Presentation on theme: "ERCOT Load Forecasting Forum"— Presentation transcript:
1ERCOT Load Forecasting Forum Pattern Recognition Technologies (PRT), Inc. On-Line Load Forecasting ServicesAl Khotanzad, Ph.D., P.E.PresidentPRT, Inc.17950 Preston Road, Suite 916Dallas, Texas(214)ERCOT Load Forecasting ForumJanuary 24, 2007
2Corporate Profile Founded in 1994 Products & Services Online load and price forecasting servicesStand-alone load and price forecasting softwareCustom forecasting solutionsClientsOver 90 energy firms consisting ofUtilities in North America & OverseasISOs, Municipalities, Coops. Government AgenciesPower marketing and trading organizationsThe Electric Power Research Institute (EPRI)First company to develop a commercial neural network based load forecaster in the early 90’s – ANNSTLF for EPRIPRT, Inc.
3Load ForecastingAccurate forecast of future demand required by all entities involved in the energy marketsElectric UtilitiesIndependent System OperatorsPower MarketersDifferent forecast horizonsLong Term: Several years out – required for planning purposesMid Term: Several weeks to months – scheduling maintenance, planning fuel supply, transactionsShort Term: Next hour to next week – daily operation, energy transactions, reliability studiesPRT, Inc.
4Short Term Load Forecasting Hourly or sub-hourly forecasts starting from next hour to next seven to ten daysForecasts used for:Unit commitmentselection of generators in operationstart up/shut down of generation to minimize operation costHydro scheduling to optimize water release from reservoirsGenerator type coordination to determine the least cost operation mode (optimum mix)Interchange scheduling and energy purchaseTransmission line loadingPower system security assessmentAccuracy has significant economic impactEven a 0.5% improvement in accuracy can result in thousands of dollars in savingsPRT, Inc.
5Factors Affecting Short Term Load 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 trendsTime of day (morning, afternoon, night)Day of week (weekdays, weekends)Holidays (Christmas, New Years)Special events (popular sporting events or TV shows)Demand side managementRandom disturbancesForecasts used for:Unit commitmentselection of generators in operationstart up/shut down of generation to minimize operation costHydro scheduling to optimize water release from reservoirsGenerator type coordination to determine the least cost operation mode (optimum mix)Interchange scheduling and energy purchaseTransmission line loadingPower system security assessmentAccuracy has significant economic impactEven a 0.5% improvement in accuracy can result in thousands of dollars in savingsPRT, Inc.
7Major STLF TechniquesAny STLF technique attempts to model the relationship between the load and factors that affect it – these relationships are nonlinear and complexRegression modelsStochastic time seriesSpectral decompositionSimilar-day searchIntelligent system based modelsSuperiority of intelligent system based techniques have been demonstrated in many studiesPRT, Inc.
8PRT’s LF TechnologiesProducts & services are based on cutting-edge intelligent system technologies of:Artificial Neural NetworksFuzzy LogicGenetic Algorithms/Evolutionary ComputingPRT, Inc.
9Artificial Neural Networks (ANNs) Neurologically inspired systems consisting of highly interconnected elementary computational units (neurons)Distributed processing by neurons results in intelligent outcomeANNs learn to perform a desired task directly from examples using special training algorithmsANNs can generalize; after training, they can produce good results for data that only broadly resembles the data they were trained on originallyANNs are nonlinear systems, well suited for real world problems that are often nonlinearPRT, Inc.
10Forecasting Using ANNs 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 effectsTrain with historical data (examples of the underlying relationship)A properly trained ANN can predict the outcome of the modeled process based on the available observationsANN 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 studentsPRT, Inc.
11Unique Aspects of PRT ANN Forecasters Architecture of ANN specifically designed for electric load forecastingOptimal set of inputs selected for load forecasting applicationNo need for frequent re-trainingQuick response to deviations between forecast and actual loadSpecial algorithms for unusual days, e.g., weekday holidaysPRT, Inc.
12Fuzzy LogicFuzzy logic (FL) is a means to transform subjective/expert knowledge about a process expressed in the form of linguistic rules into computer algorithms.FL employs fuzzy sets, fuzzy membership functions and fuzzy if-then rules to model the uncertainty in nature, and express the knowledgeA fuzzy set is a set without a crisp, clearly defined boundary, and can contain fuzzy variables with a partial degree of membershipPRT, Inc.
13Fuzzy Rule & Membership Function 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 highSubjective interpretation of “hot temperature” or “high load”Characterized by fuzzy membership function – an example shownIn Mamdani model both fuzzy premise part and consequence part are represented in linguistic terms.In Takagi-Sugeno model, linguistic term is only for the fuzzy premise partPRT, Inc.
14Fuzzy Logic Based Load Forecaster Develop applicable fuzzy membership functionsExtract relevant IF-THEN rules from historical data – There could be hundreds of such rulesDuring the forecasting phase several of the rules become activated along with some of the fuzzy membership functionFuzzy inference engine converts all this information into a final crisp forecastPRT, Inc.
15Genetic Algorithms (GAs) Genetic Algorithms (GAs) are optimization algorithms that are based on the concept of natural evolutionGAs can find the optimal solution quickly and efficiently, especially when there is little information about the solution available.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.PRT, Inc.
16GA Based ForecasterCreate M sets of forecasts (in random) for a given set of actual historical dataSort based on accuracyRetain the top K most accurate sets (stronger solutions) and discard the rest (weaker solutions) – survival of the fittestUse the retained K sets as parents to create a second generation of M solutions through mutation & crossover – repeat the processAfter several generation, the top K solutions converge toward a single solution – Strongest solutionThis is the optimal solution used as the final forecasting modelPRT, Inc.
17PRT’s e-ISOForecast Price & Load Forecasting Service 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 AmericaERCOT, PJM, NY-ISO, ISO-NE, MISO, CA-ISO, ONTARIO IESO, ALBERTA AESOHourly forecasts for current day and six days beyondHourly load forecasts for one year out using various simulated weather scenariosForecasts are posted onSubscribers use a Web browser to access and download the forecasts – available 24/7Forecasts are updated every hour or faster based on the most recent price/load/weather data that become availableWeather forecasts are used in the models - updated several times per dayPRT, Inc.
18e-ISOForecast Price & Load Forecasts ERCOTSystem-Wide & Congestion Zone Load ForecastsZonal Market Clearing Price ForecastsPJMSystem-Wide, Regional and Zonal Load ForecastsReal-Time & Day-Ahead LMP Price ForecastsISO New England (ISO-NE)System-Wide & Zonal Load ForecastsZonal Real-Time & Day-Ahead LMP Price ForecastsNew York ISO (NYISO)PRT, Inc.
19e-ISOForecast Price & Load Forecasts Midwest ISO (MISO)System-Wide Load ForecastReal-Time & Day-Ahead LMP Price Forecasts for Five Hubs andVarious CPNsCalifornia ISOZonal Supplemental Real-Time Price Forecast sONTARIO EISOSystem-Wide Price ForecastALBERTA AESOPRT, Inc.
20Forecasting EnginesMultiple models based on different technologies run in parallel generating independent forecastsA top layer of intelligence decides to:Select one of the forecasts as the final forecastCombine multiple forecasts (“Combination of Experts”) into a final forecastAccuracy is improved over use of a single modeling techniquePRT, Inc.
21Weather DataPRT has affiliations with two major weather service providers, WSI and MeteorlogixMost free internet based weather forecast services simply provide forecasts generated by NWS or other computer modelsWeather service providers bring human meteorologists in the loop who scrutinize/edit computer generated forecastsWeather forecasts updated several times throughout the dayActual temperature updated every hour and with every update, new load forecasts are generatedPRT, Inc.
22Access via the WebForecasts are posted to a dedicated password protected pageCan be accessed using any standard Web browser from any computerProvides easy access for all in the companyForecasts are displayed in tabular and graphical formsActual data of previous day and any available data of current day are displayedForecasts can be downloaded in EXCEL formatOther statistics including actual prices of past week, similar day comparisons and price bands are providedPRT, Inc.
27e-ISOForecast LF Performance for 2006 Forecasts Recorded at 8 am CTLoad: Hourly MAPE/ Daily Peak Load MAPETemperature: Hourly MAD/Daily Peak Temp MADMarketAfter-the-FactCurrentDayNext DayDay 3Day 4Day 5ERCOTLoad1.51/1.841.81/2.692.99/3.483.50/3.873.99/4.594.54/5.13Temp1.38/1.732.14/1.982.41/2.132.86/2.683.21/3.02PJM East1.24/1.331.32/1.912.49/2.642.88/2.933.25/3.453.67/3.811.48/2.002.56/2.462.73/2.783.20/3.463.68/4.05ISONE1.70/1.601.24/1.882.30/2.322.73/2.913.04/3.313.34/3.601.49/2.002.47/2.572.76/2.933.27/3.623.78/4.32NYISO1.15/1.282.21/2.402.56/2.592.73/2.742.93/3.053.26/3.472.72/2.813.04/2.993.22/3.233.53/3.603.98/4.30MISO1.21/1.391.21/1.832.29/2.542.79/3.053.20/3.533.54/3.871.29/1.702.15/2.322.37/2.482.82/3.053.32/3.57
28e-ISOForecast Performance for Forecast of Next-Day ERCOT Total Load – 2006 i Forecasts Recorded at 3:00 PM of Previous DayMonthLoad MAPETemp MADJan2.10/2.022.13Jul2.51/2.771.71Feb3.01/3.183.14Aug2.70/2.812.06Mar2.52/2.852.42Sep3.86/5.032.35Apr3.17/4.111.78Oct3.53/5.062.36May3.05/3.571.83Nov2.21/2.252.21Jun2.70/2.921.62Dec2.45/3.492.14PRT, Inc.
29Comparison of PRT and ISO LF Performance Forecasts Recorded at 8 am CTHourly MAPE/ Daily Peak Load MAPEMarketPeriodByCurrent DayNext DayDay 3Day 4Day 5ERCOT1/1-12/31PRT1.81/2.692.99/3.483.50/3.873.99/4.594.54/5.13ISO2.28/2.963.22/3.453.75/4.084.56/4.795.04/5.40PJM East1.32/1.912.49/2.642.88/2.933.25/3.453.67/3.811.70/2.003.30/2.913.32/3.233.79/3.604.27/4.13ISONE1.24/1.882.30/2.322.73/2.913.04/3.313.34/3.603.08/1.773.23/2.15-NYISO12/5-12/310.64/1.061.51/1.541.74/1.851.99/2.292.10/2.342.54/2.622.22/2.092.12/1.592.08/1.43
30e-LoadForecast Service An online load forecast service for company-specific load dataStandard Service: Hourly/sub hourly forecasts for current day and six days beyondExtended Service: Additional Hourly/sub hourly forecasts for several months and years outUser only needs to:Provide historical load data for initial model trainingUpload the most recent actual load data as it becomes available (via FTP, , provided Excel interface)All the required actual and forecast weather data acquired by PRT fromLoad and weather data quality checked and validatedForecasts posted to a dedicated and secure website in tabular and graphical formsPRT, Inc.
31e-LoadForecast Service, Cont’ Forecasts are updated every hour with preceding hour’s actual observed weatherForecasts are updated any time an actual load data is uploaded by user24/7 access throughVia Internet at any locationAn Excel Interface with built-in functions enabling user to remotely interact with the forecasting systemFTPERCOT uses this service for forecast of its eight weather zonesPRT, Inc.
32Forecasting EnginesMultiple models based on different technologies run in parallel generating independent forecastsA top layer of intelligence decides to:Select one of the forecasts as the final forecastCombine multiple forecasts (“Combination of Experts”) into a final forecastAccuracy is improved over use of a single modeling techniquePRT, Inc.
33Weather DataPRT has affiliations with two major weather service providers, WSI and MeteorlogixMost free internet based weather forecast services simply provide forecasts generated by NWS or other computer modelsWeather service providers bring human meteorologists in the loop who scrutinize/edit computer generated forecastsWeather forecasts updated several times throughout the dayActual temperature updated every hour and with every update, new load forecasts are generatedPRT, Inc.
34Other Features The provided Excel Interface allows user to: Modify forecasted temperatures and generate corresponding load forecasts – “What-If” scenariosModify predicted morning and/or afternoon peak loads. Forecasts for other hours are reshaped accordinglyView load and temperature of three most similar days (temperature wise) in the historyPRT, Inc.
35Access via the Web - View & Download Forecasts are posted to a dedicated password protected pageCan be accessed using any standard Web browser from any computerProvides easy access for all in the companyForecasts are displayed in tabular and graphical formsActual data of previous day and any available data of current day are displayedForecasts can be downloaded in EXCEL formatPRT, Inc.
36Access via Excel Interface – View, Download & Interact An Excel interface with easy-to-use built-in featuresDownload and view most current load and temperature forecasts in tabular and graphical formsModify forecasted temperatures and generate corresponding load forecastsModify predicted peak loads and reshape load forecasts accordinglyDownload and view three most similar daysUpload actual load updates to PRT’s serversPRT, Inc.
37Profile Based Forecasting Retailers operating in deregulated markets work with individual accounts that may not be metered hourly (e.g., residential load)Energy transactions and settlements are done based on hourly demandHourly load is simulated using pre-specified standard load profiles for client typeTo forecast their retail load, load profile for each account must be scaled appropriately to account for pattern of usage by that accountProfile based module of e-LoadForecast – User provides:List of current accounts in the portfolio along with their corresponding profile typeThe historical usage data for each accountBackcasted profiles for corresponding profile types are used to develop a profile multiplier (scale factor) for each account using historical meter reads.Forecasted standard profiles are multiplied by the scale factor to get the final hourly forecastPRT, Inc.
38Mid-Term/Long-Term Module Optional service includes mid-term/long-term hourly load forecastForecast horizon can be extended to five years outANN technology is used – Models are different from those used for short-term forecastingImpact of load growth is consideredWeather forecast is needed for the forecast horizonSimulated using historical weather dataThree scenarios of “Normal”, “Hot”, and “Cold” available for each month in forecast horizonAdditional scenarios for generating “High Load” and “Low Load” casesTwo statistical methods available for simulation of scenarios from historical weather dataTools are provided for easy manipulation of simulated weather – user can build heat waves/cold frontsPRT, Inc.
39Quality Control Extensive quality control system in place Actual load and temperature data continually quality checkedDetected anomalies such as spikes and gaps correctedEvery day accuracy of load and temperature forecasts for various forecasts horizons are computed and reviewed by our experienced staffCorrective action taken if degradation in quality detectedAnalysis of the causeCalibrate forecasting modelsUse of different kind of forecasting enginesPRT, Inc.
40Forecasting Service Advantages Uses state-of-the-art load forecasting modelsMore accurate forecasts than in-house systemsMore economical than maintaining an in-house systemFrees up valuable manpower & resourcesNo data hassles, IT overhead, software maintenance & upgradePerformance continuously monitored by specialists with extensive experience and background in forecastingModels are continually calibrated and upgradedConvenient access to forecasts for all who need it in the organizationUnlimited use by all in the organizationPRT, Inc.