Process Scheduling Using Input Price Rank Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manufacturers face challenges in accurately predicting short-term prices of commodities like electricity, leading to inefficiencies in scheduling processes that use time-varying inputs, especially when stockpiling is not an option, due to unpredictable price fluctuations and numerous variables affecting demand.
Innovation Solution
The solution involves predicting the 'rank' of the current input price among future prices to determine when it is economically advantageous to run a process, setting operational cutoffs based on this ranking, and adjusting these cutoffs to balance economic and inventory considerations, while minimizing process disruptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manufacturers use traditional price prediction methods to schedule processes, then they can potentially reduce manufacturing costs by running processes when input prices are low, but the predictions are inaccurate due to unpredictable short-term price fluctuations and numerous variables
Solution Approach 1:
Instead of predicting future prices directly, the patent inverts the approach by predicting the rank of current prices among future prices. This inversion transforms an intractable prediction problem (accurate price levels) into a more manageable one (relative ranking), thereby improving prediction accuracy for scheduling decisions.
Solution Approach 2:
The patent changes the parameter being predicted from absolute price levels to relative price ranks. This parameter transformation makes the prediction problem more tractable by focusing on the ordinal position of current prices in the future price distribution rather than attempting to predict exact future price values, which are influenced by too many unpredictable variables.
2Measurement precision
If manufacturers postpone process operation to wait for lower input prices, then they can reduce manufacturing costs, but production productivity decreases due to delayed output
Solution Approach 1:
The patent applies partial action by using the predicted rank to determine a threshold (cutoff) for when to operate the process. Rather than always operating or never operating, the system operates partially - only when the predicted rank indicates favorable pricing conditions - thus balancing cost optimization with maintained productivity through selective operation.
Solution Approach 2:
The patent performs preliminary action by predicting the rank of current prices before making scheduling decisions. This advance prediction allows manufacturers to proactively plan process operation timing based on expected price rankings, enabling them to schedule operations during low-price periods without severely impacting overall productivity.
3Measurement precision
If manufacturers frequently adjust process operation to capture low price periods, then they can reduce input costs, but process stability deteriorates due to frequent start-ups and shut-downs
Solution Approach 1:
The patent uses partial action by implementing a cutoff threshold based on predicted rank. Instead of responding to every price fluctuation, the system only triggers process operation when the predicted rank falls below the cutoff, filtering out minor variations and reducing frequent start-stop cycles while still capturing significant cost-saving opportunities.
Solution Approach 2:
The patent applies beforehand cushioning by using the predicted rank to anticipate favorable pricing conditions before they occur. This allows smooth, planned transitions into and out of process operation rather than reactive, frequent adjustments, cushioning the system against instability while still capturing cost advantages.
Data Source
AI summary
Disclosed are methods for scheduling when to run a process to take advantage of time-varying input prices by predicting the “rank” of the current actual input price among future input prices. The rank is that portion of a future time period during which the input price will be less than the current actual input price. An “operational cutoff” is set based on the portion of time during which a process should run in order to produce a target output amount. Periodically, the actual input price is determined, and its rank is predicted. The predicted rank is then compared with the operational cutoff. If the predicted rank is below the cutoff, then it makes economic sense to run the process. Otherwise, it would be cheaper not to run the process at present but to wait a while in expectation that the input price will drop.