Load prediction based on-line and off-line training of neural networks
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
G06G-007/63
G06G-007/00
출원번호
UP-0495482
(2006-07-28)
등록번호
US-7552100
(2009-07-01)
발명자
/ 주소
Chen, Dingguo
출원인 / 주소
Siemens Energy, Inc.
인용정보
피인용 횟수 :
16인용 특허 :
4
초록▼
A method and system is provided for predicting loads within a power system through the training of on-line and an off-line neural networks. Load data and load increments are used with an on-line load prediction scheme to generate predicted load values to optimize power generation and minimize costs
A method and system is provided for predicting loads within a power system through the training of on-line and an off-line neural networks. Load data and load increments are used with an on-line load prediction scheme to generate predicted load values to optimize power generation and minimize costs. This objective is achieved by employing a method and system which predicts short term load trends through the use of historical load data and short term load forecast data.
대표청구항▼
What is claimed is: 1. A method of energy management and short term load prediction in a power system using an off-line neural network and an on-line neural network and a load database having load data captured from the field, each neural network comprising a plurality of neurons to predict a short
What is claimed is: 1. A method of energy management and short term load prediction in a power system using an off-line neural network and an on-line neural network and a load database having load data captured from the field, each neural network comprising a plurality of neurons to predict a short term load demand on the power system, comprising: (a) implementing a first prediction scheme that uses off-line neural network training to output first on-line load predictions, the first prediction scheme comprising: receiving by an off-line neural network historical load data from a load database; training the off-line neural network with the historical load data, resulting in a trained off-line neural network with tuned parameter values Θ containing weights and biases; loading short term load forecast data into the trained off-line neural network; generating first on-line load predictions by the trained off-line neural network; (b) implementing a second prediction scheme that uses on-line neural network training to output second on-line load predictions, the second prediction scheme comprising: initializing parameter values of an on-line neural network with the tuned parameter values Θ from the trained off line neural network; receiving by the on-line neural network current load data for a current time segment from the load database; adjusting the parameter values of the on-line neural network based on the current load data; training the on-line neural network with the current load data, resulting in a trained on-line neural network; loading short term load forecast data into the trained on-line neural network; generating second on-line load predictions by the trained on-line neural network; (c) calculating performance statistics of the first and second prediction schemes wherein the performance statistics are represented by counters, comprising: comparing the first on-line load predictions with actual load values and incrementing a first counter if the difference between the first on-line load predictions and the actual load values are within a predetermined range; comparing the second on-line load predictions with actual load values and incrementing a second counter if the difference between the second on-line load predictions and the actual load values are within a predetermined range; (d) selecting either the first or second on-line load predictions as final predicted load values based on a decision algorithm using the calculated performance statistics; and (e) scheduling energy generation of the power system based on the final predicted load values. 2. The method of claim 1, further comprising selecting the first prediction scheme to use in calculating the final predicted load values if the value in the first counter is greater than the value in the second counter; and selecting the second prediction scheme to use in calculating the final predicted load values if the value in the second counter is greater than the value in the first counter. 3. The method of claim 1 wherein the step of training the off-line neural network comprises the step of calculating load increments between load data, and using the load increments to train the off-line neural network. 4. The method of claim 3, wherein the step of training the off-line neural network comprises the step of normalizing the load increments and using the normalized load increments to train the off-line neural network. 5. The method of claim 4, wherein the step of using the load increments to train the off-line neural network comprises the step of using the load increments to calculate the first on-line load predictions. 6. The method of claim 4, wherein the step of normalizing the load increments comprises the step of using a standard deviation from a best matching date calculated in off-line neural network training. 7. The method of claim 1 wherein the step of training the on-line neural network comprises the step of calculating load increments between load data, and using the load increments to train the on-line neural network. 8. The method of claim 7, wherein the step of training the on-line neural network comprises the step of normalizing the load increments and using the normalized load increments to train the on-line neural network. 9. The method of claim 8, wherein the step of using the load increments to train the on-line neural network comprises the step of using the load increments to calculate the third set of on-line predicted load values. 10. The method of claim 8, wherein the step of normalizing the load increments comprises the step of using a standard deviation from a best matching date calculated in on-line neural network training. 11. The method of claim 10 further comprising the step of using gradient optimization to train the on-line neural network. 12. The method of claim 6 further comprising the step of using gradient optimization to train the off line neural network. 13. The method of claim 2 further comprising generating an accuracy probability as a ratio of the first counter over the combined first and second counters. 14. The method of claim 13 further comprising the step of comparing the generated probabilities and selecting the load prediction scheme with the highest probability. 15. A system for energy management and short term load prediction in a power system using an off line neural network and an on-line neural network and load data captured from the field, each neural network comprising a plurality of neurons to predict a short term load demand on the power system, comprising: a processor for executing an energy management application; an off-line neural network and an on-line neural network, each in communication with the processor; a load database for storing current and historical load data accessible by the processor, the off-line neural network and the on-line neural network; wherein said energy management application causes the processor to perform the steps of: (a) implementing a first prediction scheme that uses off-line neural network training to output first on-line load predictions, the first prediction scheme comprising: receiving by an off-line neural network historical load data from a load database; training the off-line neural network with the historical load data, resulting in a trained off-line neural network with tuned parameter values Θ containing weights and biases; loading short term load forecast data into the trained off-line neural network; generating first on-line load predictions by the trained off-line neural network; (b) implementing a second prediction scheme that uses on-line neural network training to output second on-line load predictions, the second prediction scheme comprising: initializing parameter values of an on-line neural network with the tuned parameter values Θ from the trained off-line neural network; receiving by the on-line neural network current load data for a current time segment from the load database; adjusting the parameter values of the on-line neural network based on the current load data; training the on-line neural network with the current load data, resulting in a trained on-line neural network; loading short term load forecast data into the trained on-line neural network; generating second on-line load predictions by the trained on-line neural network; (c) calculating performance statistics of the first and second prediction schemes, wherein the performance statistics are represented by counters, comprising: comparing the first on-line load predictions with actual load values and incrementing a first counter if the difference between the first on-line load predictions and the actual load values are within a predetermined range; comparing the second on-line load predictions with actual load values and incrementing a second counter if the difference between the second on-line load predictions and the actual load values are within a predetermined range; (d) selecting either the first or second on-line load predictions as final predicted load values based on a decision algorithm using the calculated performance statistics; and (e) scheduling energy generation of the power system based on the final predicted load values. 16. The system of claim 15, wherein the energy management application causes the processor to perform the further step of selecting the first prediction scheme to use in calculating the final predicted load values if the value in the first counter is greater than the value in the second counter; and selecting the second prediction scheme to use in calculating the final predicted load values if the value in the second counter is greater than the value in the first counter. 17. The system of claim 16, further comprising generating an accuracy probability as a ratio of the first counter over the combined first and second counters. 18. The system of claim 15, wherein the step of training the off line neural network and the on-line neural network comprises the step of calculating load increments between load data, and using the load increments to train the off line neural network and on-line neural network. 19. The system of claim 18, wherein the step of training the off-line neural network and on-line neural network comprises the step of normalizing the load increments and using the normalized load increments to train the off-line neural network and on-line neural network. 20. The system of claim 19, wherein the step of normalizing the load increments comprises the step of using a standard deviation from a best matching date calculated in off-line neural network training. 21. A computer-readable medium for use in energy management and short term load prediction in a power system using load data captured from the field, said computer-readable medium having stored thereon instructions which when executed by a processor, cause the processor to perform the steps of: (a) implementing a first prediction scheme that uses off-line neural network training to output first on-line load predictions, the first prediction scheme comprising: receiving by an off-line neural network historical load data from a load database; training the off-line neural network with the historical load data, resulting in a trained off-line neural network with tuned parameter values Θ containing weights and biases; loading short term load forecast data into the trained off-line neural network; generating first on-line load predictions by the trained off-line neural network; (b) implementing a second prediction scheme that uses on-line neural network training to output second on-line load predictions, the second prediction scheme comprising: initializing parameter values of an on-line neural network with the tuned parameter values Θ from the trained off-line neural network; receiving by the on-line neural network current load data for a current time segment from the load database; adjusting the parameter values of the on-line neural network based on the current load data; training the on-line neural network with the current load data, resulting in a trained on-line neural network; loading short term load forecast data into the trained on-line neural network; generating second on-line load predictions by the trained on-line neural network; (c) calculating performance statistics of the first and second prediction schemes, wherein the performance statistics are represented by counters, comprising: comparing the first on-line load predictions with actual load values and incrementing a first counter if the difference between the first on-line load predictions and the actual load values are within a predetermined range; comparing the second on-line load predictions with actual load values and incrementing a second counter if the difference between the second on-line load predictions and the actual load values are within a predetermined range; (d) selecting either the first or second on-line load predictions as final predicted load values based on a decision algorithm using the calculated performance statistics; and (e) scheduling energy generation of the power system based on the final predicted load values.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (4)
Tzes Anthony (Brooklyn NY) Tsotras Vassilis (New York NY), Adaptive system and method for predicting response times in a service environment.
Nomura Masahide (Hitachi JPX) Saito Tadayoshi (Hitachi JPX) Matsumoto Hiroshi (Ibaraka JPX) Shimoda Makoto (Katsuta JPX) Kondoh Masakazu (Hitachi JPX) Miyagaki Hisanori (Ohta JPX) Sugano Akira (Katsu, Process control method and system for performing control of a controlled system by use of a neural network.
Harley,Ronald Gordon; Habelter,Thomas G.; Lambert,Franklin Cook; Mazumdar,Joy, System and method for determining harmonic contributions from non-linear loads.
Al-Mohssen, Husain; Paradis, Richard R.; Bassa, Angela S., Apparatus and method for employing weather induced facility energy consumption characterizations in a demand response dispatch system.
Al-Mohssen, Husain; Bassa, Angela S.; Paradis, Richard R., Demand response dispatch system employing weather induced facility energy consumption characterizations.
Fein, Gene S.; Merritt, Edward, System and method for data processing and transferring in a multi computer environment for energy reporting and forecasting.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.