LMP Forecasting
Methods and Applications

December 14-15, 2023 | Live Streaming Online

A Program

Click Here to register $1295

If you are unable to attend at the scheduled date and time, we make recordings available to all registrants for three business days after the event

This course examines several approaches to predicting electricity prices for multiple purposes over different time horizons.  It begins with an overview of the characteristics of electricity prices and later provides the physical and mathematical background to explain the reasons for those characteristics.   Both fundamentals-based and time series-based approaches will be described in practical terms.  Forecasting approaches will be recommended based on the end use, time horizon and which data are (un)available, but all approaches offer the potential to forecast prices at multiple locations of interest at the same time.  Distinctions will be made between when high accuracy (for both timing and magnitude) forecasts are required by some use cases and when they are not required.  Specific real-world uses cases will be explored, using real-world data.

The content will also provide many insights into the quantities to be forecast and, from there, cover guidance about forecasting prices – with or without sophisticated modeling tools.  Emphasis will be placed on forecasting prices in U.S. locational marginal pricing (LMP) style markets.   A survey of forecasting approaches will be presented, examining their applicability to specific forecasting time horizons.  And, for those who are not professional forecasters, a discussion of “Do It Yourself” forecasting will round out the course.

Learning Outcomes

Using real-world situations and illustrative examples, this instructional course will:

  • Review the characteristics of Locational Marginal Prices
  • Discuss the value of forecast accuracy in multiple use cases
  • Explain the underlying structure that determines electricity prices
  • Illustrate locational marginal price (LMP) formation
  • Address real-world applications, questions and solutions
  • Review four time horizons and the appropriate price forecasting approaches for each
  • Demonstrate what can be done with and without sophisticated modeling tools



9:00 a.m. – 5:00 p.m. Central Time

9:00 – 9:15 a.m. :: Overview and Introductions

9:15 – 10:30 a.m. :: Characteristics Of Wholesale Electricity Prices

  • Nodal and Zonal locations
  • Day-Ahead and Real-Time LMPs (1-hour and 5-minute intervals)
  • Periodicity and Seasonality
  • Typical distributions
  • Trends, Correlations and Averages
  • Discontinuities (regime shifts over time)
  • Spikes (positive and negative)
  • Negative prices
  • Predictability over different time horizons
  • Prediction accuracy required for various applications
    • Resource siting decision (which market, state and substation)
    • Resource investment and retirement decisions
    • Generation bidding (gas nomination, water release, etc.)
    • Transmission expansion plans
    • Generation investment and retirement decisions
    • PPA negotiation (CfD, basis, curtailment, negative LMP risks)
    • Trading (physical, virtual, speculative)
    • Financial transmission rights (FTR /TCR /TCC /CRR)
    • Regulatory and legislative (market reform debate)
    • Mark-to-Model

10:30 – 10:45 a.m. :: Morning Break

10:45 – 11:20 a.m. :: Two Categories of Price Forecasting Models

  • Fundamentals-based
    • Economic dispatch and constrained power flow
    • Transmission network model needed
    • Individual generator economics and location
    • Consumer location in the network
  • Data-driven
    • Time series, regression, ARIMA, ANN
    • Machine learning
    • Physics-informed models
  • Addressing uncertainty of forecast inputs

11:20 – 11:45 a.m. :: Behind the Scenes of Electricity Markets

  • Demand forecasting
  • Unit commitment
  • Economic dispatch
  • Operating reserves
  • Power flow

11:45 a.m. – 12:45 p.m. :: Components of Locational Marginal Prices

  • Marginal energy
  • Marginal loss
  • Marginal congestion
  • Shadow prices
  • Shift factors

12:45 – 1:30 p.m. :: Break for Lunch

1:30 – 3:00 p.m. :: LMP Analysis Leads to LMP Forecasting

  • Forecasting hub LMP versus node (generator) LMP
  • Discontinuity from the past (ex: a new transmission line)
  • Forecasting LMP components vs. forecasting LMP
  • Bespoke models for nodes versus simulating the entire market
  • Fundamentals vs. time series (historical prices)

3:00 – 3:15 p.m. :: Afternoon Break

3:15 – 5:00 p.m. :: Real-World Applications

  • Data sources and detective work
  • LMP forecasts depend on forecasts of the inputs
  • Measuring error, gaining confidence
  • Predicting negative LMPs
  • Predicting LMP spikes (up and down)
  • Good sites for renewables and storage
  • “Fair” forecasting vs. risk analysis
  • Real Time vs. Day Ahead LMPs
  • Credibility, transparency, defensibility

5:00 p.m. :: Course Adjourns for the Day



9:00 a.m. – 12:15 p.m. Central Time


9:00 – 10:00 a.m. :: Long-term (5 – 25 years) Price Forecasting

  • Forecasting under uncertainty (scenario building)
  • Why history is not a good predictor of the (renewable) future
  • Production cost model with deterministic and stochastic inputs
  • Capacity expansion model with exogenous inputs
  • Iterating to get a consistent outlook
  • Expecting the unexpected

10:00 – 10:30 a.m. :: Mid-term (1 – 10 years) Price Forecasting

  • Impactful inputs for forecasting hub and node LMP
  • Statistical models
  • Production cost models (market software simulation)
  • Change is the only constant

10:30 – 10:45 a.m. :: Morning Break

10:45 – 11:15 a.m. :: Short-term (1 week – 1 year) Price Forecasting

  • Parameters and input variables
  • Time series / statistical models
  • Machine learning
  • Production cost models
  • Strengths and weaknesses, pitfalls, accuracy

11:15 – 11:45 a.m. :: Intra-day and Next-day Price Forecasting

  • Choosing and Updating Inputs
  • Time Series / Statistical Models
  • Machine Learning
  • Supplemental information (e.g., planned outages)

11:15 – 11:45 a.m. :: Intra-day and Next-day Price Forecasting

  • Choosing and Updating Inputs
  • Time Series / Statistical Models
  • Machine Learning
  • Supplemental information (e.g., planned outages)

11:45 a.m. – 12:15 p.m. :: What Can Be Done without Coding

  • Creating models in MS Excel
  • Scenario-building
  • Consultants

12:15 p.m. :: Course Adjourns


Nicholas Pratley has worked in electrical power systems analysis and design for more than three decades.  He began his career as a consulting engineer designing power distribution for heavy industrial projects and moved on to become a subject matter expert in power systems analysis for software vendors and consulting firms. He has experience with generation interconnection studies and transmission planning, including more than 15 years in electricity market simulation and price forecasting. Mr. Pratley has received training on Plexos, GE MAPS and PROMOD IV production cost models, and provided training and consulting services using PROMOD for many years. Clients have included transmission owners and developers, municipal utilities deciding whether to join an ISO market or where to construct a new generating plant, competitive generation (IPP) developers forecasting transmission congestion risk for wind/solar/storage project siting and contract negotiation, and pension funds investing in renewable generation projects. His work has influenced decisions to acquire, divest, build or invest in dozens of projects worth multiple billions of dollars. In his current professional capacity, Mr. Pratley focuses on the U.S. northeast wholesale electricity markets, federal and state policy, and transmission plans with respect to large-scale renewable energy development.  He earned degrees in Electrical Engineering at McGill University and the Université de Montréal polytechnical school.