A Lagrange Relaxation Based Decomposition Algorithm for Large-Scale Offshore Oil Production Planning Optimization

Round 1

Reviewer 1 Report

The presented manuscript entitled “A Lagrange relaxation-based decomposition algorithm for large-scale offshore oil production planning optimization” proposes the well-known Lagrange decomposition to solve a complex large-scale offshore oil production optimization problem. The authors have successfully proved not only a reduction in the computational cost, but also and more importantly the improvements in terms of duality gap. As such, the reviewer congratulates the authors for the presented work. Nevertheless, the authors are encouraged to address the following issues:

1. The first paragraph is important, but the “framing” of this work subject needs to be rewritten with added literature. Please, refer to the expected role of future oil and gas production (it is a backbone role? A reserve role?), including the future rate of exploration of oil and gas fields, in a framework where China has made international compromises in terms of reducing emissions.
2. The authors should consider updating the state-of-the-art (references [2]-[6],) despite the understandable difficulty of finding works in this field. As such, consider using some reports or operation experiences from real oil and gas operators (which must deal with this optimization problem daily). In this sense, why the authors neglect what is the common robust approach used in terms of production optimization?
3. Moreover, you mention the reducing oil prices. Is that assessment based on a yearly basis during this Covid-19 pandemic or is an expected trend? If so, consubstantiate with added literature.
4. The main motivation of this work is to address the subject of production costs in a large-scale environment. But the authors fail to contextualize this problem for energy companies. Are these the main exploration cost, and what % share of gains is expected for other case studies?
5. The authors mention “To the best of our knowledge, there is little report about LR in oil and gas production optimization, especially in the offshore oil and gas domain.” The reviewer encourages the authors to follow this phrase, with brief statements, preferably in bullet points, highlighting the main contributions of this work. (From the conceptualization, key implementation details of the LR, case study relevance and optimization horizon, and finish with the % gains in terms of dual problem gap and reduced computational cost).
6. The flow guarantee mechanism cannot be compromised under the considered restrictions.
7. Please improve and review the entire text. An English proof is recommended in phrases like: “The mathematical and constraints of Z1 to Z6 are given by equation (2)- (48).” Or “Although the final result of plan optimization obtained by this method is not necessarily the global optimal solution, it can obtain the expected result in a shorter time scale, which has reference value for practical industrial application.” Or Table 6 caption.
8. Equation (49) is just a repeat of equation (1). Please, for purposes of readability avoid repeating equation, in an already extensive formulation.
9. Why was the comparison made against the ALPHAECP? How is the benchmarking representative?
10. When presenting the results in Table 6, what is the significance of the presented % relative difference?
11. The authors can save space and reduce the length of the paper by combining horizontally the Figures regarding the results. For example, Figures 4 to 6, do not require higher level of resolution and can be 4 a) to 4c) or at least to 4a) to 4b), by outlining the curves in Figures 4 and 6 under the same plot but with different colors for the total and delivered oil production.
12. In the Conclusions, the authors need to better highlight the differences (benefits and downsizes) of the LR in solving the MINLP problem versus traditional LP or MILP? Moreover, please elaborate on the adequacy of bundle and interior point methods.

Author Response

We would like to thank the reviewers for the insightful comments and the editor’s further consideration of our manuscript. We have addressed all concerns and revised the manuscript.  Detailed responses are listed below.

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Recommendation: The manuscript may be accepted for publication after the following comments are addressed

The manuscript applies a Langrangian relaxation approach to decompose a large MINLP problem for large-scale offshore oil production planning optimization problem to reduce solution time so that it can be solved within a specified duality gaps in reasonable time for practical industrial application.

Comments:
1) In the abstract, can you rephrase what is meant by "efficient and mature" algorithms?
2) Can you rephrase and clarify what is meant by "actual industrial cases requirement" lines 81-83
3) On line "88" please add the abbreviation LR as well to the Langrangian Relaxation so that it is clear when that abbreviation is used later in the text
4) Please cite articles which have used LR decomposition in MINLPs for oil and gas production optimization even if they donot include the offshore domain.
An example https://pubs.acs.org/doi/abs/10.1021/ie000755e Heever et al (2001)
5) What is ESP? Line 133
6) Please replace the word "dual gap" with "duality gap"
7) For section 2.2, can you add a citation to the original article on which Langrangian Relaxation methodology used in this manuscript is based on? Also mention novelties in the current approach, if any
8) On line 220, the equation referred to should be 48
9) Please cite the work of Fisher (1985) for equation 61 on line 293, otherwise this is quite unclear and arbitrary to a reader
10) Please summarize all the novelties of this current work in the introduction

Author Response

We would like to thank the reviewers for the insightful comments and the editor’s further consideration of our manuscript. We have addressed all concerns and revised the manuscript.  Detailed responses are listed below.

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Manuscript: A Lagrange Relaxation based decomposition algorithm for large scale offshore oil production planning optimization

The paper proposes a new Lagrange relaxation-based decomposition algorithm for the integrated offshore oil production planning optimization that minimizes the total costs. In this approach, a Lagrange multiplier is introduced to relax coupling constraints of multi-batch units, resulting in moderate scale subproblems that can be solved more efficiently. Computational results show that the proposed methodology can reduce the solution of solving the original optimization problem by upward of 43%.

 

Major Comments

1. One of the main determinants of an efficient Lagrange relaxation scheme is the numerical value of Lagrange multiplier \lambda at each iteration of the algorithm. How do you determine \lambda needed to initialize the Lagrange algorithm in Figure 3?
2. The original problem is a mixed-integer nonlinear programming problem (MINLP). It is well known that LR has issues in closing the dual gap. Does your LR have a way to close the dual gap whenever the upper and lower bounds are not changing with increasing iteration?
3. What is the problem class of the subproblems solved in the LR? Is it an NLP? If so, is there a reason why the authors used ALPHAECP to solve the subproblems? I wonder if more popular NLP solvers like IPOPT, KNITRO, or CONOPT are more performant to solve these subproblems?
4. Have the authors considered using a different MINLP solver like BARON/ANTIGONE for the full problem? And how does replacing the solvers for the subproblems affect the efficiency of the LR algorithm?
5. The evolution of the dual bounds (lower and upper bounds) as the LR iteration progresses is an important piece of information for LR. I would suggest the authors include a figure for each case study showing how the dual bounds evolve with increasing iteration.

 Minor comments

1. A lot of grammatical errors are prevalent in the manuscript. Authors should please review and fix those, e.g.

1. Line 35; replace with "last 10 years".
2. Line 81; Replace “Although these valuable progress” with “Despite the valuable progress”
3. Line 90,94, etc. Only last name is required for author citation.
4. Etc.

2. The introduction is missing a lot of key reference publications in the upstream production optimization space, particularly papers from Eduardo Campanagara’s lab. I‘d suggest authors look at these publications.
3. Can you prepare a table that lists all the variables (binary/continuous), parameters, and sets defined in Eq 2 to 48?
4. Can the authors consider including the number of nonlinear terms in Tables 5 and 6? That’s important in further giving readers an idea of the complexity of the model.

Author Response

We would like to thank the reviewers for the insightful comments and the editor’s further consideration of our manuscript. We have addressed all concerns and revised the manuscript.  Detailed responses are listed below.

Please see the attachment.

Author Response File: Author Response.docx