{"id":"83a29089-e6a5-4ee4-bbdb-50a6098e127e","arxiv_id":"1908.00710","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"A two-stage point-estimate and bee-colony heuristic solves probabilistic AC transmission expansion planning with N-1 contingencies up to 99 percent faster than a single-stage approach on tested systems.","lead":"The paper proposes a two-stage heuristic method that uses point estimate methods and a modified bee colony algorithm to speed up probabilistic AC transmission expansion planning under wind, load, and outage uncertainties. It reports over 90 percent reductions in computation time on two standard test systems, while keeping solutions close to those from an existing method.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Bernoulli line outages are routed through the continuous 2m+1 point-estimate scheme without a defined mapping from fractional line states to the AC power-flow model, so the expected-penalty objective driving the search may be undefined.","rationale":"The central claim has two parts: near-identical solution quality and extremely low computational burden. The computational savings are plausibly demonstrated by the reported times, but the near-identical quality rests on a single small-system AC comparison (Table II) plus a DC comparison to an external reference. The weakest load-bearing point is the treatment of Bernoulli line outages inside the continuous 2m+1 PEM: the paper gives no mapping from fractional line states to the AC power-flow model, so the objective E(F_pen) in Eq. (15) is underspecified. This is not merely a matter of consensus or external validity; it is an internal gap in the definition of the problem being solved. The concern can be settled by rerunning the Garver case with an exact Bernoulli treatment and comparing the resulting plan and cost. Because the paper provides no code, no formal verification, and no explicit statement of the required mapping, the correct editorial posture remains CONDITIONAL rather than ACCEPT. The reader's weakest assumption identifies the same issue, and the conditional verdict already captures the resulting correctness risk.","tokens_in":14414,"tokens_out":11587,"duration_ms":115622,"concrete_test":"Re-run the Garver 6-bus PrACTNEP case with the line-outage uncertainties handled by exact two-point evaluation (H=0 and H=1, weighted by rho_FOR and 1-rho_FOR) for the marginal penalty, or by Monte Carlo over line states, instead of the paper's 2m+1 PEM. Hold all other settings identical. If the optimal plan or expected cost in Table II changes materially (any line decision differs, or cost shifts by more than a small tolerance), the Bernoulli-through-PEM treatment is load-bearing and the central claim is not established. If the plan is unchanged, the concern is benign for this system.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II.C.3 models each line outage as a Bernoulli random variable H in {0,1} with P(H=1)=1-rho_FOR. Section III then applies the 2m+1 point-estimate method to all uncertain variables, and Section IV.B.3 states that deterministic ACPF is solved for any location determined by the 2m+1 point-estimate method. For a Bernoulli variable, the moment-matched PEM locations are mu +/- xi*sigma and mu, which are not generally 0 or 1; for rho_FOR=0.99 the two non-central locations are approximately 0.00015 and 1.000, and the central point carries a negative weight. The paper never specifies how to run the AC power flow for a line with fractional availability: whether to scale line admittance, line rating, both, or something else. The network constraints (2)-(8) and contingency constraints (5)-(6) are defined for integer numbers of in-service lines, not for fractional states. Consequently E(F_pen) in Eq. (15) is not well-defined. Since this expected-penalty objective drives MABC's search, the reported plans and the 'almost identical' claim are solutions to an underspecified optimization problem. If the rigorous method in Table II also uses the same PEM, the comparison does not validate the solution against the true Bernoulli model.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a two-stage heuristic methodology for probabilistic AC transmission network expansion planning (PrACTNEP) under wind, load, and line-outage uncertainties. Stage 1 solves a deterministic AC TNEP to provide a warm start and to derive heuristic bounds; Stage 2 reduces the number of probabilistic AC power-flow evaluations by filtering candidate line plans through corridor-count bounds (90-130% of the crisp value), line-count bounds (70-200%), an upper cost limit v_ulim = 2*v_cr, and early truncation when a contingency or point-estimate evaluation is infeasible. The optimizer is a modified artificial bee colony (MABC) algorithm, and the expected penalty is computed with the 2m+1 point-estimate method (PEM). Results are reported for Garver 6-bus and IEEE 24-bus test systems, including a comparison with a single-stage rigorous MABC on the 6-bus AC case (same plan, 99.18% time reduction) and with a published DC TNEP method on both systems (over 90% time reduction).","tokens_in":14742,"tokens_out":6217,"duration_ms":56877,"significance":"If the methodology is valid, it offers a practically useful, low-computational-cost way to obtain first-cut probabilistic AC expansion plans, a problem that is otherwise computationally prohibitive. The paper is transparent about its heuristic nature, reports reproducible test-system data, and addresses a practically relevant combination of probabilistic wind, load, and N-1 line-outage modeling. The computational gains reported are substantial. However, the central claims rest on a single rigorous validation on a 6-bus system, and the treatment of discrete line outages through the continuous PEM raises a correctness question that needs resolution before the results can be considered trustworthy.","major_comments":[{"comment":"The Bernoulli line-outage model is passed through the continuous 2m+1 point-estimate method without specifying how a fractional line state is mapped to the AC power-flow model. For a Bernoulli variable with FOR=0.99, the PEM locations are approximately 0.00015 and 1.0, with the central point at the mean 0.01; none of these values is a physically meaningful line state for the constraints (2)-(8), which are defined for integer numbers of in-service lines. No rule is given for scaling line admittance, line rating, or both for a fractional state, so E(F_pen) in Eq. (15) is not well-defined. Because this expected-penalty objective drives the MABC search, the reported plans and the 'almost identical' claim rest on an underspecified optimization problem. The authors must either treat Bernoulli outages exactly (e.g., by a two-point evaluation with probabilities rho_FOR and 1-rho_FOR) or specify and justify a consistent fractional-state power-flow model.","section":"Section II.C.3, Section III, Section IV.B.3"},{"comment":"The corridor bound (90-130%), the line-count bound (70-200%), and the upper cost limit v_ulim = 2*v_cr are empirical filters 'obtained through numerous trials' on the test systems. These filters directly restrict which candidate plans are evaluated by the probabilistic power flow. The paper's observation that the final solution cost is within the v_ulim bound is therefore partly a consequence of the filter, not an independent finding. No sensitivity analysis is provided, and no argument is given that the bounds are wide enough to contain the true optimum for the 24-bus system. As the method is explicitly sub-optimal, the authors should provide a sensitivity study or a formal justification (e.g., comparison with a relaxed or exhaustive reference solution) to show that the tuned bounds do not materially exclude high-quality plans.","section":"Section IV.B.1, Section IV.B.2"},{"comment":"The central claim that the proposed method 'is able to obtain almost identical solutions' is validated on only one small system: Table II shows that for the Garver 6-bus AC case the proposed and rigorous methods yield the same plan and cost, with a 99.18% time reduction. For the IEEE 24-bus AC case, no rigorous comparison is provided (Table IV), and the reported >90% reductions for the DC case are against a published method on possibly different hardware and implementation. The paper also reports only the best result of 50 trials, with no information about the distribution (median, min/max, standard deviation) or any optimality gap. To support the generalization claim, the authors should report trial statistics and provide at least a lower bound or a limited rigorous reference solution for the 24-bus AC case.","section":"Section V.A.2, Section V.B.2"}],"minor_comments":[{"comment":"The use of the index k both for the contingency state and for a corridor (in 'l ≠ k' and 'for l = k, k ≠ 0') creates a notation clash that makes the contingency constraints hard to parse; please distinguish the contingency index from the corridor index.","section":"Section II.A, Eqs. (3)-(6)"},{"comment":"The coefficient of variation is usually defined as sigma_x/mu_x, but Eq. (22) defines nu_x = mu_x/sigma_x; this appears to be a typo.","section":"Section III, Eq. (22)"},{"comment":"The statement that 'the final PrACTNEP cost remains within twice of the crisp ACTNEP cost' is presented as a fact; please explicitly label it as an empirical observation from the authors' trials and describe the systems over which it was tested.","section":"Section IV.B.2"},{"comment":"The 15.38% cost reduction over [5] should be qualified with the scenario assumptions (wind level, load variation, no-load-shedding condition) to ensure the comparison is commensurable.","section":"Table I"},{"comment":"The MABC parameters (csN=20, psi=2, lim=6, iter=30, wg=1.5) are taken from previous papers by the same authors; a brief statement on their sensitivity for the present problem would improve reproducibility.","section":"Section V, MABC settings"}],"recommendation":"major_revision","confidential_remarks":"The Bernoulli-PEM issue is the most serious concern: it is not merely a presentation gap but potentially renders the objective in Eq. (15) undefined. I would ask the editor to require the authors to clarify or correct this before acceptance. The tuned heuristic bounds are also central to the reported speedups, and the validation is much weaker than the abstract's 'almost identical' claim suggests."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This one is worth a look if you care about probabilistic AC transmission expansion planning. The core idea is simple and sensible: solve a deterministic ACTNEP first, then use that solution to seed a metaheuristic and to define search bounds (corridors, total lines, cost cap) so that the expensive probabilistic power flow is only run for 'worthy' candidate plans. That package is new as a combination, and the reported speedups are dramatic—99% on the 6-bus case, 90% on the DC 24-bus case. The authors are up front that they are after good sub-optimal solutions, not exact ones.\n\nWhat the paper does well: it attacks a real computational wall, the two-stage structure is a legitimate way to cut the search space, and they test on two standard systems with N-1 contingencies plus wind and load uncertainty. The deterministic-first-stage is a reasonable warm start, and the dynamic cost cap does adapt during the search.\n\nThe soft spots are real. The most serious is in the probabilistic modelling. Section II.C.3 models each line outage as a Bernoulli random variable. Then Section III says the 2m+1 point estimate method is applied to all uncertain variables. For a Bernoulli, the PEM locations are mu +/- xi*sigma and mu, which are not 0 or 1 in general. The paper never says how to run the AC power flow for a line with fractional availability—whether to scale admittance, rating, or both. The constraints in (2)-(8) are defined for integer numbers of lines. So E(F_pen) in eq. (15) is not actually well-defined on the face of the paper. If the rigorous comparison method in Table II uses the same PEM, then the 'almost identical' result is comparing two versions of the same undefined scheme, not validating against the true Bernoulli model. That's a load-bearing gap, not a rounding error.\n\nSecond, the central AC claim rests on one small system. The Garver 6-bus comparison against a rigorous method is the only evidence for 'almost identical solutions.' On IEEE 24-bus, only the proposed method is run; there is no rigorous baseline. The DCTNEP comparison to [5] is good, but that is the DC problem. The bounds (90-130% corridors, 70-200% lines, 2*v_cr cost cap) are fitted parameters, and the cost cap directly truncates the search, so the observation that final costs stay within it is partly enforced. The claim that bounds are from 'numerous trials' is plausible but not a substitute for a sensitivity study.\n\nWho is this for? A reader working on heuristic TNEP or on probabilistic power flow will find the two-stage idea and the bound strategies worth discussing. It is not a paper whose numerical claims you can rely on as-is. For peer review: I would send it to referees. The topic matters, the speedup is real if the objective can be defined, and the Bernoulli/PEM issue is fixable (e.g., treat line outages as separate discrete states or use a mixed-integer formulation for those states). It needs major revision, not desk rejection.\n\nBottom line: the engineering instinct is right, the technical execution has a hole in the middle. Worth engaging with, but do not cite the numerical results until the mapping is clarified.","headline":"Heuristic speedup for probabilistic AC TNEP with a sensible two-stage structure, but the Bernoulli-to-PEM mapping is undefined and the 'almost identical' claim rests on a single small system.","tokens_in":15190,"tokens_out":3778,"would_cite":false,"duration_ms":35971,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A two-stage point-estimate and bee-colony search solves probabilistic AC transmission planning with near-identical plans and over 99 percent less computation.","keywords":["AC transmission network expansion planning","probabilistic planning","point estimate method","modified artificial bee colony","wind power uncertainty","N-1 contingency","power flow"],"falsifier":"On the 6-bus test case, replace the shortcut samples for line outages with exhaustive enumeration of all line in/out combinations (keeping the paper's wind and load sampling), and compare the expected penalty and final plan. If the plan or the ranking of candidate plans changes, the shortcut is computing a different expectation than the stated probabilistic problem.","tokens_in":14240,"feed_emoji":"⚡","tokens_out":10519,"duration_ms":101856,"temperature":0.7,"pith_summary":"The paper targets transmission network expansion planning—deciding which new power lines to build—under wind, load, and line-outage uncertainty, using the full alternating-current (AC) power-flow model. It tries to establish that a two-stage solution method can solve this probabilistic planning problem with near-identical results to a rigorous single-stage search while cutting computation by more than 99 percent. Stage one solves the deterministic version of the problem to get a starting plan and a cost lower bound; stage two uses a $2m+1$ point estimate method inside a modified artificial bee colony (MABC) algorithm to estimate expected constraint penalties, and prunes away candidate plans that are too costly, too large, or already infeasible in early checks. On a standard 6-bus test system the method finds the same expansion plan as the rigorous method in 104.93 seconds instead of 3.57 hours, a 99.18 percent time reduction. If the claim holds, medium-sized AC transmission planning under realistic uncertainties becomes tractable on a desktop computer rather than effectively unsolvable.","feed_headline":"Grid expansion planning under uncertainty cut by 99 percent","feed_subtitle":"A two-stage AC power-flow planner matches the rigorous 6-bus plan in 105 seconds instead of 3.6 hours.","key_machinery":"The central mechanism is the pairing of a modified artificial bee colony (MABC) metaheuristic with the $2m+1$ point estimate method (PEM) for power-flow uncertainty, guarded by pruning rules. PEM approximates the expected value of a function of random inputs by solving deterministic AC power flow at $2m+1$ selected sample points—two per uncertain variable plus one joint-mean point—instead of running thousands of random scenarios. MABC proposes candidate line-addition plans, and each candidate is scored by line investment cost plus the estimated expected penalty; the pruning rules restrict which candidates receive a full probabilistic power-flow evaluation. Because each inner evaluation is a full AC power flow, the decisive factor for the reported speedup is the reduction in how many of these evaluations the search performs.","core_discovery":"The central claim is that the computational bottleneck of probabilistic AC transmission network expansion planning—the repeated power-flow solves needed to evaluate each candidate expansion plan—can be bypassed without losing plan quality. The proposed recipe solves the crisp/deterministic expansion problem first, then solves the probabilistic problem with a modified artificial bee colony search whose fitness is investment cost plus expected constraint penalty, where the expectation is estimated by the $2m+1$ point estimate method. Four pruning rules decide when a full probabilistic power flow is actually run: the number of new corridors and total new lines must sit in bands set from the deterministic solution, the cost must lie between the deterministic cost and twice that value, and evaluation stops at the first infeasible sample or the first failed single-line outage contingency. In the 6-bus test case the method returns the same plan and same total cost as the rigorous single-stage method while reducing solution time from 3.57 hours to 104.93 seconds; in the 24-bus case it reports plans and timings without a rigorous comparison, and the paper explicitly frames the results as good-quality sub-optimal solutions.","pith_inferences":["I would not carry the 99 percent time-reduction claim beyond the 6-bus case: the paper reports no rigorous comparison for the 24-bus system, so its accuracy there is untested.","The pruning bounds (e.g., final cost at most twice the crisp cost, corridor counts in a 90–130 percent band) are tuned by experiment on these systems; on other networks they could exclude the true optimum or fail to prune, so the speedup is not guaranteed to transfer without retuning.","The same two-stage pattern—solve an inexpensive deterministic surrogate, then run a probabilistic refinement with early termination—could be applied to other expensive stochastic mixed-integer nonlinear planning problems, not only transmission expansion."],"forward_implications":["Single-line-outage (N-1) security checks can be included in probabilistic AC expansion planning for medium test systems in minutes rather than hours.","Planners can retain full AC power-flow fidelity while modeling wind, load, and line-availability uncertainties, reducing the need to fall back on simplified linearized models.","The reported time savings make repeated metaheuristic trials affordable, which matters because the reported plans are best-of-50 results and tuning depends on running many trials.","The deterministic solution serves as both a warm start and a cost floor, so the method guarantees that probabilistic planning cost never falls below the crisp-plan cost."],"supporting_citations":[{"why":"Supplies the point-estimate scheme that estimates expected penalties from a few deterministic evaluations.","marker":"[23]"},{"why":"Extends the scheme to probabilistic power flow, providing the $2m+1$ evaluation framework used in the search.","marker":"[24]"},{"why":"Supplies the modified artificial bee colony algorithm and the parameter settings used for candidate plan generation.","marker":"[25]"},{"why":"Provides the probabilistic DC TNEP comparison results and the wind/load probabilistic models used in the test cases.","marker":"[5]"},{"why":"Provides the AC transmission expansion formulation and test-system data on which the probabilistic AC planning is built.","marker":"[28]"},{"why":"Supplies the 24-bus test-system data used for the larger planning case.","marker":"[29]"},{"why":"Provides the security-constrained AC TNEP approach that the paper extends to probabilistic line-outage and N-1 conditions.","marker":"[26]"}],"fun_headline_variants":["Two-stage planner cuts expansion grid delays by 99%","AC grid expansion solved 100x faster with uncertainties","Probabilistic grid planning: 105 seconds vs 3.6 hours","Sub-optimal but fast: new TNEP method matches plan in 2 minutes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that a line's random in/out state can be fed through the shortcut statistical method even when the sampled states fall between 'line present' and 'line absent'—an assumption the paper does not justify.","fun_headline_variants_meta":{"raw":{"variants":["Two-stage planner cuts expansion grid delays by 99%","AC grid expansion solved 100x faster with uncertainties","Probabilistic grid planning: 105 seconds vs 3.6 hours","Sub-optimal but fast: new TNEP method matches plan in 2 minutes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000221,"raw_usage":{"total_tokens":1473,"prompt_tokens":990,"completion_tokens":483,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":606,"completion_tokens_details":{"reasoning_tokens":408}},"tokens_in":606,"tokens_out":483,"duration_ms":4874,"temperature":1.0,"reasoning_tokens":408,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:36:16.198723+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On the 6-bus test case, replace the shortcut samples for line outages with exhaustive enumeration of all line in/out combinations (keeping the paper's wind and load sampling), and compare the expected penalty and final plan. If the plan or the ranking of candidate plans changes, the shortcut is computing a different expectation than the stated probabilistic problem.","supporting_citations":[{"cited_title":"An efficient point estimate method for probabilistic analysis,","cited_arxiv_id":null,"evidence_quote":"Supplies the point-estimate scheme that estimates expected penalties from a few deterministic evaluations."},{"cited_title":"Point Estimate Schemes to Solve the Probabilistic Power Flow,","cited_arxiv_id":null,"evidence_quote":"Extends the scheme to probabilistic power flow, providing the $2m+1$ evaluation framework used in the search."},{"cited_title":"Transmission network expansion planning using a modified artificial bee colony algorithm,","cited_arxiv_id":null,"evidence_quote":"Supplies the modified artificial bee colony algorithm and the parameter settings used for candidate plan generation."},{"cited_title":"Transmission Expansion Planning of Systems With Increasing Wind Power Integration,","cited_arxiv_id":null,"evidence_quote":"Provides the probabilistic DC TNEP comparison results and the wind/load probabilistic models used in the test cases."},{"cited_title":"Power system transmission network expansion planning using AC model ,","cited_arxiv_id":null,"evidence_quote":"Provides the AC transmission expansion formulation and test-system data on which the probabilistic AC planning is built."},{"cited_title":"Optimization of AC Transmission System Planning,","cited_arxiv_id":null,"evidence_quote":"Supplies the 24-bus test-system data used for the larger planning case."},{"cited_title":"Security Constrained AC Transmission Network Expansion Planning ,","cited_arxiv_id":null,"evidence_quote":"Provides the security-constrained AC TNEP approach that the paper extends to probabilistic line-outage and N-1 conditions."}],"review_version":1}