Optimizing Merchant Power Generation for Deregulated Power Markets

Challenge

Matching electrical energy consumption with the right level of supply is crucial, because excess electricity cannot be stored, unless converted to other forms, which incurs additional costs and resources. At the same time, underestimating energy consumption could be fatal, with excess demand overloading the supply line and even causing blackouts. Clearly, there are tangible benefits in closely monitoring the energy consumption; be they by city, state or country.

Solution

A great variety of methods for predicting electricity demand are being used by electrical companies, which apply to short-term and long-term forecasting. With the large-scale adoption of machine learning and especially deep learning techniques, accurately predicting future energy consumption becomes increasingly possible. Accurate predictions provide two-fold benefits: first, operations staff can gain key insights into future energy demand, providing opportunities to plan their power plant operations and improve overall efficiency. Secondly, forecasts provide a benchmark to single out anomalously high/low energy consumption periods and alert the operations staff.

Approach

A great variety of methods for predicting electricity demand are being used by electrical companies, which apply to short-term and long-term forecasting. With the large-scale adoption of machine learning and especially deep learning techniques, accurately predicting future energy consumption becomes increasingly possible. Accurate predictions provide two-fold benefits: first, operations staff can gain key insights into future energy demand, providing opportunities to plan their power plant operations and improve overall efficiency. Secondly, forecasts provide a benchmark to single out anomalously high/low energy consumption periods and alert the operations staff.

Deep learning techniques such as Recurrent neural networks (RNNs) and LSTMs which are a variant of RNNs have historically proven better at forecasting time series data than linear methods such as ARIMA or SARIMA. Here at BKO services, we use some of the most advanced deep learning techniques such as Bayesian Deep learning methods to forecast energy demands. This is advantageous over conventional deep learning methods as Bayesian deep learning gives a confidence interval for every forecast instead of just a point estimate and we can take into account the room of error as well. This room of error is particularly important because it helps the operations staff to know the best and worst possible scenarios beforehand, giving them a competitive advantage in the market.

COLLECT

DATA

ORGANIZE

DATA

BUILD

MODELS

CODIFY

MODELS

OPTIMIZE

&

MONITOR


Conclusion

BKO Services designed and built a machine learning based solution for an Independent Power Producer that will enable the IPP to improve profitability by several millions of dollars annually.

Get In Touch

Project
Project
Freelance
Business
Affiliate