Presentation is loading. Please wait.

Presentation is loading. Please wait.

Pattern Recognition Technologies (PRT), Inc. On-Line Load Forecasting Services Al Khotanzad, Ph.D., P.E. President PRT, Inc. 17950 Preston Road, Suite.

Similar presentations


Presentation on theme: "Pattern Recognition Technologies (PRT), Inc. On-Line Load Forecasting Services Al Khotanzad, Ph.D., P.E. President PRT, Inc. 17950 Preston Road, Suite."— Presentation transcript:

1 Pattern Recognition Technologies (PRT), Inc. On-Line Load Forecasting Services Al Khotanzad, Ph.D., P.E. President PRT, Inc Preston Road, Suite 916 Dallas, Texas (214) ERCOT Load Forecasting Forum January 24, 2007

2 PRT, Inc.2 Corporate Profile l Founded in 1994 Products & Services l Online load and price forecasting services l Stand-alone load and price forecasting software l Custom forecasting solutions Clients l Over 90 energy firms consisting of  Utilities in North America & Overseas  ISOs, Municipalities, Coops. Government Agencies  Power marketing and trading organizations  The Electric Power Research Institute (EPRI) l First company to develop a commercial neural network based load forecaster in the early 90’s – ANNSTLF for EPRI

3 PRT, Inc.3 Load Forecasting l Accurate forecast of future demand required by all entities involved in the energy markets  Electric Utilities  Independent System Operators  Power Marketers l Different forecast horizons  Long Term: Several years out – required for planning purposes  Mid Term: Several weeks to months – scheduling maintenance, planning fuel supply, transactions  Short Term: Next hour to next week – daily operation, energy transactions, reliability studies

4 PRT, Inc.4 Short Term Load Forecasting l Hourly or sub-hourly forecasts starting from next hour to next seven to ten days l Forecasts used for:  Unit commitment –selection of generators in operation –start up/shut down of generation to minimize operation cost  Hydro scheduling to optimize water release from reservoirs  Generator type coordination to determine the least cost operation mode (optimum mix)  Interchange scheduling and energy purchase  Transmission line loading  Power system security assessment l Accuracy has significant economic impact l Even a 0.5% improvement in accuracy can result in thousands of dollars in savings

5 PRT, Inc.5 Factors Affecting Short Term Load l Factors affecting short-term load are:  Mix of customer in the service area (residential, commercial, industrial)  Weather condition (temperature, humidity, cloud cover, wind speed)  Seasonal effects & recent load trends  Time of day (morning, afternoon, night)  Day of week (weekdays, weekends)  Holidays (Christmas, New Years)  Special events (popular sporting events or TV shows)  Demand side management  Random disturbances l Forecasts used for:  Unit commitment –selection of generators in operation –start up/shut down of generation to minimize operation cost  Hydro scheduling to optimize water release from reservoirs  Generator type coordination to determine the least cost operation mode (optimum mix)  Interchange scheduling and energy purchase  Transmission line loading  Power system security assessment l Accuracy has significant economic impact l Even a 0.5% improvement in accuracy can result in thousands of dollars in savings

6 PRT, Inc.6 Example of Load and Temperature

7 PRT, Inc.7 Major STLF Techniques l Any STLF technique attempts to model the relationship between the load and factors that affect it – these relationships are nonlinear and complex  Regression models  Stochastic time series  Spectral decomposition  Similar-day search  Intelligent system based models l Superiority of intelligent system based techniques have been demonstrated in many studies

8 PRT, Inc.8 PRT’s LF Technologies l Products & services are based on cutting-edge intelligent system technologies of:  Artificial Neural Networks  Fuzzy Logic  Genetic Algorithms/Evolutionary Computing

9 PRT, Inc.9 Artificial Neural Networks (ANNs) l Neurologically inspired systems consisting of highly interconnected elementary computational units (neurons) l Distributed processing by neurons results in intelligent outcome l ANNs learn to perform a desired task directly from examples using special training algorithms l ANNs can generalize; after training, they can produce good results for data that only broadly resembles the data they were trained on originally l ANNs are nonlinear systems, well suited for real world problems that are often nonlinear

10 PRT, Inc.10 Forecasting Using ANNs l A key feature of ANNs is their ability to learn a complex pattern mapping, i.e., model the underlying relationship between a set of variables and an outcome that is a function of them –  Future Load – Function of past loads and weather, recent load trends, upcoming weather, calendar effects l Train with historical data (examples of the underlying relationship) l A properly trained ANN can predict the outcome of the modeled process based on the available observations l ANN based predictors employed in a wide variety of forecasting applications such as prediction of: electric load, weather, gas consumption, stock market, economic trends time series data, future sales, traffic patterns and grade point average of students

11 PRT, Inc.11 Unique Aspects of PRT ANN Forecasters l Architecture of ANN specifically designed for electric load forecasting l Optimal set of inputs selected for load forecasting application l No need for frequent re-training l Quick response to deviations between forecast and actual load l Special algorithms for unusual days, e.g., weekday holidays

12 PRT, Inc.12 Fuzzy Logic l Fuzzy logic (FL) is a means to transform subjective/expert knowledge about a process expressed in the form of linguistic rules into computer algorithms. l FL employs fuzzy sets, fuzzy membership functions and fuzzy if- then rules to model the uncertainty in nature, and express the knowledge l A fuzzy set is a set without a crisp, clearly defined boundary, and can contain fuzzy variables with a partial degree of membership

13 PRT, Inc.13 Fuzzy Rule & Membership Function l An example of a typical fuzzy IF-Then rule :  IF next-day temperature is hot, and today’s temperature is hot, THEN next-day load is high l Subjective interpretation of “hot temperature” or “high load” Characterized by fuzzy membership function – an example shown

14 PRT, Inc.14 Fuzzy Logic Based Load Forecaster l Develop applicable fuzzy membership functions l Extract relevant IF-THEN rules from historical data – There could be hundreds of such rules l During the forecasting phase several of the rules become activated along with some of the fuzzy membership function l Fuzzy inference engine converts all this information into a final crisp forecast

15 PRT, Inc.15 Genetic Algorithms (GAs) l Genetic Algorithms (GAs) are optimization algorithms that are based on the concept of natural evolution l GAs can find the optimal solution quickly and efficiently, especially when there is little information about the solution available. l GAs emulate natural evolution, and make use of four operators, including reproduction, crossover, mutation, and survival of the fittest to produce and keep the optimal solutions.

16 PRT, Inc.16 GA Based Forecaster l Create M sets of forecasts (in random) for a given set of actual historical data l Sort based on accuracy l Retain the top K most accurate sets (stronger solutions) and discard the rest (weaker solutions) – survival of the fittest l Use the retained K sets as parents to create a second generation of M solutions through mutation & crossover – repeat the process l After several generation, the top K solutions converge toward a single solution – Strongest solution l This is the optimal solution used as the final forecasting model

17 PRT, Inc.17 PRT’s e-ISOForecast Price & Load Forecasting Service l e-ISOForecast is an on-line real-time price & load forecasting service that has been set up for all wholesale power markets/ISOs in North America  ERCOT, PJM, NY-ISO, ISO-NE, MISO, CA-ISO, ONTARIO IESO, ALBERTA AESO l Hourly forecasts for current day and six days beyond l Hourly load forecasts for one year out using various simulated weather scenarios l Forecasts are posted on l Subscribers use a Web browser to access and download the forecasts – available 24/7 l Forecasts are updated every hour or faster based on the most recent price/load/weather data that become available l Weather forecasts are used in the models - updated several times per day

18 PRT, Inc.18 e-ISOForecast Price & Load Forecasts l ERCOT  System-Wide & Congestion Zone Load Forecasts  Zonal Market Clearing Price Forecasts l PJM  System-Wide, Regional and Zonal Load Forecasts  Real-Time & Day-Ahead LMP Price Forecasts l ISO New England (ISO-NE)  System-Wide & Zonal Load Forecasts  Zonal Real-Time & Day-Ahead LMP Price Forecasts l New York ISO (NYISO )  System-Wide & Zonal Load Forecasts  Zonal Real-Time & Day-Ahead LMP Price Forecasts

19 PRT, Inc.19 e-ISOForecast Price & Load Forecasts l Midwest ISO (MISO)  System-Wide Load Forecast  Real-Time & Day-Ahead LMP Price Forecasts for Five Hubs and Various CPNs l California ISO  System-Wide Load Forecast  Zonal Supplemental Real-Time Price Forecast s l ONTARIO EISO  System-Wide Load Forecast  System-Wide Price Forecast l ALBERTA AESO  System-Wide Load Forecast  System-Wide Price Forecast

20 PRT, Inc.20 Forecasting Engines l Multiple models based on different technologies run in parallel generating independent forecasts l A top layer of intelligence decides to:  Select one of the forecasts as the final forecast  Combine multiple forecasts (“Combination of Experts”) into a final forecast l Accuracy is improved over use of a single modeling technique

21 PRT, Inc.21 Weather Data l PRT has affiliations with two major weather service providers, WSI and Meteorlogix l Most free internet based weather forecast services simply provide forecasts generated by NWS or other computer models l Weather service providers bring human meteorologists in the loop who scrutinize/edit computer generated forecasts l Weather forecasts updated several times throughout the day l Actual temperature updated every hour and with every update, new load forecasts are generated

22 PRT, Inc.22 Access via the Web l Forecasts are posted to a dedicated password protected page l Can be accessed using any standard Web browser from any computer l Provides easy access for all in the company l Forecasts are displayed in tabular and graphical forms l Actual data of previous day and any available data of current day are displayed l Forecasts can be downloaded in EXCEL format l Other statistics including actual prices of past week, similar day comparisons and price bands are provided

23 PRT, Inc.23 e-ISOForecast Main Page

24 PRT, Inc.24 e-ISOForecast PJM Segment

25 PRT, Inc.25 e-ISOForecast Load Forecast View

26 PRT, Inc.26 e-ISOForecast LF View - Graphical

27 Market After- the-Fact Current Day Next Day Day 3Day 4Day 5 ERCOT Load1.51/ / / / / /5.13 Temp1.38/ / / / /3.02 PJM East Load1.24/ / / / / /3.81 Temp1.48/ / / / /4.05 ISONELoad1.70/ / / / / /3.60 Temp1.49/ / / / /4.32 NYISOLoad1.15/ / / / / /3.47 Temp2.72/ / / / /4.30 MISOLoad1.21/ / / / / /3.87 Temp1.29/ / / / /3.57 e-ISOForecast LF Performance for 2006 Forecasts Recorded at 8 am CT Load: Hourly MAPE/ Daily Peak Load MAPE Temperature: Hourly MAD/Daily Peak Temp MAD

28 PRT, Inc.28 e-ISOForecast Performance for Forecast of Next-Day ERCOT Total Load – 2006 i Forecasts Recorded at 3:00 PM of Previous Day MonthLoad MAPE Temp MAD MonthLoad MAPE Temp MAD Jan2.10/ Jul2.51/ Feb3.01/ Aug2.70/ Mar2.52/ Sep3.86/ Apr3.17/ Oct3.53/ May3.05/ Nov2.21/ Jun2.70/ Dec2.45/

29 MarketPeriod ByCurrent Day Next Day Day 3Day 4Day 5 ERCOT1/1- 12/31 PRT1.81/ / / / /5.13 ISO2.28/ / / / /5.40 PJM East 1/1- 12/31 PRT1.32/ / / / /3.81 ISO1.70/ / / / /4.13 ISONE1/1- 12/31 PRT1.24/ / / / /3.60 ISO3.08/ / NYISO12/5- 12/31 PRT0.64/ / / / /2.34 ISO2.54/ / / /1.43 Comparison of PRT and ISO LF Performance Forecasts Recorded at 8 am CT Hourly MAPE/ Daily Peak Load MAPE

30 PRT, Inc.30 e-LoadForecast Service l An online load forecast service for company-specific load data l Standard Service: Hourly/sub hourly forecasts for current day and six days beyond l Extended Service: Additional Hourly/sub hourly forecasts for several months and years out l User only needs to:  Provide historical load data for initial model training  Upload the most recent actual load data as it becomes available (via FTP, , provided Excel interface) l All the required actual and forecast weather data acquired by PRT from l Load and weather data quality checked and validated l Forecasts posted to a dedicated and secure website in tabular and graphical forms

31 PRT, Inc.31 e-LoadForecast Service, Cont’ l Forecasts are updated every hour with preceding hour’s actual observed weather l Forecasts are updated any time an actual load data is uploaded by user l 24/7 access through  Via Internet at any location  An Excel Interface with built-in functions enabling user to remotely interact with the forecasting system  FTP  l ERCOT uses this service for forecast of its eight weather zones

32 PRT, Inc.32 Forecasting Engines l Multiple models based on different technologies run in parallel generating independent forecasts l A top layer of intelligence decides to:  Select one of the forecasts as the final forecast  Combine multiple forecasts (“Combination of Experts”) into a final forecast l Accuracy is improved over use of a single modeling technique

33 PRT, Inc.33 Weather Data l PRT has affiliations with two major weather service providers, WSI and Meteorlogix l Most free internet based weather forecast services simply provide forecasts generated by NWS or other computer models l Weather service providers bring human meteorologists in the loop who scrutinize/edit computer generated forecasts l Weather forecasts updated several times throughout the day l Actual temperature updated every hour and with every update, new load forecasts are generated

34 PRT, Inc.34 Other Features l The provided Excel Interface allows user to:  Modify forecasted temperatures and generate corresponding load forecasts – “What-If” scenarios  Modify predicted morning and/or afternoon peak loads. Forecasts for other hours are reshaped accordingly  View load and temperature of three most similar days (temperature wise) in the history

35 PRT, Inc.35 Access via the Web - View & Download l Forecasts are posted to a dedicated password protected page l Can be accessed using any standard Web browser from any computer l Provides easy access for all in the company l Forecasts are displayed in tabular and graphical forms l Actual data of previous day and any available data of current day are displayed l Forecasts can be downloaded in EXCEL format

36 PRT, Inc.36 Access via Excel Interface – View, Download & Interact l An Excel interface with easy-to-use built-in features  Download and view most current load and temperature forecasts in tabular and graphical forms  Modify forecasted temperatures and generate corresponding load forecasts  Modify predicted peak loads and reshape load forecasts accordingly  Download and view three most similar days  Upload actual load updates to PRT’s servers

37 PRT, Inc.37 Profile Based Forecasting l Retailers operating in deregulated markets work with individual accounts that may not be metered hourly (e.g., residential load) l Energy transactions and settlements are done based on hourly demand l Hourly load is simulated using pre-specified standard load profiles for client type l To forecast their retail load, load profile for each account must be scaled appropriately to account for pattern of usage by that account l Profile based module of e-LoadForecast – User provides:  List of current accounts in the portfolio along with their corresponding profile type  The historical usage data for each account l Backcasted profiles for corresponding profile types are used to develop a profile multiplier (scale factor) for each account using historical meter reads. l Forecasted standard profiles are multiplied by the scale factor to get the final hourly forecast

38 PRT, Inc.38 Mid-Term/Long-Term Module l Optional service includes mid-term/long-term hourly load forecast l Forecast horizon can be extended to five years out l ANN technology is used – Models are different from those used for short- term forecasting l Impact of load growth is considered l Weather forecast is needed for the forecast horizon  Simulated using historical weather data  Three scenarios of “Normal”, “Hot”, and “Cold” available for each month in forecast horizon  Additional scenarios for generating “High Load” and “Low Load” cases  Two statistical methods available for simulation of scenarios from historical weather data l Tools are provided for easy manipulation of simulated weather – user can build heat waves/cold fronts

39 PRT, Inc.39 Quality Control l Extensive quality control system in place l Actual load and temperature data continually quality checked l Detected anomalies such as spikes and gaps corrected l Every day accuracy of load and temperature forecasts for various forecasts horizons are computed and reviewed by our experienced staff l Corrective action taken if degradation in quality detected  Analysis of the cause  Calibrate forecasting models  Use of different kind of forecasting engines

40 PRT, Inc.40 Forecasting Service Advantages l Uses state-of-the-art load forecasting models l More accurate forecasts than in-house systems l More economical than maintaining an in-house system l Frees up valuable manpower & resources l No data hassles, IT overhead, software maintenance & upgrade l Performance continuously monitored by specialists with extensive experience and background in forecasting l Models are continually calibrated and upgraded l Convenient access to forecasts for all who need it in the organization l Unlimited use by all in the organization


Download ppt "Pattern Recognition Technologies (PRT), Inc. On-Line Load Forecasting Services Al Khotanzad, Ph.D., P.E. President PRT, Inc. 17950 Preston Road, Suite."

Similar presentations


Ads by Google