Dynamic Optimization in MATLAB and Python

[@sakshisinhal8152]

Why do we get the value of state variable and total variable in the end as 200 and 198?

[@apm]

I get values less than 1. Here is the source code to check your work: apmonitor.com/do/index.php/Main/DynamicOptimizationBenchmarks

[@40NoNameFound-100-years-ago]

Hello, thanks for such a great video. I have a question regarding the CSV file. Why do we have to create that file? I mean in Matlab If we use fmincon and ode45, we will solve the optimization problem and use the 'stairs' function to generate the control signal

[@apm]

Using fmincon and an ode integrator such as ode45 is a shooting (sequential) method. This is a simultaneous optimization method. Try Gekko for the latest version of the package: apmonitor.com/wiki/index.php/Main/GekkoPythonOptimization

[@mellorf]

Great contribution! How can I install the optimization server in my local computer?

[@apm]

It is available for Windows or Linux: apmonitor.com/wiki/index.php/Main/APMonitorServer

[@apm]

Alternatively, use Python Gekko and set remote=False

[@itzipitzitzi]

When you speak about dynamic optimization, you are referring to Model predictive control as you talk in other videos? How is this related to optimal control?

[@apm]

Dynamic optimization is more general than MPC. MPC typically involves a receding horizon and is typically with linear models in industrial practice. I use the term Dynamic Optimization to differentiate that we are solving an optimization problem with a dynamic model but there are no restrictions for linearity, MV or CV tuning (both are allowed), open-loop stable systems, or continuous processes. Dynamic optimization can also be used to describe the data reconciliation problem where model predictions are aligned to measured values. The Dynamic Optimization course is hosted here: apmonitor.com/do We just started a new session (Jan-Apr 2018) if you are interested in joining.

[@al_yafeai6125]

Can we solve it in MATLAB only? like use ode45 and fmincon

[@apm]

Yes, you can use shooting methods but they are much less efficient and often don't find a solution to large or complex problems. More information is at apmonitor.com/do on how to solve these problems with Python Gekko, APM MATLAB, and some simple examples with fmincon and solvers such as ode45. The problem with shooting methods (fmincon + ode45) is that you end up converging the system equations multiple times (wasted effort) instead of only once at the solution.

[@akshaysunil2852]

How to do this for a reservoir optimization model

[@apm]

Here is a review of reservoir optimization techniques: Udy, J., Hansen, B., Maddux, S., Peterson, D., Heilner, S., Stevens, K., Lignell, D., Hedengren, J.D., Review of Field Development Optimization of Waterflooding, EOR, and Well Placement Focusing on History Matching and Optimization Algorithms, Processes, 5(3), 34, 2017, doi:10.3390/pr5030034. www.mdpi.com/2227-9717/5/3/34

[@virajgir7613]

Can you suggest how to do with PSO in matlab and python?

[@apm]

I don't have any content on that but here is something on simulated annealing that may help: apmonitor.com/me575/index.php/Main/SimAnnealHW

[@amardeepmishra5547]

can I do the same in linux?

[@rrc]

yes, can run on Linux, MacOS, Windows, etc. It will run any platform that can run Python, Matlab, or Julia.

[@amardeepmishra5547]

+John Hedengren thanks a lot John, it solves a lot of problem!