A wind farm site selection decision method

Abstract

The present application relates to a kind of wind farm site selection decision-making method, comprising the following steps: obtaining the relevant influence factor data of historical reference wind farm site selection case, and the operation measurement index data after the case is implemented, and establish three-level evaluation network structure;Based on the inference conditional probability table of variable evaluation network structure, the cost function reflecting the effectiveness of the score corresponding measurement index data is established by three-level evaluation network structure;Based on the least squares method optimization, the minimum value of cost function is obtained, and the corresponding conditional probability table is obtained;The relevant influence factor data of candidate wind farm is obtained;Based on the conditional probability table obtained, the candidate wind farm is evaluated, and the optimal wind farm site selection decision is made according to the evaluation score.Compared with prior art, the present application can avoid the adverse effects of subjective factors of expert opinions on the final selection, and innovatively introduce objective measurement index as the optimization evaluation standard, realize objective, reasonable, systematic wind farm site selection scheme decision.

Description

Technical Field

[0001] This invention relates to the field of offshore wind farm site selection decision-making, and in particular to a wind farm site selection decision-making method based on a custom cost function and sequential least squares optimization. Background Technology

[0002] With the rapid development and progress of the wind power industry, the site selection of offshore wind farms has become a key issue in the field. Factors influencing offshore wind farm site selection include production efficiency, construction costs, operating costs, and construction risks.

[0003] The technical problems existing in the traditional offshore wind farm site selection methods include: over-reliance on expert opinions brings uncertain subjective factors to the decision-making results, and the lack of objective evaluation and measurement may lead to deviations between the site selection decision results and the expected effects. Summary of the Invention

[0004] The purpose of this invention is to provide a wind farm site selection decision method based on a custom cost function and sequential least squares optimization, so as to achieve reasonable and feasible wind farm site selection under objective evaluation indicators.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A wind farm site selection decision-making method based on a custom cost function and sequential least squares optimization includes the following steps:

[0007] S1. Obtain relevant influencing factor data from historical reference wind farm site selection cases, as well as operational measurement index data after the implementation of the cases, and establish a three-level evaluation network structure based on the relevant influencing factor data for comprehensive analysis.

[0008] S2, based on the variable evaluation network structure inference conditional probability table, establishes a cost function that reflects the effectiveness of the data of the corresponding evaluation indicators through a three-level evaluation network structure;

[0009] S3, the minimum value of the cost function is obtained based on the sequential least squares method, and the corresponding conditional probability table is obtained;

[0010] S4, obtain relevant influencing factor data for candidate wind farms;

[0011] S5. Based on the conditional probability table obtained in step S3, evaluate the candidate wind farms and make the optimal wind farm site selection decision based on the evaluation scores.

[0012] In S1, the relevant influencing factor data includes four categories: production efficiency, construction cost, operating cost, and construction risk. Among them, the influencing factors of the production efficiency category include average wind speed and the sea area involved in the project; the influencing factors of the construction cost category include water depth, distance of the wind farm from the shore, and equipment transportation distance; the influencing factors of the operating cost category include salt spray concentration, the sum of the grid connection distance and 0.5 times the cable length, and the wind direction consistency coefficient; and the influencing factors of the construction risk category include annual typhoon frequency, the sum of the annual average typhoon intensity and 0.7 times the annual maximum typhoon intensity, and annual average wave height.

[0013] An operational performance indicator sequence (a1, a2, ... a3) is established based on operational performance indicator data from n wind farms. i ,…,a n ), where a i Let i represent the i-th wind farm.

[0014] The three-level evaluation network structure is as follows: the first level is the influencing factor layer, which represents the specific influencing factors under each category; the second level is the decision standard layer, which represents the category of influencing factors; and the third level is the target layer, which represents the final comprehensive evaluation.

[0015] S2 includes the following steps:

[0016] S21, perform data fuzzing on the influencing factors data of historical reference wind farm cases;

[0017] S22, based on the conditional probability table of the variable evaluation network structure, calculate the comprehensive score b of the i-th wind farm. i And establish a reference wind farm scoring sequence (b1, b2, ..., b i …,b n The evaluation network structure inference conditional probability table determines the parameters in the table based on the hierarchical structure of the evaluation network and the evaluation level of each predefined unit in the network structure. The values ​​of the parameters are variable.

[0018] S23, based on the operational measurement indicator sequence (a1, a2, ..., a n ) and wind farm scoring sequence (b1,b2,…,b n A cost function is established to measure the effectiveness of the scoring sequence.

[0019] The cost function is:

[0020]

[0021] in, is the auxiliary calculation sequence; x is the sequence of independent variables composed of the parameters of the variable evaluation network structure inference conditional probability table; n is the number of reference wind farm selection cases.

[0022] In S3, based on the sequential least squares method, the solution to the minimum value of the cost function is transformed into the following problem:

[0023]

[0024] Where x is a sequence of independent variables consisting of the parameters of the conditional probability table for the evaluation network structure inference. h i G represents an equality constraint. j This represents an inequality constraint; m is the number of equality constraints; p is the number of inequality constraints.

[0025] In S3, the iterative calculation process based on the sequential least squares method includes the following steps:

[0026] S31, construct an approximate quadratic programming problem and solve the problem to obtain the search direction;

[0027] S32, determine the search step size based on the line search method, and update the current solution in the search direction;

[0028] S33, Convergence test.

[0029] In S31, based on the principle of sequential least squares, in the current solution x l At this point, construct the following approximate quadratic programming problem:

[0030]

[0031] in

[0032]

[0033] Let λ be the Lagrange function, λ and μ be the Lagrange multipliers for equality and inequality constraints, respectively, and l be the iteration number. This represents calculating the gradient. Let f(x) denote the Laplace operator, where f(x) is the cost function;

[0034] After solving this quadratic programming problem, the corresponding search direction d after the l-th iteration is obtained. l .

[0035] In step S32, the current solution is updated based on the search step size and search direction:

[0036] x l+1 =x l +α l d l

[0037] Where, α lThe search step size obtained using the line search method:

[0038]

[0039] α is the step size.

[0040] In S33, when the search direction d l When the norm or the change in the cost function is less than a set cutoff threshold, the algorithm terminates and obtains the optimal solution f(x) for the current cost function. l ) and the optimal solution x in the corresponding conditional probability table l .

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] (1) This invention comprehensively considers multiple factors affecting the decision-making of offshore wind farm site selection. Each wind farm involves multiple relevant raw data, which improves the comprehensiveness and accuracy of the final decision.

[0043] (2) This invention introduces objective measurement index data as optimization target data and optimizes it based on the sequential least squares method, thereby improving the objectivity, rationality and scientific nature of the final decision.

[0044] (3) This invention avoids reliance on expert opinions and innovatively solves the problem of decision-making errors caused by the uncertainty of subjective factors in expert opinions during evaluation and decision-making. Attached Figure Description

[0045] Figure 1 This is a flowchart of the method of the present invention;

[0046] Figure 2 This is a diagram of the three-level evaluation network structure in an embodiment of the present invention;

[0047] Figure 3 This is a flowchart of the sequential least squares iterative process in an embodiment of the present invention. Detailed Implementation

[0048] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0049] This embodiment provides a wind farm site selection decision-making method based on a custom cost function and sequential least squares optimization, such as... Figure 1 As shown, it includes the following steps:

[0050] S1. Obtain relevant influencing factor data from historical reference wind farm site selection cases, as well as operational measurement index data after the implementation of the cases, and establish a three-level evaluation network structure based on the relevant influencing factor data for comprehensive analysis.

[0051] In this embodiment, the acquired relevant influencing factor data includes four categories: production efficiency, construction cost, operating cost, and construction risk. The production efficiency category includes average wind speed and the sea area involved in the project. The construction cost category includes water depth, wind farm offshore distance, and equipment transportation distance. The operating cost category includes salt spray concentration, the sum of grid connection distance and 0.5 times the cable length, and the wind direction consistency coefficient. The construction risk category includes annual typhoon frequency, the sum of the annual average typhoon intensity and 0.7 times the annual maximum typhoon intensity, and the annual average wave height. An operational measurement index sequence (a1, a2, ... a) is established based on the operational measurement index data of n wind farms. i ,…,a n ), where a i Let represent the i-th wind farm. Furthermore, based on the current reference case data, a three-level evaluation network structure is established, such as... Figure 2 As shown, the first level is the influencing factor layer, which represents the specific influencing factors under each category; the second level is the decision-making standard layer, which represents the category of influencing factors; and the third level is the target layer, which represents the final comprehensive evaluation.

[0052] The operational metrics selected in this embodiment are the average annual benefits of the wind farm, expressed in millions of US dollars.

[0053] S2, based on a variable evaluation network structure inference conditional probability table, establishes a cost function that reflects the validity of the data of the corresponding evaluation indicators through a three-level evaluation network structure.

[0054] S2 includes the following steps:

[0055] S21, perform data fuzzing processing on the influencing factors data of historical reference wind farm cases.

[0056] The specific operation of fuzzifying the influencing factor data is as follows: calculate the average value of each influencing factor. The standard deviation σ is also included, and the corresponding fuzzy standard is divided into five evaluation levels, from worst to best: VB (Very Bad), B (Bad), A (Average), G (Good), and VG (Very Good). The original data values ​​corresponding to 100% membership of VB and VG evaluations are... The original data value corresponding to the 100% membership degree of B and G is... A. The original data value corresponding to a 100% membership degree. Finally, based on the principle of triangular fuzzification, the data is fuzzified to obtain evaluation data consisting of evaluation types and their membership degree combinations. Specifically, for practical situations... The corresponding VB or VG depends on whether increasing the value of the current influencing factor contributes to the wind farm evaluation; the evaluations for B and G are similar. Finally, the fuzzification criteria are obtained from the specific data in the implementation example, as shown in Table 1.

[0057] Table 1. Fuzzification Standards

[0058]

[0059]

[0060] S22, based on the conditional probability table of the variable evaluation network structure, calculate the comprehensive score b of the i-th wind farm. i And establish a reference wind farm scoring sequence (b1, b2, ..., b i …,b n The evaluation network structure inference conditional probability table determines the parameters in the table based on the hierarchical structure of the evaluation network and the evaluation level of each predefined unit in the network structure. The values ​​of the parameters are variable.

[0061] S23, based on the operational measurement indicator sequence (a1, a2, ..., a n ) and wind farm scoring sequence (b1,b2,…,b n A cost function is established to measure the effectiveness of the scoring sequence.

[0062] For the final obtained score (b1, b2, ..., b n If we can make the proportions of each item as close as possible to the sequence of evaluation indicators (a1, a2, ..., a...), n The closer the corresponding proportions are, the higher the validity of the rating is considered. Consider the following cost function, which is the sequence (b1k, b2k, ..., bk) obtained by multiplying each desired final rating by the same proportion coefficient k. n k) can approximate the evaluation metric sequence (a1, a2, ..., a) as closely as possible. n If the proportions of the two original sequences are equal, the corresponding ratios will be most similar, and the corresponding scoring validity will be the highest.

[0063]

[0064] Obviously, currently only if a t =b t Only when k is there a possibility that the part i = t is 0, and any a t ≠b tIn cases k, the cost function increases based on the difference in the ratios of the two factors. Therefore, the process of finding the optimal scoring network is transformed into finding the optimal scoring sequence (b1, b2, ..., b...). n The process of obtaining the cost function W(k) is then transformed into the process of minimizing the cost function W(k).

[0065] To facilitate further calculations, the cost function can be transformed into the following form to make it more mathematically feasible. At the same time, it is clear that minimizing the objective W(k) after the following transformation will not affect the determination of the optimal score sequence:

[0066]

[0067] For this function, first find its minimum value, and then consider differentiating both sides of the equation with respect to k:

[0068]

[0069] make Then, when At that time, the minimum value of the cost function W(k) can be found accordingly:

[0070]

[0071] Furthermore, a custom cost function is established to measure the effectiveness of the scoring sequence:

[0072]

[0073] in, is the auxiliary calculation sequence; x is the sequence of independent variables composed of the parameters of the variable evaluation network structure inference conditional probability table; n is the number of reference wind farm selection cases.

[0074] S3. The minimum value of the cost function is obtained by optimizing the sequence least squares method, and the corresponding conditional probability table is obtained.

[0075] Based on the sequential least squares method, the solution to the minimum value of the cost function is transformed into the following problem:

[0076]

[0077] Where x is a sequence of independent variables consisting of the parameters of the conditional probability table for the evaluation network structure inference. h i G represents an equality constraint. j This represents an inequality constraint; m is the number of equality constraints; p is the number of inequality constraints.

[0078] Specifically, the factors that decisively influence model accuracy are the setting of the conditional probability table and the weighted allocation of the final total score for the four categories in the overall evaluation. Consider defining all units involved in the conditional probability table setting, as well as the weights in the final weighted allocation, as functional variables. Regarding the number of functional variables, since both influencing factors and decision categories are divided into five levels, the first decision category (production efficiency category) involves 5... 2 =25 criteria, with the second to fourth categories of decision-making (construction cost, operating cost, and construction risk) involving 5 3 There are 125 criteria, each involving 5 possible evaluation outcomes. Adding the 4 coefficients involved in the final weight allocation, there are a total of 25*5 + 125*5*3 + 4 = 2004 independent variables. Correspondingly, constraints are added: for equality constraints, the sum of the 5 possible evaluation outcomes in each criterion should be 1, and the sum of the 4 coefficients in the final weight allocation should also be 1, resulting in 25 + 125*3 + 1 = 401 equality constraints; for inequality constraints, all independent variables should be real numbers between 0 and 1, resulting in 2*2004 = 4008 inequality constraints. The effectiveness of the final score is then optimized within the constrained solution space to determine a more effective conditional probability table configuration.

[0079] like Figure 3 As shown, the iterative calculation process based on the sequential least squares method includes the following steps:

[0080] S31, construct an approximate quadratic programming problem and solve the problem to obtain the search direction;

[0081] S32, determine the search step size based on the line search method, and update the current solution in the search direction;

[0082] S33, Convergence test.

[0083] In S31, based on the principle of sequential least squares, in the current solution x l At this point, we can construct the following approximate quadratic programming problem:

[0084]

[0085] in

[0086]

[0087] Let λ be the Lagrange function, λ and μ be the Lagrange multipliers for equality and inequality constraints, respectively, and l be the iteration number. This represents calculating the gradient. Let f(x) denote the Laplace operator, where f(x) is the cost function;

[0088] After solving this quadratic programming problem, the corresponding search direction d after the l-th iteration is obtained. l .

[0089] In S32, the current solution is updated based on the search step size and search direction:

[0090] x l+1 =x l +α l d l

[0091] Where, α l The search step size obtained using the line search method:

[0092]

[0093] α represents the step size. Note that in practice, to accelerate the calculation of the search step size, an inaccurate line search method based on the Amichow condition or Wolf condition is generally used to optimize the search step size calculation.

[0094] In S33, when the search direction d l When the norm or the change in the cost function is less than a set cutoff threshold, the algorithm terminates and obtains the optimal solution f(x) for the current cost function. l ) and the optimal solution x in the corresponding conditional probability table l .

[0095] S4, obtain relevant influencing factor data for candidate wind farms.

[0096] To address this, actual wind farm data not included in the reference wind farm case studies were used as a substitute, so that the feasibility of the method could be verified based on the evaluation metrics after the scores were obtained. A total of five wind farm data sets were used to simulate candidate wind farm data, as shown in Table 2:

[0097] Table 2 Simulated Candidate Wind Farm Data Table

[0098]

[0099] S5. Based on the conditional probability table obtained in step S3, evaluate the candidate wind farms and make the optimal wind farm site selection decision based on the evaluation scores.

[0100] Based on the existing data on relevant influencing factors of candidate wind farms, the data was first processed by fuzzification according to Table 1:

[0101] Table 3. Fuzzification of Simulated Candidate Wind Farm Data

[0102]

[0103] Based on existing data on relevant influencing factors of candidate wind farms and the optimal conditional probability table parameters obtained through final optimization, a comprehensive evaluation reasoning is performed in a three-level evaluation network. Utility values ​​of 0, 0.35, 0.55, 0.85, and 1.00 are assigned to five levels of evaluation (VB, B, A, G, VG), respectively. The weighted comprehensive score is calculated, as shown in Table 4.

[0104] Table 4 shows the final score obtained based on the method of this invention.

[0105]

[0106] The final score, obtained based on the average method reasoning, is shown in Table 5:

[0107] Table 5 shows the final scores obtained based on the average method reasoning.

[0108] OWF-1 OWF-2 OWF-3 OWF-4 OWF-5 Overall rating 0.65 0.68 0.35 0.73 0.52 Ranking 3 2 5 1 4

[0109] The actual performance metrics for the simulated candidate wind farm cases are shown in Table 6:

[0110] Table 6. Actual performance metrics data for simulated candidate wind farm cases.

[0111]

[0112] Comparing Tables 4, 5, and 6, it is found that the final score obtained by this method is almost identical to the target metric, with a corresponding cost function value of 0.12. This score effectively reflects the magnitude of the target metric (annual revenue in this example). In contrast, the final score obtained based on the averaging method does not provide a clear reference for site selection decisions, with a corresponding cost function value of 11.35, failing to reflect the actual site selection quality of wind farms. This demonstrates the effectiveness and scientific validity of the method presented in this invention.

[0113] Ultimately, based on the comprehensive scoring results, the candidate wind farm OWF-1 with the highest score was selected as the optimal decision.

[0114] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A wind farm siting decision method based on a user-defined cost function and a sequential least squares optimization, characterized in that, The method comprises the following steps: S1, obtaining relevant influence factor data of historical reference wind farm site selection cases and operation measurement index data after implementation of the cases, and establishing a three-level evaluation network structure for comprehensive analysis based on the relevant influence factor data; The relevant influence factor data obtained includes four categories of influence factors, namely, production benefit category, construction cost category, operation cost category and construction risk category, wherein the influence factors of the production benefit category include average wind speed and scheme sea area, the influence factors of the construction cost category include water depth, wind farm offshore distance and equipment transportation distance, the influence factors of the operation cost category include salt mist concentration, sum of power grid access distance and 0.5 times cable length, and wind direction consistency coefficient, and the influence factors of the construction risk category include typhoon annual frequency, sum of annual average typhoon intensity and 0.7 times annual maximum typhoon intensity, and annual average wave height; Based on n Operation measurement index data of the wind farm is used to establish an operation measurement index sequence wherein, represents the i th wind farm; S2, based on the variable evaluation network structure reasoning condition probability table, a cost function reflecting the effectiveness of the score corresponding to the measurement index data is established through the three-level evaluation network structure; S2 comprises the following steps: S21, data fuzzification processing is performed on the influence factor data of the historical reference wind farm cases; S22, based on the variable evaluation network structure inference conditional probability table, calculating the comprehensive score of the first wind farm i and establishing a reference wind farm score sequence wherein the evaluation network structure inference conditional probability table determines parameters in the table according to the hierarchical structure of the evaluation network and the evaluation grade of each unit predefined in the network structure, and the values of the parameters are variable.​ S23, based on the sequence of operational metrics with the sequence of wind farm scores , a cost function is established to measure the effectiveness of the sequence of scores, the cost function being: wherein, is a sequence of auxiliary calculations; is a sequence of arguments composed of variable evaluation network structure inference conditional probability table parameters; n is the number of reference wind farm selection cases; S3, the minimum value of the cost function is obtained based on the sequence least square method optimization, and the corresponding condition probability table is obtained; S4, obtaining relevant influence factor data of the candidate wind farm; S5, based on the condition probability table obtained in step S3, the candidate wind farm is evaluated, and optimal wind farm site selection decision is made according to the evaluation score.

2. The wind farm site selection method based on the self-defined cost function and the sequential least squares optimization according to claim 1, characterized in that, The three-level evaluation network structure is specifically: the first level is the influence factor layer, which represents the specific influence factors under each category; the second level is the decision standard layer, which represents the influence factor categories; and the third level is the target layer, which represents the final comprehensive evaluation.

3. The method of claim 1, wherein the method is characterized by, In S3, based on the sequence least square method, the solution of the minimum value of the cost function is converted into the following problem: wherein, is a sequence of arguments consisting of the variable network structure inference conditional probability table parameters, represents equality constraints, represents inequality constraints; m is the number of equality constraints; p is the number of inequality constraints.​ 4. The wind farm site selection method based on the self-defined cost function and the sequential least squares optimization according to claim 1, characterized in that, In S3, the iterative calculation process based on the sequence least square method comprises the following steps: S31, constructing an approximate quadratic programming problem and solving the problem to obtain a search direction; S32, determining a search step based on a line search method, and updating the current solution in the search direction; S33, convergence judgment; In the S31, based on the principle of sequence least square method, the current solution At this point, the following approximate quadratic programming problem is constructed: In S32, the current solution is updated according to the search step and the search direction: For Lagrange functions, These are the Lagrange multipliers for equality and inequality constraints, respectively. l For the number of iterations, This represents calculating the gradient. Represents the Laplace operator. The cost function; After solving the quadratic programming problem, the search direction corresponding to the lth iteration is obtained .

5. The method of claim 4, wherein the method is characterized in that, ​ wherein is the search step obtained based on the line search method: is the step size.

6. The method of claim 5, wherein the method is characterized by, In the S33, when the change amount of the norm or cost function is smaller than a set cutoff threshold value, the algorithm is terminated, and the current cost function optimal solution is obtained corresponding conditional probability table optimal solution corresponding conditional probability table optimal solution ​

Images