H+J
0≦Weightt+h+j+k≦1,
when H+J<(t+h+j+k)≦H+J+K
Weightt+h+j+k=0, when (t+h+j+k)>H+J+K
wherein Weightt+h+j+k is weight placed on the stitched forecast computed h+j+k intervals into a forecast horizon.

### Claim 13 (depends on 10)

The method as recited in claim 10, wherein:
the near-term forecast has a horizon that is between 10-minutes and 60-minutes ahead, and is based on load stability;
the stitched forecast has a horizon that is between 30 minutes to 60 minutes ahead; and
the unified forecast has a horizon that is at least two hours ahead.

### Claim 14 (depends on 10)

The method as recited in claim 10, wherein the ramp rate forecast is configured using sigma values ({circumflex over (σ)}tR) that are obtained from a generalized autoregressive conditional heteroskedasticity (GARCH) forecast.

### Claim 15 (depends on 10)

The method as recited in claim 10, wherein blending the ramp rate forecast to the short-term forecast to form a unified forecast comprises:
weighting a stitched forecast at intervals into a forecast horizon, wherein the weight is reduced to 0.0 over plurality of intervals.

### Claim 16 (depends on 10)

The method as recited in claim 10, wherein blending the ramp rate forecast to the short-term forecast to form the unified forecast comprises:
bridging a gap between the near-term and short-term forecasts with a plurality of five-minute ramp rate forecasts.

### Claim 17 (depends on 10)

The method as recited in claim 10, wherein blending the ramp rate forecast to the short-term forecast to form the unified forecast comprises:
varying a point at which a near-term forecast transitions to the ramp rate forecast within the unified forecast based on system load stability.

### Claim 18

A system of producing a unified near-term and short-term system load forecasting, comprising:
one or more processors configured with executable instructions to perform:
receiving data of supervisory control and data acquisition (SCADA) measurements of an electrical grid;
filtering the data according to a ramp rate filter test and a load level filter test;
smoothing the filtered data according to an augmented Savitzky-Golay filter, wherein a lift multiplier alters Savitzky-Golay smoothing by a factor that is a function of time;
forecasting electrical demand on the electrical grid using a unified framework of near-term and short-term forecast forecasts, wherein the unified framework comprises:
a near-term forecast;
ramp rate forecasts extending beyond an endpoint of the near-term forecast horizon;
a blending of the ramp rate forecast to a short-term forecast to form a unified forecast; and
using the uniform forecast to regulate a spinning reserve of the electrical grid.

### Claim 19 (depends on 18)

The system as recited in claim 18, wherein the blending of the ramp rate forecast to the short-term forecast to form the unified forecast comprises:
bridging a gap between the near-term and short-term forecasts with a plurality of five-minute ramp rate forecasts.

### Claim 20 (depends on 18)

The system as recited in claim 18, additionally comprising:
forecasting approximately 0-15 minutes into the future using the near-term forecast;
forecasting approximately 15-60 minutes into the future using the ramp rate forecasts extending beyond the near-term forecast;
forecasting approximately 60-120 minutes into the future using the blending of the ramp rate forecast to the short-term forecast;
forecasting beyond approximately 120 minutes into the future using, at least in part, the short-term forecast.