How to automatically optimize production schedule against electricity costs

New concepts allowing industrial demand-side management (iDSM) tested live in steel production

分享本页

Scheduling the steel-making process

Scheduling production in a steel melt shop is not easy, partly due to the extreme processing and material temperatures. For example, each production delay leads to cooling and later a reheating need. Therefore, there is a strong demand in the industry for automatic production schedule optimization. Attention must also be paid to other implementation aspects such as enabling different melt shop configurations and product portfolios; appropriate graphical user interfaces (GUIs); integration with other IT systems, eg, enterprise resource planning (ERP), energy management systems and process control systems. Without all these aspects, even the most sophisticated scheduling optimization model can never be deployed in a real production environment.

In contrast to energy efficiency strategies, which aim to produce the same using less energy, demand-side response focuses on profitable shifting of the load in time

Collaboration between energy management and production planning

ABB developed new concepts allowing industrial demand-side management (iDSM) by automatic optimization of the production schedule against the electricity costs. The first step toward the iDSM solution was to investigate the use of monolithic models for the integration schemes shown below in the middle.

The last graph depicts the idea in which an additional time grid is added to the original scheduling formulation in order to check the electricity consumption in each of the slots formed. The electricity provider or the electricity market defines the energy price for each of these time slots (15 to 60 min). Theoretically, this holistic-model-based optimization may lead to a so called global optimum, ie, the best possible solution with respect to both the production and electricity costs. However, holistic models are very often complicated or impossible to solve within a reasonable time, so some refinement is required.

The scientific challenge arises in simultaneously optimizing the production schedule and the electricity purchase strategy.

Here, at least a partly prespecified production schedule is assumed

Here, the production schedule and the electricity purchase strategy are simultaneously or iteratively optimized

The grid delineating the electricity price time slots

Refining the models

In production processes that require multiple production steps, such as batch-oriented steel manufacturing, typically not all equipment is continuously occupied. This allows flexibility to adapt production according to energy management needs. Multistage production processes also usually have buffers to store raw material, and intermediate and final products – for limited periods of time. In the melt shop process, for example, intermediate products are very hot so inadequate coordination of subsequent production stages causes energy loss through cooling. Another constraint is that large electricity consumers commonly have to commit their forecasted load pattern upfront and suffer financial penalties for deviations from it.

In the work done by ABB, the continuous-time (exact) melt-shop scheduling model has been refined to take into account both the electricity price as well as deviations from a committed load curve. The benefit of this approach is that the energy considerations can be included in the original scheduling model by adding new decision variables to map the electricity consumption for each grid-defined time slot. This results in feasible solutions with clear energy savings potential.

However, this basic approach is not efficient for more complex instances. Therefore, various alternative approaches have also been looked at including other modeling philosophies – eg, resource-task network – as well as decomposition algorithms.

Mixed-integer linear programming (MILP) techniques have improved significantly and can now solve problems that are several magnitudes larger than those overcome a few decades ago.

Energy pricing and usage scenarios

Based on realistic data, a hypothetical case study has been carried out to investigate how strongly three different energy tariff scenarios might influence the energy bill for a typical 24-hour scheduling problem. The scenarios are each assumed to buy a fixed amount of electricity at a known rate using a base load contract. The total energy bill can be reduced by reselling any surplus electricity. The committed load aspect is also taken into consideration.

The first scenario represents a day with “normal” electricity prices in the volatile day-ahead market. When the scheduling driven by energy price is employed, the net electricity cost is around $ 110,000.

The second scenario uses weather-driven prices, which result in an additional cost of $ 27,000. The third scenario ignores energy price considerations, ie, only production throughput is optimized – resulting in a cost double that of the second scenario. This demonstrates how much the plant could potentially save by collaborative scheduling and energy optimization on a day with extreme prices.

In this case study, energy-driven scheduling contributes to significant reductions of the electricity bill. However, comparison of the schedules of scenarios two and three clearly shows that the energy-driven schedule tries to avoid extreme prices of the peak hours (marked in red and orange in the graph below at the expense of extending the overall make-span (the total time needed for production).

Some of the production operations are delayed – perhaps incurring reheating costs. In the study, the cost of thermal losses has not been included in the savings calculation. However, with realistic cooling models it is certainly possible to account for potential costs that can be associated with production delays.

Scheduling solution

A new ABB production scheduling system based on MILP technology was deployed in a very complex melt shop belonging to Acciai Speciali Terni SpA, a member of ThyssenKrupp and one of the world’s leading producers of stainless steel flat products. With the new system in place, the production scheduler is able to automatically and optimally create a new schedule, or manually update an existing
one, for up to seven days of production within just a few minutes. The system is flexible enough to support different melt shop configurations, as well as to include all other information necessary – such as processing, transportation, setup and cleanup times – to generate a feasible production schedule. It also takes into account maintenance plans, the current status of the melt shop and availability of different equipment, due dates, penalties for lateness and violation of holdup times between stages in the process, etc. In addition, the steel plant created a Web-based GUI that allows the user to flexibly select what to optimize and schedule.

Total electricity cost for various energy prices and optimization strategies

Comparison of schedules driven by energy cost and makespan cost

Web-based GUI of the new production scheduling system. Prices can change by the hour

The production scheduler is able to automatically and optimally create a new schedule, or manually update an existing one, for up to seven days of production within just a few minutes

The list on the left in the second graph shows production steps or units, such as the electric arc furnace (FEA) and ladle furnaces (ASEA). For each unit, there are two rows. The “monitoraggio,” row is related to monitoring and shows the current status of the unit and what really has happened there. The “programma” row shows what has been planned/scheduled in each unit. Thus, the GUI also enables other departments to initiate appropriate actions in order to minimize potential losses and production delays – for example, reschedule or postpone production slightly due to a high electricity price. The new scheduling system is not only linked to other internal IT systems such as ERP and process control, but also to the external day-ahead electricity market, in order to dynamically cater for volatile electricity prices.

Furthermore, in cooperation with ABB, the steel plant has also integrated into the new scheduling system an advanced solution for optimization of the production schedule, taking into account electricity prices as well. This advanced solution enables the plant to optimize its electricity costs and make-span and thus play a more active role in demand-side response programs and support grid reliability and safety.

The continuous-time melt-shop scheduling model has been refined to take into account both the electricity price as well as deviations from a committed load curve.

Benefits

It has been shown that the implementation has improved the coordination between different production stages in the melt shop and thus decreased the holdup times between these, reducing energy consumption. The system has also been recognized as a very useful tool for running various simulations and what-if analyses. The benefits are estimated to be in the range of 2 to 5 percent – a considerable saving given the large energy budgets involved.

Flexibility is key

The complexity of production scheduling is increasing in industries outside of steelmaking too, mainly due to smaller and more customized orders. Production plants now have to be agile and flexible to respond to short-term changes. These industries also face the complexity arising from variable, but potentially more affordable, electricity pricing on an hourly basis in the day-ahead market. Consequently, combined energy and production planning processes must always be well integrated with real-time data.

Having a full offering of process and grid automation, ABB has the tools to realize a proper matching of supply and demand using internal buffers in the process and production load shifting for a wide range of industries.