Building system with selective use of data for probabilistic forecasting

A method for generating input forecast data, the method comprising:

• obtaining, by a processing circuit, a historical prediction set of measurements and a historical actual set of measurements for a plurality of time-steps, each of the plurality of time-steps associated with a historical value of the historical prediction set of measurements and a historical value of the historical actual set of measurements;
• determining, by the processing circuit, whether historical values of the historical prediction set of measurements are within a tolerance threshold of corresponding historical values of the historical actual set of measurements;
• identifying, by the processing circuit, time-steps that are associated with historical values of the historical prediction set of measurements that are outside of the tolerance threshold;
• retrieving, by the processing circuit, a current prediction set of measurements comprising current values associated with the plurality of time-steps;
• replacing, by the processing circuit, each current value of the current prediction set of measurements that is associated with the identified time-steps with a predetermined value to generate an updated prediction set of measurements;
• and
• providing, by the processing circuit, the updated prediction set of measurements to a prediction model.

A non-transitory computer-readable storage medium having instructions stored thereon that, upon execution by a processor, cause the processor to perform operations to generate forecast input data, the operations comprising:

• obtaining a historical prediction set of measurements and a historical actual set of measurements for a plurality of time-steps, each of the plurality of time-steps associated with a historical value of the historical prediction set of measurements and a historical value of the historical actual set of measurements;
• determining whether historical values of the historical prediction set of measurements are within a tolerance threshold of corresponding historical values of the historical actual set of measurements;
• identifying time-steps that are associated with historical values of the historical prediction set of measurements that are outside of the tolerance threshold;
• retrieving a current prediction set of measurements comprising current values associated with the plurality of time-steps;
• replacing each current value of the current prediction set of measurements that is associated with the identified time-steps with a predetermined value to generate an updated prediction set of measurements;
• and
• providing the updated prediction set of measurements to a prediction model.