Mountain wind farm construction scheduling optimization method based on wind measurement tower measured meteorological data

Abstract

This invention provides an optimization method for construction scheduling of mountain wind farms based on meteorological data measured by meteorological towers. The method includes the following processes: Data acquisition: acquiring measured time-series wind speed data at multiple height levels from meteorological towers in the site area; Calibration and wind speed estimation: based on the measured time-series wind speed data, calibrating the wind shear index of the site area through logarithmic linear regression, and using this data to estimate the wind speed at the hub height of each wind turbine location. The estimation results are used to determine the feasibility of construction and hoisting; Probabilistic calendar construction and risk perception optimization. This invention transforms the judgment of preconditions for mountain wind power construction from relying on subjective experience to a data-driven quantitative assessment system, achieving a fundamental shift from passive response to proactive avoidance of climate risks, improving scheduling efficiency, and shortening the planned construction period.

Description

Technical Field

[0001] This invention relates to the field of wind power engineering construction management technology, and in particular to a method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower. Background Technology

[0002] Mountain wind farms typically consist of 15 to 20 wind turbines, with a project construction period of approximately 12 to 15 months. Taking the 6.25MW model as an example, the transportation and hoisting of the main components of a single wind turbine face extremely high constraints: the tower (single section) weighs 60 to 90 tons, and the load-bearing limit of mountain roads is a key bottleneck; the nacelle assembly (including the impeller and transmission system) weighs approximately 22 tons, is enormous, and has significant constraints on its ability to navigate mountain curves; a single blade weighs approximately 35 tons and is over 60 meters long, with the most stringent constraints on its ability to navigate mountain curves.

[0003] Currently, the scheduling of mountain wind power construction generally relies on the personal experience of project managers and uses Excel Gantt charts to manually create plans. This has revealed the following four types of defects in actual engineering: Inaccurate judgment of preconditions for hoisting operations: When making a conservative decision, the project manager chose not to operate even when the wind speed actually met the requirements, due to a lack of data to support the decision, resulting in the large crawler crane running idle and waiting, accumulating several days of lost construction time; when making an aggressive decision, the project manager forced operation when the wind speed exceeded the standard, which posed a serious risk of hoisting safety accidents and equipment collision damage.

[0004] The time required for road access to recover after rainfall cannot be quantitatively predicted: the transportation of heavy load-bearing items in mountainous areas places extremely high demands on the road surface's load-bearing capacity. After heavy rainfall, it typically takes 3 to 5 days for the road surface to regain its load-bearing capacity, but existing methods cannot quantitatively assess the recovery process, leading to frequent conflicts between transportation plans and actual road conditions, and causing a chain reaction of delays in the overall project schedule.

[0005] Seasonal transport restrictions are not systematically incorporated into the plan: softened roads during the rainy season and icy conditions in winter make it impossible to transport large items during certain periods, and the existing manual scheduling methods lack modeling for such systemic constraints.

[0006] The effects of multiple constraints and coupling are ignored, and rework is frequent: the chain reaction of hoisting delays, transportation waiting, platform preparation delays, and equipment arrival relies entirely on subjective judgment in existing methods, and plan generation requires multiple rounds of manual iteration, which is inefficient.

[0007] While large crawler cranes are typically equipped with real-time wind speed sensors at the top of their booms, which can automatically alarm and shut down when speeds exceed limits, this sensor addresses the issue of "real-time monitoring at the execution level" (a passive, reactive response). However, how to identify climate risks in advance during the construction planning phase and make forward-looking predictions at the planning level (proactive avoidance) to reduce the probability of forced shutdowns during the execution phase remains a technological gap. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an optimization method for construction scheduling of mountain wind farms based on meteorological data measured by wind towers. This method aims to solve problems in mountain wind power construction such as long-term reliance on the subjective experience of project managers, inaccurate quantitative prediction of hoisting windows, inability to dynamically assess road recovery time after rainfall, and frequent rework due to multiple constraints coupled in the construction plan. It achieves automatic generation of scheduling plans, proactive prevention of climate risks, and dynamic response to hard road condition constraints.

[0009] To achieve the above objectives, this invention provides an optimization method for construction scheduling of mountain wind farms based on meteorological data measured by a wind tower, comprising the following processing steps: Data acquisition: Acquire measured time-series wind speed data at multiple height levels from the meteorological towers in the site; Calibration and wind speed calculation: Based on the measured time-series wind speed data, the wind shear index of the site is calibrated using log-linear regression. and using the The wind speed at the hub height of each wind turbine location is calculated, and the calculation results are used to determine the feasibility of construction and hoisting. Probabilistic Calendar Construction: Based on historical hub height and wind speed data, a daily probability distribution of hoisting feasibility is statistically constructed throughout the year. The It reflects the historical probability of meeting hoisting weather conditions for each calendar day, and is used to quantify climate risk; Risk perception optimization: The probability distribution A climate risk term is embedded in the fitness function of a genetic algorithm, and a construction scheduling scheme is generated through the genetic algorithm.

[0010] Preferably, the In the calibration process, daily average wind speeds were simultaneously measured at three levels (30m, 50m, and 70m) of the wind measurement tower. Low wind speed data were filtered out, and logarithmic linear OLS regression through the origin was used to calibrate the actual wind speed data of the field area. value: ; in, For reference height, and output The standard error (SE) and 95% confidence interval (CI) were determined, and the calibration was repeated seasonally to obtain... The seasonal variation pattern.

[0011] As a preferred option, based on calibration Calculate the wind speed at the hub height of each wind turbine location in the site. The formula is: ; in, This is a preset altitude correction factor. For wind turbine The altitude difference between the wind measurement tower and the wind measurement tower.

[0012] As a preferred embodiment, the statistically constructed daily hoisting feasibility probability distribution P(d) for the whole year undergoes the following three-level determination: The first level is the baseline feasibility assessment: Statistical historical data shows that the calendar day meets the criteria of "average daily wind speed at hub height". The preset hoisting limit wind speed threshold exists, and there are continuous wind speeds on that day. The baseline feasibility probability is obtained by calculating the percentage of days within the "condition-satisfying time window with a preset minimum time window length" that meets the conditions. ; The second level is the P90 gust risk correction: when the P90 wind speed exceeds the preset gust risk threshold on that day, Adjust or reduce to 0; The third level is turbulence intensity degradation correction: when the turbulence intensity exceeds the preset turbulence threshold on a given day, Downgraded to a semi-feasible state; The second level takes precedence over the third level in the determination, and the modified P(d) sequence is smoothed by a cyclic sliding mean with a preset window length.

[0013] As a preferred option, data acquisition also includes acquiring ambient temperature, humidity, and atmospheric pressure data; The method also includes a road constraint generation step: based on measured air pressure and humidity data, identify rainfall events and calculate the road surface bearing capacity recovery time to generate daily dynamic passage feasibility masks for various types of large cargo; In risk perception optimization, the dynamic accessibility mask is used as a hard constraint for the transportation task and input into the genetic algorithm. The genetic algorithm is guided by a penalty term to avoid infeasible days.

[0014] Preferably, the road surface bearing capacity recovery time The calculation formula is: ; in, The preset baseline recovery days, As a baseline threshold for rainfall duration, For the duration of rainfall, The system presets an impact coefficient and classifies cargo passage for the engine compartment assembly, individual blades, and individual tower sections based on road surface recovery days and cargo weight. It then outputs the classification passage results to generate a dynamic passage feasibility mask.

[0015] Preferably, the fitness function of the genetic algorithm for: ; in, To constrain violations and penalties; Configurable weight parameters; climate risk item The calculation formula is: ; in, Automatically calculated from historical measured data of the wind measurement tower, For the lifting weight, The total number of hoisting tasks; the genetic algorithm uses an island model parallel search mechanism, an adaptive mutation rate mechanism, and an elite retention mechanism to solve the population evolution problem.

[0016] Preferably, the method also includes a visualization output step, in which the construction scheduling plan is output as a visual construction scheduling Gantt chart on the software display terminal to guide on-site construction.

[0017] This invention also proposes a construction scheduling optimization system for mountain wind farms based on meteorological data measured by a wind tower, including... Physical input layer: used to collect measured time-series wind speed data at multiple height levels, as well as ambient temperature, humidity and atmospheric pressure data, through the field anemometer tower sensors; Processing and computing layer: used to run the construction scheduling optimization method for mountain wind farms, transform the data into quantitative evaluation results of construction preconditions, and generate construction scheduling schemes through genetic algorithms; Visualization output layer: Used to output the construction scheduling plan as a visual construction scheduling Gantt chart on the software interface.

[0018] Compared with existing traditional manual experience-based scheduling methods, this invention has the following significant advantages: 1. Improved scheduling efficiency: By integrating physical sensors, processor computing, and visualization architecture, the traditional 3-5 day multi-round manual rework scheduling is shortened to about 20 seconds for automatic generation, improving efficiency by more than 99%.

[0019] 2. Reduced planned construction period: Due to systematic and detailed modeling of climate conditions and multi-process coupling, redundant waiting caused by experience-based scheduling was avoided. In a real project verification in Shanxi Province, the planned construction period was shortened from 385 days based on experience-based scheduling to 241 days based on algorithm optimization, a reduction of 37.4%.

[0020] 3. Transportation plan conflicts are reduced to zero: The dynamic response to rainfall events and road bearing capacity recovery time is achieved by using the air pressure, humidity and temperature data of the wind tower. The generated dynamic traffic feasibility mask is used as a hard constraint to limit the algorithm optimization, which completely eliminates the conflict between the heavy-load and large-item transportation plan and the actual road conditions after rainfall.

[0021] 4. Decision-making has shifted from "passive response" to "proactive foresight": This breaks through the traditional limitation of relying solely on sensors on top of the crane for post-event shutdown. At the planning level, a daily lifting probability calendar is used to guide the optimization direction, reducing the idle time losses of large crawler cranes. Attached Figure Description

[0022] Figure 1 This is a system architecture block diagram of the mountain wind farm construction scheduling optimization system based on meteorological data measured by a wind tower, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the data flow and technology chain of the mountain wind farm construction scheduling optimization method based on meteorological data measured by a wind tower, according to an embodiment of the present invention. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0024] Reference Figures 1-2 This invention provides an optimization method for construction scheduling of mountain wind farms based on meteorological data measured by a wind tower, including the following processing steps: Data acquisition: Acquire measured time-series wind speed data at multiple height levels from the meteorological towers in the site; Calibration and wind speed calculation: Based on the measured time-series wind speed data, the wind shear index of the site is calibrated using log-linear regression. and using the The wind speed at the hub height of each wind turbine location is calculated, and the calculation results are used to determine the feasibility of construction and hoisting. Probabilistic Calendar Construction: Based on historical hub height and wind speed data, a daily probability distribution of hoisting feasibility is statistically constructed throughout the year. The P(d) reflects the probability that the hoisting weather conditions are met historically for each calendar day, and is used to quantify the climate risk; Risk perception optimization: The probability distribution P(d) is embedded into the climate risk term of the fitness function of the genetic algorithm, and a construction scheduling scheme is generated through the genetic algorithm.

[0025] This invention uses a mountain wind farm project in Xiyang County, Jinzhong City, Shanxi Province as the main implementation scenario. This mountain wind farm is located in a typical Loess Plateau mountainous terrain, and includes 15-20 wind turbines. The overall planned construction period is approximately 15 months (about 456 days), covering the entire process from foundation construction and hoisting to grid connection and commissioning. In this embodiment, the wind turbine has a single unit capacity of 6.25MW, and the turbine locations are at an altitude of 1600-1700m. In this high-altitude mountainous terrain, due to the lower-than-standard air density and extremely complex topography, the hoisting and transportation of large and heavy equipment are subject to extremely stringent physical and environmental constraints.

[0026] Specifically, the main components' weight and mountain construction constraints for a single 6.25MW unit are as follows: 1. Tower (single section): A single section weighs 60 to 90 tons, and the load-bearing limit of mountain roads is the key bottleneck restricting their passage; 2. Nacelle assembly (including impeller and transmission system): weighing approximately 22 tons, its large size significantly restricts its passability on mountain curves; 3. Single blade: Each blade weighs about 35 tons, but its length exceeds 60 meters, making it most challenging to pass through mountain bends.

[0027] To achieve proactive pre-construction scheduling and error avoidance, this embodiment adopts a "physical sensor acquisition - processor calculation" approach. The system architecture features a "visualized software interface output." Firstly, at the physical input layer, historical mountain climate data (sampling history of 1-2 years, 2 years for Pian Guan County, and 1 year for other areas) is provided by physical sensors from three wind towers deployed in the site. This data serves as the driving force for all subsequent precondition judgment models. In the specific data acquisition and preprocessing operations, the system exports measured data in CSV format from the historical wind tower database in the site. Fields include timestamps, wind speeds (m / s) at 30m / 50m / 70m / hull height, etc. The system supports data adaptation at various time resolutions, from minute-by-minute to 10-minute intervals, and has automatic detection capabilities for various file encodings such as UTF-8-SIG and GBK, and can automatically match Chinese and English column names.

[0028] After acquiring the raw time series data, the system performs preprocessing steps: 1. Automatically filter and remove abnormal noise values ​​that exceed reasonable physical limits; 2. For data with consecutive missing durations not exceeding a preset time, linear interpolation is used for completion; 3. Aggregate the original high-frequency data at a 1-minute frequency into a 10-minute average, and then further aggregate it into daily statistics on a daily basis (including daily average wind speed, P90 wind speed, whether there is a continuous time window that can be hoisted, etc.), and input these into the calibration and optimization models of the processing and computing layer.

[0029] Preferably, the In the calibration process, daily average wind speeds were simultaneously measured at three levels (30m, 50m, and 70m) of the wind measurement tower. Low wind speed data were filtered out, and logarithmic linear OLS regression through the origin was used to calibrate the actual wind speed data of the field area. value: ; in, For reference height, and output The standard error (SE) and 95% confidence interval (CI) were determined, and the calibration was repeated seasonally to obtain... The seasonal variation pattern.

[0030] In this embodiment of the invention, the wind shear index The measured calibration serves to quantify daily hoisting weather risks at the construction execution level. During algorithm execution, the reference height... The input time series data points are set to 50m. The wind speeds at both 30m and 50m must be greater than the preset low wind speed filtering threshold (typically 1m / s in this embodiment). This filters out data from windless or lightly windy sections, thereby improving the statistical accuracy of hub wind speed estimation in subsequent construction scheduling.

[0031] To verify the ability of the method of this invention to capture the local circulation effect of mountainous terrain, this embodiment uses historical data from wind measurement towers in three real mountainous wind farms in Shanxi Province with similar altitudes but different topography for cross-project comparative calibration. The calibration results are as follows: 1. Shanxi Project A (Pianguan County): Wind measurement tower altitude 1570m, year-round actual measurements The value is 0.1143, and the fit is... The value is 0.2553, and the 95% confidence interval of the calibrated output is [0.1084, 0.1202]. The upper bound of its 95% confidence interval, 0.1202, is significantly lower than the commonly used industry standard default empirical value of 0.143, with a deviation of -20.1%.

[0032] 2. Shanxi Project B (Xishuijie Township): Wind measurement tower altitude 1672m, year-round actual measurements The value is 0.1102, and the fit is... The value is 0.4941, and the 95% confidence interval of the calibrated output is [0.1041, 0.1163]. The upper bound of its 95% confidence interval, 0.1163, is significantly lower than 0.143, with a deviation of -22.9%.

[0033] 3. Shanxi Project C (Mafang Township): Wind measurement tower altitude 1553m, year-round actual measurements The value is 0.1254, and the fit is... The value is 0.3346, and the 95% confidence interval for the calibrated output is [0.1146, 0.1362]. The upper bound of its 95% confidence interval, 0.1362, is still lower than 0.143, with a deviation of -12.3%.

[0034] The aforementioned real data shows that, at a 95% confidence level, the wind shear index of mountain wind farms is statistically significantly lower than the default value of 0.143 for plains areas. Using the traditional empirical value of 0.143 to extrapolate the wind speed at the hub height at high altitudes will lead to a systematic overestimation error, making the prediction of the hoisting window overly conservative, thus wasting a significant number of valuable actually operable days. Furthermore, for the three sites with an altitude difference within 150m, The measured range reached 0.0152 (coefficient of variation 9.3%), proving that the complex slope aspect and ridge morphology of the mountains make it impossible to replace it with any single empirical value. Furthermore, the results of seasonal repeated calibration show that Pianguan County in summer and autumn... The range of values ​​even reached 53.8% (0.0880 vs 0.1353), indicating that the variation range of the same field area in different seasons even exceeded the spatial differences between different fields. It is necessary to use the measured calibration method proposed in this invention for dynamic and systematic deep coupling.

[0035] As a preferred option, based on calibration Calculate the wind speed at the hub height of each wind turbine location in the site. The formula is: ; in, This is a preset altitude correction factor. For wind turbine The altitude difference between the wind measurement tower and the wind measurement tower.

[0036] Due to the significant elevation variations and localized airflow effects in mountainous terrain, this invention calculates the specific locations of each wind turbine within the site. When considering wind speed at wheel hub height, micro-terrain correction is introduced. In this embodiment, a preset altitude correction coefficient is used. The typical value is set at 0.00008 / m. This coefficient is designed based on the unique topographical flow-guiding effect of mountainous terrain; for every 100m increase in elevation at the wind turbine location, the local wind speed increases by approximately 0.8% due to the terrain's lifting effect. The system reads data from each wind turbine location... The actual altitude difference relative to the wind measurement tower Combined with the data uniquely determined by actual measurements This value enables precise and differentiated assessment of climate risks at each aircraft location, avoiding safety or work stoppage risks caused by a one-size-fits-all macro-calculation.

[0037] As a preferred embodiment, the statistically constructed daily hoisting feasibility probability distribution P(d) for the whole year undergoes the following three-level determination: The first level is the baseline feasibility assessment: Statistical historical data shows that the calendar day meets the criteria of "average daily wind speed at hub height". The preset hoisting limit wind speed threshold exists, and there are continuous wind speeds on that day. The baseline feasibility probability is obtained by calculating the percentage of days within the "condition-satisfying time window with a preset minimum time window length" that meets the conditions. ; The second level is the P90 gust risk correction: when the P90 wind speed exceeds the preset gust risk threshold on that day, Adjust or reduce to 0; The third level is turbulence intensity degradation correction: when the turbulence intensity exceeds the preset turbulence threshold on a given day, Downgraded to a semi-feasible state; The second level takes precedence over the third level in the determination, and the modified P(d) sequence is smoothed by a cyclic sliding mean with a preset window length.

[0038] In this embodiment, for each calendar day The probability calculation parameter configuration and logic details are as follows: 1. First-level judgment: The preset hoisting wind speed limit is configured as 8 m / s (which can be flexibly configured according to the specifications of the crawler crane and wind turbine equipment), and the preset minimum time window length is configured as 2 hours. Based on this, the percentage of days in historical time series data that fully meet the requirements is calculated, and the result is output. .

[0039] 2. Second-level judgment: The preset gust risk threshold is set at 9 m / s. When the historical P90 wind speed (reflecting the 10% extreme high wind speed level of the day) of a certain calendar day exceeds 9 m / s, it indicates that the probability of a drastic sudden gust of wind occurring that day is extremely high. At this time, regardless of whether the daily average wind speed meets the standard, the system will forcibly reduce the final hoisting probability for that day. The correction was reduced to 0.

[0040] 3. Third-level judgment: The preset turbulence threshold (standard deviation of wind speed / mean wind speed) is set to 0.15. When the turbulence intensity exceeds 0.15, or when on-site turbulence time-series data is missing, the system determines that the day belongs to a semi-feasible state with high weather risk and forcibly downgrades the probability value to 0.5 times the original base probability. If P90 has already exceeded the standard and returned to 0 in the second level, the third-level judgment will not be performed.

[0041] To eliminate statistical noise from random daily weather events and to mitigate the boundary fault effect at the beginning and end of the year, the system uses a 15-day preset window length for the 365-day period. The sequence is smoothed using a cyclic moving average. Specifically, the sequence is extended by 15 days from both ends (i.e., the last 15 days are connected to the beginning, and the first 15 days are connected to the last), and the extended cyclic sequence is smoothed using an equal-weighted moving average within a window. Finally, the middle 365 days are truncated as the final probability calendar output.

[0042] Based on the above model, this embodiment calculates the historical measured data of three mountainous areas, and the statistical summary results of the daily probability calendar are as follows: Shanxi Project A: The average probability of hoisting is 47.8% throughout the year. The best hoisting window is August (probability 62.9%) and December (probability 61.2%), while the worst hoisting window is April (due to frequent sudden winds, probability only 22.4%). The average hoisting time throughout the year is 19.7 hours / day.

[0043] Shanxi Project B: The average annual hoisting probability is 34.5%, the best hoisting window is August (probability 42.7%), the worst window is February (20.6%) and April to June (consistently less than 22%), and the average annual hoisting time is 15.5 hours / day.

[0044] Project C in Shanxi: The average probability of hoisting is 37.5% throughout the year. The best hoisting window is August (65.0%) and October (58.1%), while the worst window is from January to April (consistently less than 27%). The average hoisting time throughout the year is 15.2 hours / day.

[0045] The calculation results show that the Pearson correlation coefficients of the probability distribution patterns of the three closely spaced site areas are only between 0.337 and 0.519 (average 0.451), exhibiting significant site-specific climate characteristics. Continuing to rely on traditional experience or fixed production windows will inevitably cause each project to miss its unique optimal weather bonus days. Furthermore, none of the three site areas experience any weather conditions throughout the year. The absolute safe period is only around 60%, even at its best, indicating that the weather window for mountain hoisting is generally limited. Accurate quantitative calculation of the daily probability calendar is the core mechanism for achieving "data-driven risk perception".

[0046] Preferably, the data acquisition also includes acquiring ambient temperature, humidity, and atmospheric pressure data; The method also includes a road constraint generation step: based on measured air pressure and humidity data, identify rainfall events and calculate the road surface bearing capacity recovery time to generate daily dynamic passage feasibility masks for various types of large cargo; In the risk perception optimization, the dynamic access feasibility mask is used as a hard constraint for the transportation task and input into the genetic algorithm. The genetic algorithm is guided by a penalty term to avoid infeasible days.

[0047] In this embodiment, the exported wind measurement tower data includes complete meteorological indicators such as ambient temperature (°C), ambient humidity (%), and ambient pressure (hPa). Because the transportation of large, heavy-duty items (with single items weighing up to 60-90 tons) places extremely high demands on the axle load-bearing capacity of mountain roadbeds, heavy rainfall can lead to roadbed softening. To achieve dynamic road condition perception in wind farm areas without dedicated rainfall sensors, the system utilizes existing air pressure, humidity, and temperature sensor data from the wind measurement tower to indirectly identify rainfall through meteorological dynamic correlation characteristics.

[0048] The specific rules for identifying rainfall events are as follows: when the system detects that within a preset time window (6 hours in this embodiment), the cumulative decrease in atmospheric pressure exceeds a preset pressure decrease threshold (the standard threshold in this embodiment is 5 hPa, corresponding to the pressure gradient change of a typical atmospheric front or low-pressure trough weather system in meteorology), and the highest ambient humidity on that day exceeds a preset humidity threshold (85% in this embodiment), a rainfall event is determined to have occurred, and the duration of the rainfall is automatically recorded. The algorithm system in this embodiment also has an adaptive robust mechanism: when humidity sensor data is missing, the system will automatically lower the air pressure drop threshold to 3 hPa as an independent basis for rainfall determination. All the above threshold parameters can be calibrated and dynamically adjusted through cross-referencing of historical construction logs and meteorological data.

[0049] Preferably, the road surface bearing capacity recovery time The calculation formula is: ; in, The preset baseline recovery days, As a baseline threshold for rainfall duration, For the duration of rainfall, The system presets an impact coefficient and classifies cargo passage for the engine compartment assembly, individual blades, and individual tower sections based on road surface recovery days and cargo weight. It then outputs the classification passage results to generate a dynamic passage feasibility mask.

[0050] In this embodiment, the preset parameters in the formula are derived from actual historical data of construction logs of three mountain wind farm projects in Shanxi Province through mathematical statistical regression and joint calibration analysis. Their specific values ​​are configured as follows: preset baseline recovery days. Day, baseline threshold for rainfall duration Hourly rainfall duration impact coefficient Wind speed influence coefficient Temperature influence coefficient .in The values ​​represent the average measured wind speed and air temperature during the subsequent window period after rainfall, respectively, to quantitatively simulate the promoting effect of evaporation environment on soil subgrade moisture evaporation and bearing capacity recovery.

[0051] This embodiment utilizes the aforementioned model to detect a total of 42 rainfall events in historical meteorological sequences (including 21 in spring, 7 in summer, 7 in autumn, and 7 in winter, with pressure drops ranging from 3.0 to 88.4 hPa), and parameter calibration was performed based on construction shift logs. A typical rainfall event example is as follows: On November 26, 2019, a strong low-pressure cyclone passed through Xishuijie Township, causing a sudden pressure drop of 88.4 hPa. The model estimated the recovery time of the heavy-load road surface bearing capacity. The estimated recovery time was 5.0 days, which matches the actual shutdown record in the field log; on February 8, 2018, a weak frontal system passed through Pianguan County, with a pressure drop of 4.5 hPa (the humidity deficiency triggered the independent criterion), and the model estimated the recovery time to be 4.3 days.

[0052] When generating dynamic access feasibility masks, the system changes the traditional extensive management model of "banning all large items from passing for several days after rain" and instead executes a refined cargo classification access control matrix as shown below, based on the single-item heavy tonnage of different goods (in terms of weight). (Taking a typical scenario as an example) Day 1 of recovery: The nacelle assembly (approximately 22t), a single blade (approximately 35t), and a single tower section (60-90t) are all [prohibited from passage]. Day 2 of recovery: The nacelle assembly (approximately 22t), a single blade (approximately 35t), and a single tower section (60-90t) are all [prohibited from passage]. Day 3 of recovery: The lighter [nacelle assembly (approx. 22t)] has met the road load-bearing capacity standard and is now [allowed to pass]; while the heavier [single blade (approx. 35t)] and the heaviest [tower section (60-90t)] remain [prohibited from passing]; After full recovery (day 4 and onwards): All large items will be switched to "Allow passage".

[0053] By outputting the above-mentioned hierarchical access judgment results daily for 365 days of the year, the system will automatically generate a dynamic access feasibility mask composed of a matrix of 0s and 1s, which will be used as a hard constraint condition input into the processor's genetic algorithm.

[0054] Preferably, the fitness function of the genetic algorithm for: ; in, To constrain violations and penalties; Configurable weight parameters; climate risk item The calculation formula is: ; in, Automatically calculated from historical measured data of the wind measurement tower, For the lifting weight, The total number of hoisting tasks; the genetic algorithm uses an island model parallel search mechanism, an adaptive mutation rate mechanism, and an elite retention mechanism to solve the population evolution problem.

[0055] In this embodiment, the design and parameter details of each term of the fitness function of the multi-objective optimization algorithm are as follows: 1. Configurable weight parameters: Based on the actual situation of wind power projects, since the penalty for delay in construction is usually the highest, followed by construction cost, and resource balance affects overall efficiency, this embodiment scientifically configures them as follows: .

[0056] 2. Routine scheduling target item: Schedule normalization item Cost normalization item Resource balance item .

[0057] 3. Climate Risk Items : Historical measured probability calendar from wind measurement towers Automatic drive, by multiplying by a normalized lifting weight coefficient This allows the algorithm to automatically and dynamically schedule heavier lifting tasks on dates with a higher probability of weather safety during production scheduling, thus achieving intelligent risk avoidance.

[0058] 4. Constraints on penalties for violations For various soft and hard constraints faced by the project, including exceeding labor limits, exceeding construction resource limits, exceeding the number of main cranes, exceeding the limits for parallel transportation in a single mountain corridor, exceeding the limits for actual crane lifting capacity - high-altitude wind speed, infeasibility of heavy-load roads (transportation was scheduled on dates with a dynamic feasibility mask of 0), and violations of prerequisite topology relationships, the violation ratios were calculated and accumulated according to their weights. Among them, violations of prerequisite relationships were subject to the highest weight penalty to ensure that the generated solution is absolutely feasible in terms of topology.

[0059] The complete execution process of the risk-aware genetic algorithm used in this embodiment includes the following seven specific stages: Phase 1, Chromosome Encoding and Initialization: Each individual is encoded by a mixture of three sets of genes: ① Task execution sequence genes ① Generated from a valid topological sort sequence through random shuffling to ensure task prerequisites; ② Hoisting window offset gene Random values ​​are selected within [0, optimal window number] to control the selection preference within the candidate optimal time window; ③ Delay quantity gene The value is randomly selected within the range of [0, 20] days, which controls the number of days the task is delayed relative to the earliest possible start date.

[0060] Phase Two, Chromosome Decoding: According to The tasks in the task list are arranged in order, and their actual start dates are calculated one by one. First, the earliest possible start date is determined based on the preceding tasks; if it is a hoisting task, then... Select the corresponding index window from the optimal hoisting window list and align the dates; for large-item transportation tasks, scan the dynamic feasibility mask daily starting from the earliest start date, skip infeasible dates, and limit the time to the total project duration.

[0061] Phase Three, Risk Perception Adaptability Assessment: For the decoded solution, calculate the four normalized objectives and their correlation with the above formula. The sum of penalty terms is used to output the fitness value.

[0062] Phase Four, Island Parallel Evolution: The total population is evenly distributed into four independent evolutionary subpopulations (i.e., "islands," with 60 individuals per island). During each generation, each island independently performs the following basic operations: Elite Preservation (the top 5% of fittest individuals are directly copied to the next generation without modification), Tournament Selection (three individuals are randomly selected, and the best one is chosen as the parent), and Sequential Crossover (OX) (in the process of crossover between two parents). Segment exchange occurs between gene segments to maintain a valid arrangement, while Arithmetic crossover operator) and mutation operation (randomly selected from exchange mutation, delayed perturbation mutation or offset mutation).

[0063] Phase 5, Inter-island migration: Every 25 generations, the 5 apex individuals with the highest fitness are selected from each island and migrate along a circular topological path. The gene is replicated to neighboring islands, forcibly replacing the five individuals with the lowest fitness on the target island. This mechanism, by periodically introducing superior external genes, effectively breaks the trap of a single local optimum, maintains the diversity of evolutionary directions, and promotes global convergence.

[0064] Phase Six, Early Termination Judgment: After each generation of evolution, check whether the global optimal fitness has improved. If there is no improvement for 50 consecutive generations, the iteration is automatically terminated in advance to avoid invalid calculations.

[0065] Phase 7, Local Search Refinement: After the global evolution of the algorithm terminates, perform 200 neighborhood local searches on the selected optimal individual (each time randomly adjusting the delay offset of a single task). (The algorithm may randomly swap the execution order of two adjacent tasks, or replace them if the new solution is better.) The results of this embodiment show that the algorithm successfully converged and triggered early stopping in the 79th generation, with a total computation time of about 20 seconds, and the final convergence fitness stabilized at -1.5508.

[0066] Preferably, the method also includes a visualization output step, in which the construction scheduling plan is output as a visual construction scheduling Gantt chart on the software display terminal to guide on-site construction.

[0067] In this embodiment, the construction scheduling Gantt chart generated by the visualization output layer intuitively presents the planned time arrangements for each wind turbine and each process (such as foundation pouring, large tower component transportation, blade transportation, on-site hoisting, etc.), allowing the project manager to directly command on-site vehicle scheduling, equipment entry, and large crawler crane relocation operations on the software display terminal.

[0068] To quantify the beneficial technical effects of the method of this invention, this embodiment conducts a comprehensive empirical comparison between the algorithm-generated scheme and the traditional Gantt chart scheduling scheme based on the manual experience of project managers. The comparative experimental data results are as follows: 1. Scheduling scheme generation time: Traditional manual experience-based scheduling methods involve multiple rounds of manual iteration and repeated rework due to conflicts of various constraints, and it takes 3 to 5 days to generate a scheme; while the method of this invention automatically optimizes and solves through algorithms, and can automatically generate a scheme in only about 20 seconds, improving scheduling efficiency by more than 99%.

[0069] 2. Planned construction period: The traditional experience-based production scheduling scheme is 385 days (about 15 months); however, the method of this invention, under the premise of fully avoiding climate risks and road condition conflicts, finds the optimal solution with the shortest construction period, and the planned total construction period is greatly shortened to 241 days (about 8 months), which is a total reduction of 144 days for the project, and the construction period reduction rate is 37.4%.

[0070] 3. Quantification of Climate Risk in Lifting Operations: Traditional empirical methods are completely unable to quantitatively express this risk, relying solely on subjective memory and guesswork. In contrast, the method of this invention can clearly calculate the probability of each task throughout the entire period, and the climate risk score of the final optimal solution is: This has enabled the digitalization and high transparency of decision-making.

[0071] 4. Number of transportation plan conflicts: In manual experience-based solutions, conflicts frequently occur due to discrepancies between the transportation plan for large items and the actual road surface bearing capacity after heavy rain, causing vehicles to get stuck on mountain roads. The number of such conflicts is uncertain. However, the method of this invention uses dynamic road traffic masks to rigidly constrain and avoid conflicts, thus reducing the number of transportation plan conflicts to zero.

[0072] 5. Accuracy of road constraint modeling: Traditional methods can only set static approximate prohibition periods, with extremely limited accuracy; this method dynamically updates the constraints based on meteorological dynamics for 42 rainfall events identified in actual measurements, significantly improving the accuracy of the constraints.

[0073] This invention also proposes a construction scheduling optimization system for mountain wind farms based on meteorological data measured by a wind tower, including... Physical input layer: used to collect measured time-series wind speed data at multiple height levels, as well as ambient temperature, humidity and atmospheric pressure data, through the field anemometer tower sensors; Processing and computing layer: used to run the construction scheduling optimization method for mountain wind farms, transform the data into quantitative evaluation results of construction preconditions, and generate construction scheduling schemes through genetic algorithms; Visualization output layer: Used to output the construction scheduling plan as a visual construction scheduling Gantt chart on the software interface.

[0074] The system constructed in this embodiment of the invention has the following organizational relationships and functional correspondences at the hardware and software levels: 1. Physical Input Layer: The hardware relies on three representative wind measurement towers arranged according to regulations within the site. Their sensors include multiple anemometers, electronic thermometers, volumetric hygrometers, and precision barometers located at 30m, 50m, 70m, and hub height. The sensors perform continuous online data acquisition at an ultra-high frequency of 1 minute or 10 minutes, accumulating a 1.5-2 year historical time-series database, which serves as the sole physical data source driving all upper-level precondition evaluation models and optimization solvers of the entire scheduling system.

[0075] 2. Processing and Computation Layer: This layer consists of a microcomputer software processing system that runs the wind shear measurement calibration model, the daily hoisting probability calendar model, the dynamic pavement bearing capacity assessment model, and the risk perception parallel genetic algorithm described in this invention. Its function is to convert the collected physical meteorological data into quantitative probabilities and hard constraints for specific processes, such as "whether hoisting and transportation are possible," and then utilize parallel CPU computing power for population optimization to automatically generate the optimal production scheduling results.

[0076] 3. Visualization Output Layer: The system terminal's graphical user interface (GUI). It automatically converts the digitally optimal chromosome scheme calculated by the processing and calculation layer into a visualized construction scheduling Gantt chart that conforms to the reading habits of project managers and on-site commanders, and connects each wind turbine and process sequentially. This chart is directly used to guide and issue daily construction command instructions for mountain wind farms.

[0077] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for optimizing construction scheduling of a mountainous wind farm based on measured meteorological data of a wind measurement tower, characterized in that, The following processing steps are included: Data acquisition: Acquire measured time-series wind speed data at multiple height levels from the meteorological towers in the site; Calibration and wind speed estimation: Based on the measured time-series wind speed data, the wind shear index of the field area is calibrated by logarithmic linear regression. and using the Calculate the wind speed at the hub height of each wind turbine location, and use the calculation results to determine the feasibility of construction and hoisting. Probabilistic calendar construction: Based on historical hub height and wind speed data, a daily probability distribution P(d) of hoisting feasibility is constructed throughout the year. P(d) reflects the probability of meeting hoisting climate conditions on each calendar day in history, and is used to quantify climate risk. Risk perception optimization: The probability distribution P(d) is embedded into the climate risk term of the fitness function of the genetic algorithm, and a construction scheduling scheme is generated through the genetic algorithm.

2. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 1, characterized in that, In the calibration process, daily average wind speeds were simultaneously measured at three levels (30m, 50m, and 70m) of the wind measurement tower. Low wind speed data were filtered out, and logarithmic linear OLS regression through the origin was used to calibrate the actual wind speed data of the field area. value: ； in, For reference height, and output The standard error (SE) and 95% confidence interval (CI) were determined, and the calibration was repeated seasonally to obtain... The seasonal variation pattern.

3. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 1, characterized in that, Based on calibration Calculate the wind speed at the hub height of each wind turbine location in the site. The formula is: ； in, This is a preset altitude correction factor. For wind turbine The altitude difference between the wind measurement tower and the wind measurement tower.

4. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 1, characterized in that, The statistically constructed daily hoisting feasibility probability distribution P(d) for the entire year is determined through the following three levels of criteria: The first level is the baseline feasibility assessment: Statistical historical data shows that the calendar day meets the criteria of "average daily wind speed at wheel hub height". The preset hoisting limit wind speed threshold exists, and there are continuous wind speeds on that day. The baseline feasibility probability is obtained by calculating the percentage of days within the "condition-satisfying time window with a preset minimum time window length" that meets the conditions. ; The second level is the P90 gust risk correction: when the P90 wind speed exceeds the preset gust risk threshold on that day, Adjust or reduce to 0; The third level is turbulence intensity degradation correction: when the turbulence intensity exceeds the preset turbulence threshold on a given day, Downgraded to a semi-feasible state; The second level takes precedence over the third level in the determination, and the modified P(d) sequence is smoothed by a cyclic sliding mean with a preset window length.

5. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 1, characterized in that, Data acquisition also includes acquiring ambient temperature, humidity, and atmospheric pressure data; The method also includes a road constraint generation step: based on measured air pressure and humidity data, identify rainfall events and calculate the road surface bearing capacity recovery time to generate daily dynamic passage feasibility masks for various types of large cargo; In risk perception optimization, a dynamic accessibility mask is used as a hard constraint input to the genetic algorithm for the transportation task, and a penalty term guides the genetic algorithm to avoid infeasible days.

6. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 5, characterized in that, The road surface bearing capacity recovery time The calculation formula is: ； in, The preset baseline recovery days, As a baseline threshold for rainfall duration, For the duration of rainfall, The system presets an impact coefficient and classifies cargo passage for the engine compartment assembly, individual blades, and individual tower sections based on road surface recovery days and cargo weight. It then outputs the classification passage results to generate a dynamic passage feasibility mask.

7. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 1 or 5, characterized in that, The fitness function of the genetic algorithm for: ； in, To constrain violations and penalties; Configurable weight parameters; climate risk item The calculation formula is: ； in, Automatically calculated from historical measured data of the wind measurement tower, For the lifting weight, The total number of hoisting tasks; the genetic algorithm uses an island model parallel search mechanism, an adaptive mutation rate mechanism, and an elite retention mechanism to solve the population evolution problem.

8. The method for optimizing the construction scheduling of mountain wind farms based on meteorological data measured by a wind tower as described in claim 1, characterized in that, It also includes a visualization output step, in which the construction scheduling plan is output as a visual construction scheduling Gantt chart on the software display terminal to guide on-site construction.

9. A construction scheduling optimization system for mountain wind farms based on meteorological data measured by a wind tower, characterized in that, include Physical input layer: used to collect measured time-series wind speed data at multiple height levels, as well as ambient temperature, humidity and atmospheric pressure data, through the field anemometer tower sensors; Processing and computing layer: used to run the method according to any one of claims 1 to 8, to convert the data into quantitative evaluation results of construction preconditions, and to generate a construction scheduling scheme through a genetic algorithm; Visualization output layer: Used to output the construction scheduling plan as a visual construction scheduling Gantt chart on the software interface.

Images