Construction site temporary power capacity dynamic configuration method and system based on construction progress coupling load prediction and storage medium

Abstract

The application provides a construction site temporary power capacity dynamic configuration method, system and storage medium based on construction progress coupled load prediction, which can consider future load uncertainty, maintain equipment allocation and total load conservation, locate capacity bottlenecks along the real power supply path, and convert bottleneck checking results into next prediction period capacity configuration state, and relates to the technical field of construction site temporary power capacity management. The method comprises the following steps: acquiring multi-source data of a construction site; predicting a plurality of active load quantile prediction values reflecting different degrees of conservatism within a future set time window; determining the operation state of each power-consuming equipment and the uncalibrated expected active load within the future period to obtain the allocated load of each power-consuming equipment; calculating the load state of each power supply and distribution link and identifying bottleneck links with capacity risks; generating a capacity adjustment strategy; and outputting the capacity adjustment strategy.

Description

Technical Field

[0001] This invention relates to the field of temporary power capacity management technology at construction sites, specifically a method, system, and storage medium for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction. Background Technology

[0002] Temporary power systems at construction sites typically consist of temporary transformers, primary distribution boxes, secondary distribution boxes, switch boxes, branch cables, and construction equipment. As construction progresses through stages such as foundation work, main structure installation, electromechanical installation, decoration, and finishing touches, tower cranes, construction hoists, welding equipment, drainage equipment, ventilation equipment, lighting, and office loads continuously move in and out of the site, altering the concurrent operation of similar equipment and the spatial distribution of loads. Therefore, even when the total load on a single transformer is within acceptable limits, some distribution boxes or branch cables may still experience insufficient capacity margins, potentially leading to unplanned power outages and disruptions to construction progress.

[0003] Existing temporary power capacity configurations for construction are mostly based on the total equipment capacity and demand factor determined at the initial stage of construction, with insufficient consideration for changes in subsequent construction phases, equipment operating times, and concentrated loads on local circuits. Partial load forecasting methods only provide a single total load forecast value and do not decompose the forecast results to different levels of equipment and radial topology, making it difficult to directly determine whether the risk occurs in the transformer, a certain level of distribution box, or a specific branch cable.

[0004] Furthermore, construction sites are affected by complex factors such as construction phases, equipment plans, shift schedules, and the environment. Directly allocating the predicted total load to the rated power would not reflect the dynamic adjustability of each equipment combination and would lack practical feasibility. Therefore, a technical solution is needed that can consider the uncertainty of future loads, maintain the balance between equipment allocation and total load, locate capacity bottlenecks along the actual power supply path, and convert the bottleneck verification results into the capacity configuration state for the next forecast period. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, system and storage medium for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction, which can take into account the uncertainty of future load, maintain the balance of equipment allocation and total load, locate capacity bottlenecks along the actual power supply path, and convert the bottleneck verification results into the capacity configuration state of the next prediction period.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction, comprising the following steps: S1. Obtain multi-source data from the construction site, including construction schedule data, equipment parameter data, operating status data, historical electricity consumption data, on-site operating condition data, and power supply and distribution topology data; based on the power supply and distribution topology data, determine the power supply path information from each electrical device to the power supply source, and perform time alignment processing on the multi-source data to form a temporary electricity consumption dataset arranged according to a unified time scale. S2. Based on the current time and its historical time series information in the temporary power consumption dataset, determine the information for prediction, and predict multiple active load quantile prediction values ​​reflecting different levels of conservatism within a future set time window according to the information for prediction. S3. Based on the construction schedule data and on-site operation condition data, determine the commissioning status of each electrical equipment and the uncalibrated expected active load in the future period, and perform a conservation allocation of the uncalibrated expected active load according to the total load constraints corresponding to the multiple active load quantile prediction values ​​to obtain the allocated load of each electrical equipment. S4. Based on the electrical parameters of each electrical device, the allocated load is converted into capacity demand, and the equivalent active load and equivalent reactive load of each power supply link in the power supply and distribution topology data are determined by accumulating the load step by step from the end device side to the power supply side along the power supply path information. S5. Calculate the load status of each power supply and distribution link based on the rated capacity or allowable current carrying capacity of each power supply and distribution link, and identify bottleneck links with capacity risks based on the comparison results between the load status and the preset capacity conditions. S6. Calculate the required adjustment capacity based on the current load status and target load rate of the bottleneck link, and generate a capacity adjustment strategy based on the power supply guarantee priority and adjustability of the downstream electrical equipment of the bottleneck link. The capacity adjustment strategy includes at least one or more of the following: load shifting measures, load transfer measures, or capacity expansion prompts. S7. Output the capacity adjustment strategy, write the confirmed capacity adjustment strategy into the configuration status of the next prediction cycle, and execute subsequent predictions, verifications and configurations in a rolling manner as construction progress and operation data are updated.

[0007] Furthermore, the construction schedule data includes the division of construction phases, the start and end times of each phase, the progress ratio, and the equipment entry and exit schedule; the equipment parameter data includes rated power, power factor, rated current, power supply priority, and start-up impact coefficient; the operating status data includes equipment start-up and shutdown status, operating current, active power, and meter readings; and the on-site operating condition data includes shift information, weekday / holiday information, weather information, and concurrent operation scenario information.

[0008] Furthermore, the power supply path information is an ordered set of nodes from the power device access node through each level of parent nodes to the power root node, and an ordered set of branch cables connecting adjacent nodes; the power supply path information is updated as the access location of the power device or the power supply and distribution topology changes.

[0009] Furthermore, the plurality of active load percentile prediction values ​​include: the 50th percentile prediction value P, which characterizes the normal load level. 50 The 90th percentile predicted value P, representing a higher load level 90 And the 95th percentile predicted value P, which characterizes the capacity pressure in the high-load zone. 95 ;P 50 P 90 and P 95 satisfy Monotonicity constraints.

[0010] Furthermore, the prediction of the plurality of active load quantile prediction values ​​includes: Let the sampling interval be Δt, and the future set time window contain H sampling periods. The future prediction label is defined as y(t)=max{P(t+1),P(t+2),...,P(t+H)}, where H is the number of sampling periods contained in the future set time window. The label does not include the load at time t. The training set, validation set, and test set are divided in chronological order, and isolation intervals of no less than H sampling periods are set between adjacent data subsets; The uncertainty prediction model outputs active load quantile predictions with quantiles q of 0.50, 0.90, and 0.95, respectively, and performs cumulative maximum value correction according to the quantile from low to high to ensure... .

[0011] Furthermore, the predictive features include: the construction stage, progress ratio, workday attributes, day and night shifts, temperature, rainfall, total active power, apparent power, transformer load rate, maximum branch load rate, maximum distribution node load rate, number of active devices, and number of risk assets at the current time t; as well as historical load values, historical load rate values, and historical number of active devices at several sampling times prior to the current time t, and / or rolling statistical values ​​calculated based on the above historical time series data, wherein the rolling statistical values ​​include at least the mean, maximum value, and standard deviation within a previously set time window.

[0012] Furthermore, equipment i The uncalibrated expected active power load within the prediction window corresponding to time t is: ; in, For equipment i Rated power; The commissioning coefficient is used to characterize the equipment. i Is it in operation at time t? The load factor is used to characterize the equipment. i The expected load level at time t.

[0013] Further, the step of conserving and allocating the uncalibrated expected active load according to the total load constraints corresponding to the multiple active load quantile prediction values ​​includes: When the sum of the uncalibrated expected active loads is greater than a preset positive number ε, the allocation weight of each electrical device is as follows for each prediction time t and each quantile q: ; in, This represents the sum of the unspecified expected active power loads of all electrical equipment at time t. The load distribution of each electrical device at the quantile q is as follows: ; in, Indicates device i Distributed active load at time t and quantile q.

[0014] Furthermore, the step-by-step accumulation of information along the power supply path includes: For any distribution node n, the active power, reactive power of the equipment directly connected to the node, and the aggregated results of each sub-node are summed to obtain the equivalent active load of node n. and equivalent reactive load ; For any branch cable l The equivalent active load of cable l is obtained by summing the active and reactive power of the devices along the branch path that include the cable. and equivalent reactive load ; The step-by-step accumulation is performed in the order of the switch box to the secondary distribution box, the primary distribution box, and the transformer.

[0015] Furthermore, the preset capacity conditions include at least two risk level thresholds, and the load status is divided into normal status, warning status, and over-limit status according to the risk level thresholds; the bottleneck link is the power distribution node or branch cable where the load status is in the warning status or over-limit status.

[0016] Furthermore, the calculation of the required adjustment capacity includes: calculating the required reduction of apparent power and the required reduction of active power; The required reduction in apparent power is calculated using the following formula: ; in, The apparent capacity requirement of bottleneck component b at quantile q; The available capacity of bottleneck link b Target load rate; when At that time, the required reduction in active power for: ; in, Let be the current active power of bottleneck component b at quantile q, when hour, .

[0017] Furthermore, the adjustable capability includes the ability of the device to reduce power, the power reduction capability of device i Assign active load to it at present Minimum sustaining power (the minimum power required for a device to maintain standby or basic operation) difference.

[0018] A dynamic configuration system for temporary power capacity at construction sites based on construction schedule coupled with load forecasting is used to execute the aforementioned dynamic configuration method for temporary power capacity at construction sites based on construction schedule coupled with load forecasting, including: The data acquisition and standardization module is used to perform step S1. The feature and quantile prediction module is used to perform step S2. The equipment scenario and load allocation module is used to execute the S3 step; A radial topology aggregation module is used to perform step S4; The capacity verification module is used to perform step S5; and Configure the decision and status management module to execute steps S6 and S7.

[0019] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for dynamically configuring temporary power capacity at construction sites based on construction progress coupled with load prediction.

[0020] The beneficial effects of this invention are as follows: By predicting multiple active load quantile forecasts reflecting different levels of conservatism within a future set time window, this invention transforms the uncertainty of future construction power load into a multi-level capacity verification scenario, realizing graded capacity screening and early warning under different levels of conservatism; by using construction schedule data and on-site operation condition data to determine the operating status of each electrical device and the unmarked expected active load in the future period, and then using the total load quantile forecast as a constraint for conservation allocation, the total active power of each electrical device is kept consistent with the corresponding quantile forecast value, which avoids the problem of contradiction between total load forecast and equipment-level allocation, and makes the forecast results engineering-applicable at the equipment level; by converting the allocated load into capacity demand and grading it along the power supply path information from the end equipment side to the power supply side... The system uses a cumulative approach to accurately locate local bottlenecks in distribution boxes or cables when the total transformer load is within limits. It calculates the load status and identifies bottlenecks based on the rated capacity or allowable current carrying capacity of each power supply and distribution link. Verification can be completed using only standard parameters, eliminating the need for complex iterative calculations or additional sensor deployments. This high computational efficiency makes it suitable for rapid, rolling execution on-site. The system calculates the required adjustment capacity based on the current load status of the bottleneck link and the target load rate. It also generates capacity adjustment strategies, including peak shifting, transfer, or expansion, based on the priority and adjustability of downstream electrical equipment. Confirmed strategies are written into the configuration status of the next forecast cycle and updated and rolled over as construction progresses. This creates a traceable closed-loop management system for forecasting, verification, and capacity configuration, enabling dynamic and continuous optimization of temporary power capacity configuration as construction progresses. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention; Figure 3 This is a schematic diagram of the radial temporary power distribution topology and equipment mapping of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] like Figure 1 As shown, the method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction of the present invention includes the following steps: S1. Obtain multi-source data from the construction site, including construction schedule data, equipment parameter data, operating status data, historical electricity consumption data, on-site operating condition data, and power supply and distribution topology data; based on the power supply and distribution topology data, determine the power supply path information from each electrical device to the power supply source, and perform time alignment processing on the multi-source data to form a temporary electricity consumption dataset arranged according to a unified time scale.

[0024] In this invention: The construction schedule data refers to the time arrangement data used to guide the orderly development of various operations on the construction site, including at least the division of construction phases, the start and end times of each phase, the progress ratio, and the equipment entry and exit schedule. The equipment parameter data refers to a set of static or quasi-static data that characterizes the inherent electrical characteristics, rated capacity, and operation and maintenance attributes of each electrical device, including at least rated power, power factor, rated current, power supply guarantee priority, and starting impact coefficient. The operational status data refers to dynamic data that is collected in real time or reported in near real time, reflecting the current working status of electrical equipment and power supply and distribution systems at the construction site. It includes at least the equipment start-up and shutdown status, operating current, active power, and meter readings. The meter readings are used to provide raw electricity data, which are extracted and calculated to obtain the operating parameters such as operating current and active power, and serve as a source for accumulating historical electricity consumption data over time. The historical electricity consumption data refers to the time series data of load, electricity consumption and operating parameters that have occurred and been recorded over a period of time in the past, which is used to capture electricity consumption patterns and train prediction models. The power supply and distribution topology data refers to a set of data that describes the electrical connection relationships, hierarchical relationships, and access locations of power supply and distribution facilities such as transformers, distribution boxes, switch boxes, and cables in the temporary power supply system at the construction site. The on-site operation condition data refers to real-time or near-real-time dynamic factor data affecting the operation and load level of construction equipment, including at least one or more of the following: shift information, workday / holiday information, weather information, and concurrent operation scenario information; among them, concurrent operation scenario refers to a scenario in which multiple electrical equipment or processes operate simultaneously or overlappingly within the same time period, resulting in superimposed electrical load; weather information refers to the set of external meteorological factor data that affect the operation status, load level, and operating conditions of the power supply and distribution system during the dynamic configuration of temporary power capacity at the construction site, which may include temperature, rainfall, wind force level, extreme weather warning information, etc.

[0025] In this invention, specifically, the power supply path information is an ordered set of nodes from the power device access node through each level of parent nodes to the power source root node, and an ordered set of branch cables connecting adjacent nodes; the power supply path information is updated as the access location of the power device or the power supply and distribution topology changes.

[0026] After acquiring multi-source data from the construction site, power supply path information is generated for each device based on the device access nodes, parent-child relationships of distribution nodes, and upstream and downstream relationships of branches in the power supply and distribution topology data. The multi-source data is then time-aligned to form a temporary power consumption dataset arranged according to a unified time scale.

[0027] S2. Based on the current time and its historical time series information in the temporary power consumption dataset, determine the information for prediction, and predict multiple active load quantile prediction values ​​reflecting different levels of conservatism within a future set time window according to the information for prediction.

[0028] To fully reflect the current power consumption situation and environmental impact at the construction site, and to capture the time-dependent pattern of load changes as construction progresses, the information used for prediction in this invention includes: (i) Current moment characteristics: construction stage, progress ratio, workday attributes, day and night shifts, temperature, rainfall, total active power, apparent power, transformer load rate, maximum branch load rate, maximum distribution node load rate, number of active equipment and number of risk assets at current moment t; among them, the number of risk assets refers to the number of power supply and distribution links whose load rate at current moment t exceeds the preset warning threshold, which is used to characterize the overall risk exposure of the power supply system.

[0029] (ii) Historical time-series characteristics: historical load values, historical load rate values, and historical number of active devices at several sampling times prior to the current time t, and / or rolling statistical values ​​calculated based on the above historical time-series data. The rolling statistical values ​​include at least the mean, maximum value, and standard deviation within a set time window in the past. Preferably, the set time window is the past 1 hour, 4 hours, and 1 day.

[0030] The information used for prediction is derived from the temporary power consumption dataset generated in step S1. Features such as construction stage, progress ratio, workday attributes, shifts, temperature, and rainfall are directly read from multi-source data; total active power is directly read from operating status data, and apparent power is calculated from voltage and current or obtained directly from meters; transformer load rate, maximum branch load rate, and maximum distribution node load rate are calculated from the ratio of current apparent capacity to rated capacity / allowable current carrying capacity; the number of active devices is obtained by counting the number of devices currently in operation; the number of risky assets is obtained by counting the number of power supply and distribution links whose current load rate exceeds the warning threshold; historical load values, historical load rate values, and historical active device numbers are obtained by lag extraction of historical time-series data from the temporary power consumption dataset; and rolling statistics (mean, maximum, standard deviation) are calculated based on the aforementioned historical time-series data within a previously set time window.

[0031] In some embodiments, the active load quantile is obtained using any one or more combinations of the following methods: (1) Historical quantile statistical method: Select historical samples with the same construction stage, workday attribute, time period and scenario as the current time from the historical electricity consumption data, and count the active load quantile value of the historical samples in the future prediction window as the load quantile prediction value at the current time.

[0032] (2) Sliding window continuation method: Taking the current time t as the cutoff point, the empirical quantile value of the load sequence within the past set time window is extracted as the prediction result of the corresponding quantile value in the future prediction window.

[0033] (3) Hybrid prediction method: The constructed prediction features are input into the trained prediction model to obtain the first active load quantile prediction value; at the same time, the historical same-period quantile statistical method is used to obtain the second active load quantile prediction value; based on the similarity between the current time and the historical samples or the construction stage to which the current time belongs, the fusion weight of the first prediction value and the second prediction value is adaptively determined, and the fused active load quantile value is output.

[0034] In this invention, to fully leverage the ability of historical data to characterize load variation patterns while considering the complex impact of multiple factors coupled during the construction phase on future loads, preferably, the constructed prediction features are input into a trained prediction model, which outputs the active load quantile values ​​within the future prediction window starting from time t. The active load quantile values ​​include at least the 50th percentile predicted value P. 50 90th percentile predicted value P 90 And the 95th percentile predicted value P 95 , where P 50 P is used to characterize the normal level of future load. 90 P is used to characterize a higher level of future load. 95 This is used to characterize the capacity pressure of future loads in high-load ranges. By simultaneously outputting predicted values ​​at three different quantile levels, the uncertainty of future construction power load is transformed into a multi-level capacity verification scenario, providing input for subsequent capacity configuration under different levels of conservatism.

[0035] In some embodiments, quantile regression, gradient boosting regression, or time series prediction models are used as prediction models, and are trained or configured with quantile loss parameters of 0.50, 0.90, and 0.95, respectively, to output the corresponding active load quantile values.

[0036] In some embodiments, to ensure the engineering rationality of the quantile prediction results, the active load quantile values ​​output by the prediction model are non-negatively clipped and adjusted according to P... 50 P 90 P 95 The cumulative maximum value correction is performed in the order that it satisfies P. 50 ≤P 90 ≤P 95 Monotonicity constraints.

[0037] In some embodiments, to ensure the reliability and authenticity of the prediction model training and evaluation, and to prevent inflated prediction performance due to future information leakage, the future prediction window contains H sampling periods, where H ≥ 1; the future prediction label is defined as y(t) = max{P(t+1), P(t+2), ..., P(t+H)}, which is the maximum load value at each sampling time within the future prediction window. The label does not include the load at the current time t. The training set, validation set, and test set are divided in chronological order, and an isolation interval of no less than H sampling periods is set between adjacent data subsets to ensure that the future prediction window corresponding to any data subset does not cross over to another data subset. The uncertainty prediction model outputs active load quantile prediction values ​​with quantiles q of 0.50, 0.90, and 0.95, respectively, and performs cumulative maximum value correction according to the quantile from low to high to ensure... Monotonicity constraints.

[0038] S3. Based on the construction schedule data and on-site operation condition data, determine the operating status of each electrical device and the uncalibrated expected active load in the future period, and perform a conservation allocation of the uncalibrated expected active load according to the total load constraints corresponding to the multiple active load quantile prediction values ​​to obtain the allocated load of each electrical device.

[0039] Specifically, based on the construction phase divisions and equipment entry / exit schedules in the construction progress plan data, and the shift information, weekday / holiday information, weather information, and concurrent operation scenario information in the on-site operation condition data, the operating status and load factor of each electrical device in the future period are determined, and the uncalibrated expected active power load of the equipment is generated. The uncalibrated expected active power load of the equipment is the preliminary expected power of each electrical device under the condition of no total load constraint.

[0040] In this invention, the uncalibrated expected load of each electrical device can be determined in the following ways. For example, historical records with the same construction stage, workday attribute, and time period as the current moment can be filtered from the temporary power consumption dataset, and the average power of each electrical device in the historical record set can be calculated as the uncalibrated expected load; or, based on the construction schedule data and on-site operation condition data, the comprehensive operation coefficient of each device can be determined by looking up a table in the preset operation coefficient table, and the uncalibrated expected load can be obtained by multiplying it by the rated power.

[0041] However, the above methods have certain limitations in some application scenarios: the effectiveness of the historical data method depends on the sufficiency and representativeness of historical samples, and its accuracy will decrease when there are rapid changes during the construction phase or when there is a lack of historical precedents; the simplified table lookup method is based on the calculation of preset coefficients, and its adaptability is limited when the actual working conditions on site deviate significantly from the preset conditions. The above methods can also be implemented in scenarios where historical data is sufficient or the working conditions are stable.

[0042] To quantify the construction phase divisions and equipment entry / exit schedules in the construction progress plan data, along with shift information, weekday / holiday information, weather information, and concurrent operation scenario information in the on-site operation condition data, into the operational status and load level of each electrical device, so that the calculation of the expected load of the equipment reflects both the basic constraints of the construction plan and can adapt to changes in dynamic on-site conditions, this invention preferably defines the uncalibrated expected active power load of device i within the prediction window corresponding to time t as: ; in, Rated power of the equipment; This is the commissioning coefficient. This is the load factor. The commissioning factor is determined based on the equipment entry and exit schedule in the construction progress plan data and the shift information and workday / holiday information in the on-site operation condition data, and is used to characterize the equipment. i Is it in operation at time t (e.g., Indicates that it has been put into operation. (Indicates shutdown). The load factor is determined based on weather information and concurrent operation scenario information from the construction phase and the on-site operating conditions data. Specifically, the load factor is determined by the following formula: ; in, The construction stage to which time t belongs; For device i, specify the device category; The baseline load factor; This is a weather correction factor; This is a correction factor for concurrent operations.

[0043] Baseline load factor The coefficients can be obtained from the "Construction Stage - Equipment Category" two-dimensional coefficient table, representing the typical load-bearing level of this type of equipment in that construction stage. For example, in the main structure stage: tower cranes are 0.55, steel bar processing equipment is 0.70, concrete pumping equipment is 0.75, and temporary lighting is 0.40; in the decoration and finishing stage: tower cranes are 0.25, steel bar processing equipment is 0.15, decoration tools are 0.65, and temporary lighting is 0.60.

[0044] Weather correction factor Determined based on weather information. For example: 1.0 for sunny or cloudy weather; 0.6 to 0.8 for outdoor equipment during light rain; 0 for tower cranes, construction hoists, and other high-altitude equipment (i.e., deemed to be out of service) during moderate rain or winds of level 6 or above, and 1.0 to 1.2 for drainage equipment; 1.1 to 1.3 for ventilation and cooling equipment when the daily maximum temperature exceeds 35℃, and 0.8 to 0.9 for other equipment.

[0045] Concurrent operation correction factor Determined based on concurrent operation scenario information, this value characterizes the decrease in the actual load rate of a single device when multiple processes are performed simultaneously within the same time period. For example: when the number of concurrent processes in the same power distribution area is 1, the value is 1; when it is 2, the value is 0.9; when it is 3, the value is 0.85; and when it is 4 or more, the value is 0.8.

[0046] The initial values ​​of the above coefficient table are preset according to the approved temporary power supply organization design and equipment nameplate parameters. After the system is running, it is further calibrated based on historical power consumption data. That is, for each equipment category under the combination of construction stage s, weather category w, and concurrent scenario category c, the mean or specified quantile of the ratio of the measured active power to the rated power of the equipment in the historical samples under that condition is taken as the updated combination coefficient value. When the sample size is less than the preset number of records, the preset value is retained. To ensure the rationality of the project, the above results are cropped in the [0,1] interval. Through the independent configuration of the commissioning coefficient and the load coefficient, this calculation method can flexibly adapt to the differentiated operating characteristics of each equipment under different construction stages and site conditions.

[0047] After obtaining the uncalibrated expected active load of each electrical device, the uncalibrated expected load of the device is conservatively scaled according to the total load constraint corresponding to the multiple active load quantile prediction values ​​to obtain the device-level quantile load.

[0048] Specifically, for each prediction time t and each quantile q, it is first determined whether the sum of the uncalibrated expected active load of each electrical device is greater than the preset positive number ε.

[0049] When the sum of the uncalibrated expected active loads is greater than a preset positive number ε, the allocation weight of each electrical device is calculated according to the following formula: ; in, This represents the sum of the unspecified expected active power loads of all electrical equipment at time t. This weight reflects the equipment's... i The proportion of all equipment without specified expected load.

[0050] Then, the total load forecast value corresponding to the quantile q obtained in step S2 is used. According to the weight allocation Distributed to each electrical device, resulting in equipment-level quantile loads: ; in, This represents the distributed active power load of device i at time t and quantile q. This process ensures that the sum of the active power of all electrical devices remains consistent with the predicted value of the corresponding quantile, thus satisfying the conservation constraint. ; Meanwhile, the name, category, power factor, and power supply path information of each device are retained during the allocation process for use in subsequent steps. Through the above-mentioned conservation allocation, the engineering feasibility of the total load forecast results at the device level is ensured, while avoiding the problem of contradictions between the total load forecast and the device-level allocation.

[0051] In special cases: when the sum of the uncalibrated expected active loads is less than or equal to ε, it indicates that the total uncalibrated expected load of each electrical device at the current moment is too small (e.g., the equipment is in a state of large-scale shutdown or construction interval). In this case, directly allocating according to the weight formula will lead to an unreasonable amplification of the load allocated to each device, introducing a large error. Therefore, the historical power ratio of the devices under the same construction stage, time period, and scenario is used as the allocation weight; if an effective ratio still cannot be obtained (e.g., there is no historical record for reference in the same construction stage), the weight is determined by the rated power ratio of each device or by an equal-ratio allocation method, and the corresponding reasons are recorded for subsequent traceability and review.

[0052] Wherein, ε is a preset validity threshold, which is pre-set based on the total capacity of equipment and minimum metering accuracy at the construction site, and is used to determine whether the sum of uncalibrated expected loads has statistical validity.

[0053] S4. Based on the electrical parameters of each electrical device, the allocated load is converted into capacity demand, and the equivalent active load and equivalent reactive load of each power supply link in the power supply and distribution topology data are determined by accumulating the load step by step from the end device side to the power supply side along the power supply path information.

[0054] Specifically, based on the power factor of each electrical device, the allocated active load obtained in step S3 will be... Converted to reactive power: ; in, Let i be the power factor of device i; Let be the phase difference angle between the voltage and current of device i. Meanwhile, the apparent power of device i at the quantile q is: ; Then, the active power of each electrical device is... and reactive power Along the corresponding node path and branch path, respectively, the nodes are accumulated step by step from the leaf node to the transformer root node according to the radial topological parent-child relationship.

[0055] In some embodiments, the step-by-step accumulation along the power supply path information can be implemented in the following manners: (1) Full traversal method: for each power supply and distribution link, traverse all power consumption devices, and respectively accumulate the active power and reactive power of the devices whose power supply path includes this link; (2) Incremental update method: only for devices whose operation status or allocated load has changed, perform incremental correction on the existing accumulation result of each link along the power supply path. In the present invention, in order to make the accumulation direction consistent with the direction of the physical power supply path, so that each node only needs to process the aggregation result of its direct child nodes for upward transmission, without re-traversing the full device list at each level, preferably, for any distribution node n, sum the active power, reactive power of the devices directly connected to this node and the aggregation results of each child node respectively, to obtain the equivalent active load of node n and equivalent reactive load ; for any branch cable l , sum the active power and reactive power of the devices whose branch path includes this cable respectively, to obtain the equivalent active load of the cable l and equivalent reactive load . The step-by-step accumulation is performed according to a post-order traversal from the switch box to the secondary distribution box, the primary distribution box and the transformer, that is, starting from the leaf node (the last-stage distribution box or switch box) of the power supply and distribution topology, accumulating step by step upwards until reaching the power supply root node (transformer). This method makes full use of the intermediate results already aggregated by the child nodes, avoids repeatedly traversing the same device, and significantly reduces the computational complexity.

[0056] In the above steps, the active power P and the reactive power Q, as independent accumulation quantities, are respectively accumulated step by step along the power supply path, which ensures power conservation and provides a data basis for calculating the apparent capacity requirement of each distribution node and branch cable in step S5. For scenarios where only the equivalent power factor can be obtained, the apparent capacity can also be directly converted according to , where and are respectively the apparent capacity requirement and active power of the power supply and distribution link at the quantile q, is the equivalent power factor of this power supply and distribution link; is the equivalent power factor of this power supply and distribution link at the quantile q, that is, the phase difference angle between the equivalent terminal voltage and the equivalent total current.

[0057] S5. Calculate the load state of each power supply and distribution link according to the rated capacity or allowable current-carrying capacity of each power supply and distribution link, and identify bottleneck links with capacity risks according to the comparison result between the load state and a preset capacity condition.

[0058] ​Specifically, after obtaining the equivalent active and reactive loads of each distribution node and branch cable at the quantile q by accumulating them step by step from bottom to top along the power supply path through step S4, the load status of each power supply and distribution link is further calculated based on the rated capacity or allowable current carrying capacity of each power supply and distribution link.

[0059] For any distribution node in the power supply and distribution topology (including transformer nodes, distribution box nodes at all levels, and switch box nodes), its available capacity is determined as follows: ;or ; Among them, when the distribution node has rated capacity The former method is used when the distribution node only has the rated current. The latter form is used at that time. The rated voltage of the node (in V). The rated current of the node is expressed in amperes (A).

[0060] Apparent capacity demand of distribution node n under quantile q for: ; in, and These are the equivalent active load and equivalent reactive load of distribution node n at the quantile q, respectively, obtained by accumulating them step by step in step S4.

[0061] Load factor of distribution node n at quantile q and safety margin They are respectively: ; .

[0062] For any branch cable in the power supply and distribution topology, its current is calculated as follows: ; in, branch cables l The apparent capacity requirement at quantile q is obtained by accumulating along the branch path in step S4. This refers to the rated voltage (in volts) of the branch cable.

[0063] Branch cables l Available carrying capacity The allowable current carrying capacity is adjusted based on ambient temperature, laying method, and grouping conditions: ; in, This refers to the rated current carrying capacity of the branch cable under reference conditions. This is the temperature correction factor. This is a correction factor for the laying method. These are grouped correction factors. The above correction factors are determined based on the actual laying conditions at the construction site and current standards.

[0064] Branch cables l Load rate at quantile q and safety margin They are respectively: ; .

[0065] Branch cables l Equivalent available capacity calculated based on the revised allowable current carrying capacity for: ; To ensure consistency in load status assessment results across all power supply and distribution links, this invention uses load rate as the primary characterizing indicator for load status, and the load rate is uniformly calculated using apparent capacity (i.e., ...). The active power (kW) is not directly divided by the rated capacity (kVA) to avoid capacity misjudgment due to power factor differences.

[0066] In order to differentiate the capacity tension levels of different power supply and distribution links and provide a basis for the differentiated generation of subsequent capacity adjustment strategies, preferably, the preset capacity conditions include at least two risk level thresholds, which are used to classify the load status of power supply and distribution links into different risk levels.

[0067] In this invention, the risk levels include at least a normal state, a warning state, and an over-limit state: when or When this occurs, it is considered a normal state; when or When this occurs, it is determined to be a warning state; when or When this occurs, it is determined to be an over-limit state.

[0068] in, As the warning threshold, The threshold value is the one that exceeds the limit. The above risk level threshold is set as a configurable parameter in the configuration file and can be adjusted according to the safety margin, power supply level and management requirements of the construction site. It is used for algorithm screening rather than replacing engineering specifications.

[0069] For example, in one embodiment of the present invention, a load rate below 85% is defined as a normal state, 85% to less than 100% is defined as a warning state, and reaching or exceeding 100% is defined as an over-limit state. The above thresholds are for illustrative purposes only; in actual engineering projects, specific thresholds should be determined according to approved temporary power supply designs and current standards.

[0070] In this invention, the bottleneck refers to a power distribution node or branch cable whose load condition is determined to be in a warning state or an over-limit state. For P 50 P 90 and P 95 Three quantile scenarios are used to independently determine the load status and identify bottlenecks; when any power supply and distribution link is at P 95 When a scenario is determined to be in a warning state or an over-limit state, it will be identified as a potential bottleneck and a risk warning will be issued in advance.

[0071] Once the bottleneck is identified, record the asset type (distribution node or branch cable), asset number, quantile scenario, load rate, safety margin, and risk level of the bottleneck for use in the subsequent S6 step to generate capacity regulation strategies.

[0072] S6. Calculate the required adjustment capacity based on the current load status and target load rate of the bottleneck link, and generate a capacity adjustment strategy based on the power supply guarantee priority and adjustability of the downstream electrical equipment of the bottleneck link. The capacity adjustment strategy includes at least one or more of the following: load shifting measures, load transfer measures, or capacity expansion prompts.

[0073] Specifically, after identifying the bottleneck in step S5, for each bottleneck that is determined to be in a warning state or an over-limit state, the capacity reduction required to reduce its load rate to the target load rate is calculated, and a capacity adjustment strategy is generated based on the power supply guarantee priority and adjustability of downstream equipment of the bottleneck.

[0074] In this invention, calculating the required regulation capacity includes calculating the required reduction of apparent power and the required reduction of active power.

[0075] In some embodiments, the required reduction in apparent power can be achieved using a reverse calculation method based on the target capacity, a calculation method based on the load factor difference, or a calculation method based on the reduction ratio. All of these methods are based on the fundamental principle of the difference between the current capacity demand and the available capacity at the target load factor; the only difference lies in their mathematical expression.

[0076] In order to differentially determine the capacity adjustment target according to different risk levels of the bottleneck link, so that the adjustment intensity matches the severity of the risk — moderate adjustment is performed in the early warning state to restore the safety margin, stronger adjustment measures are taken in the over-limit state to quickly alleviate the overload risk, and meanwhile, invalid adjustment is avoided when the load is already within the safe range, the present invention preferably, the apparent power to be reduced is calculated by the following formula: For bottleneck link b, its current apparent capacity demand at quantile q is , the current active power is , and the available capacity is . Let the target load rate be , then the apparent power to be reduced is: ; wherein, is determined according to the risk level of the bottleneck link. Exemplarily, for the early warning state, the early warning threshold (e.g., 85%) can be taken as the target load rate; for the over-limit state, a review target value not greater than 1 (e.g., 95% or 100%) can be taken. When , , it means no adjustment is required.

[0077] In some embodiments, the active power to be reduced adopts a direct calculation method based on active power or a calculation method based on active load rate. In order to directly connect with the calculation of the apparent power reduction, maintain the consistency of the data flow, and at the same time avoid introducing the power factor as an additional parameter to simplify the calculation process, preferably, when , the active power to be reduced is: ; wherein, when , .

[0078] Determine the set of all electrical equipment downstream of the bottleneck link b. The downstream equipment refers to the electrical equipment that includes the bottleneck link in the power supply path, that is, all equipment whose access nodes are located downstream of the bottleneck link.

[0079] In the present invention, the adjustable capability includes the reducible power of the equipment.

[0080] The power supply priority and available power reduction for each electrical device in the downstream equipment set are obtained. The power supply priority is a pre-defined attribute value that indicates the importance of continuous power supply for the electrical device in the construction process and safety assurance. The higher the power supply priority, the later the device is selected as a target for peak shifting or power reduction. For example, the power supply priority can be divided into three levels: high, medium, and low. High-priority devices (such as main tower cranes, fire pumps, and critical welding equipment) are not included in the first round of peak shifting by default; medium-priority devices (such as general processing equipment) are included in peak shifting considerations when low-priority devices are insufficient; low-priority devices (such as auxiliary lighting and small tools) are given priority as peak shifting targets.

[0081] Based on the order of maximum to minimum power reduction capacity, priority should be given to devices with lower power supply security priority for peak-shifting. i Power reduction Assign active load to it at present With minimum sustaining power Difference: ; Accumulate the power reduction potential of the selected equipment until the accumulated power reduction potential meets the requirements. This forms a set of peak-shifting devices. If the cumulative power reduction capacity of low-priority devices is insufficient, then medium-priority power supply devices are added.

[0082] When there are candidate loops of the same level and the candidate loops have remaining capacity margins, load transfer measures can be generated.

[0083] Specifically, load transfer requires the following conditions to be met simultaneously: (1) The voltage level of the candidate circuit is the same as the voltage level of the circuit where the bottleneck link is located; (2) The distribution level of the candidate circuit is the same as that of the circuit where the bottleneck link is located (such as different secondary distribution box branches under the same primary distribution box). (3) After the candidate circuit takes over the transferred load, its load rate does not exceed the target load rate. That is, the remaining available capacity of the candidate circuit satisfies: ; in, The remaining available capacity of candidate loop j at the target load rate: ; in, Let j be the available capacity of candidate loop j. Let j be the current apparent capacity requirement of the candidate loop.

[0084] When all the above conditions are met, the target loop and the transfer capacity (not exceeding the transferable load of the remaining available capacity) are determined, and a load transfer measure is formed. When there are multiple candidate loops, the candidate loop with the largest remaining available capacity or the most balanced load rate after transfer is preferentially selected as the target loop.

[0085] When a capacity gap still exists after the implementation of peak-shifting and transfer measures, that is: ; in, To determine the active power reduction measures, the target available capacity of the bottleneck is determined, and expansion suggestions are generated.

[0086] Target available capacity Calculate using the following formula: ; The capacity expansion notification includes asset information for bottleneck areas, current capacity requirements, target available capacity, and expansion suggestions. The expansion notification does not automatically change the formal temporary power topology; it requires manual review and confirmation by the on-site safety officer in conjunction with the approved temporary power organization design.

[0087] The capacity configuration results should record at least the following information: prediction period; quantile scenario used (P... 50 / P 90 / P 95 The following information is required: the type, number, and risk level of the bottleneck asset; the current load rate and the target load rate; the required reduction in active power; the set of off-peak equipment and the off-peak periods for each equipment; when the transfer conditions are met, the target loop and transfer capacity are also recorded; when there is still a capacity gap, the target available capacity and expansion prompts are also recorded.

[0088] S7. Output the capacity adjustment strategy, write the confirmed capacity adjustment strategy into the configuration status of the next prediction cycle, and execute subsequent predictions, verifications and configurations in a rolling manner as construction progress and operation data are updated.

[0089] After the capacity configuration result is generated in step S6, it is sent to the temporary power management terminal for display. The temporary power management terminal displays the bottleneck location, calculation basis, and capacity configuration result, which is confirmed by on-site management personnel in conjunction with construction procedures, equipment safety requirements, and the approved temporary power organization design. If the candidate equipment cannot be staggered or the candidate transfer circuit does not meet the remaining capacity requirements, the system further outputs the target available capacity and a manual review prompt, without automatically changing the formal temporary power topology.

[0090] The confirmed capacity configuration results are written to the configuration status table and used as the target for equipment plan adjustments, candidate path constraints, or capacity verification in the next forecast cycle, thereby achieving closed-loop management of forecasting, verification, and capacity configuration. Subsequently, the system updates as construction progress and operational data are updated, continuously executing subsequent forecasts and verifications.

[0091] This invention also provides a dynamic configuration system for temporary power capacity at construction sites based on construction progress coupled with load forecasting, used to execute the above-mentioned dynamic configuration method for temporary power capacity at construction sites based on construction progress coupled with load forecasting, including: The data acquisition and standardization module is used to perform step S1. The feature and quantile prediction module is used to perform step S2. The equipment scenario and load allocation module is used to execute the S3 step; A radial topology aggregation module is used to perform step S4; The capacity verification module is used to perform step S5; and Configure the decision and status management module to execute steps S6 and S7.

[0092] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction.

[0093] This invention effectively avoids evaluation bias caused by current values ​​being mixed with future labels and cross-dataset label overlap by setting future prediction labels starting from time t and establishing isolation intervals of prediction window length between the training set, validation set, and test set. This makes the prediction results more realistically reflect the load of future periods, improving the reliability and engineering credibility of the predictions. Furthermore, by simultaneously outputting P... 50 P 90 and P 95 The load forecast values ​​at three different quantile levels transform the uncertainty of future construction power load into a multi-level capacity verification scenario—P 50 Used to assess normal load levels, P 90 For capacity screening at higher load levels, P 95This system is used for capacity pressure early warning in high-load areas, suitable for advance screening and early warning of temporary power supply during construction. Utilizing the construction phase divisions and equipment entry / exit schedules in the construction progress plan, combined with information on shifts, workdays, weather, and concurrent scenarios in the on-site operating conditions, it generates unspecified expected loads for equipment. Then, it performs conservation scaling constrained by the total load quantile prediction value, ensuring that the sum of active power of each electrical device is consistent with the corresponding quantile prediction value. This avoids contradictions between total load prediction and equipment-level allocation, and makes the prediction results engineering-applicable at the equipment level. By generating node paths and branch paths from the access node to the transformer root node for each device, and accumulating the equivalent active and reactive loads of each device from bottom to top according to the radial topology, it can generate the total expected load of the transformer. When the load is not exceeded, the system continues to locate local bottlenecks in distribution boxes or cables, breaking through the limitations of traditional methods that only focus on the total capacity of transformers, and achieving precise location of capacity bottlenecks. Capacity verification only requires conventional parameters such as rated capacity, rated current, allowable current carrying capacity, and equipment power factor of each power supply and distribution link, without the need for complex iterative calculations or additional sensor deployments. It has high computational efficiency and is suitable for rapid execution on edge computing terminals at actual construction sites. By associating the capacity configuration results with downstream equipment of bottleneck assets, power supply priority, and circuit margin, and writing them into the configuration status table of the next prediction cycle after confirmation by on-site management personnel, prediction, physical verification, and capacity configuration form a traceable closed-loop management, realizing dynamic and continuous optimization of temporary power capacity configuration as construction progresses.

[0094] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The specific implementation methods of S1 to S7 are first given according to S101 to S108, and then the calculation process is illustrated through a quantile prediction and capacity configuration example; wherein the risk threshold, equipment parameters, time window, and model parameters can all be configured according to the engineering project and are not intended to limit the scope of protection of the present invention.

[0095] As a specific implementation of steps S1 to S7, such as Figure 1 and Figure 2 As shown, the temporary power management device receives construction schedule data, equipment ledgers, equipment operating status, smart meter load data, weather data, and shift data. It also reads transformer, distribution box, branch cable, and equipment connection relationships. In one embodiment, the smart meters upload transformer master meter and key distribution box data at 15-minute intervals; the equipment management system updates the equipment plan when equipment enters or leaves the site or when construction procedures change. The temporary power management device includes a processor, memory, and communication interfaces. The memory stores prediction models, equipment path mappings, capacity parameters, and configuration status tables. The communication interfaces are connected to the smart meters, the equipment management system, and the temporary power management terminal.

[0096] S101: Data standardization and quality inspection.

[0097] Using equipment number, node number, branch number, timestamp, and construction stage number as association keys, and unifying rated power as kW, apparent power as kVA, current as A, and voltage as V, the system sequentially checks for time continuity, primary key duplication, equipment mapping integrity, existence of upstream and downstream nodes of branches, topological acyclicity, and unique transformer root node. The handling method for any missing values ​​is recorded, and the original input file remains read-only. For each device, a node path to the transformer root node is generated based on the access node and parent-child relationships, and a branch path is generated using the branch numbers between adjacent nodes; the path mapping is updated as the device access location or power distribution topology changes.

[0098] S102: Construct dataset samples.

[0099] Assuming a sampling interval of 15 minutes and a prediction window of 1 hour, then H=4. Based on the construction stage, progress ratio, workday, day / night shift, temperature, rainfall, total active power, apparent power, transformer load rate, maximum branch load rate, maximum distribution node load rate, number of active devices, and number of risk assets at time t, 1-, 4-, 8-, and 96-step lag values ​​for total active power, transformer load rate, and number of active devices, as well as the mean, maximum, and standard deviation over the past 1 hour, 4 hours, and 1 day, are constructed. Rolling statistics are first performed with an overall lag of 1 step, followed by rolling calculations.

[0100] The equipment combination characteristics during the construction phase include the number of high, medium, and low power supply priority equipment, the total rated power, the total expected power, and the total expected power of the phase plan. The target value is taken as the maximum value among P(t+1) to P(t+4). The last four time steps of the dataset are not included in the training and evaluation because the future window is incomplete.

[0101] S103: Training and Uncertainty Prediction.

[0102] The samples were divided into training, validation, and test sets in chronological order, with proportions of 60%, 20%, and 20%, respectively. Four isolation intervals were set at each of the two split boundaries. Numerical features were imputed using the median of the training set statistics, while categorical features were imputed to "unknown" and one-hot encoded.

[0103] Load quantiles can be obtained through historical period statistics, moving window quantile statistics, quantile regression, gradient boosting quantile prediction, time series prediction models, or combinations of the above methods. This invention does not limit the specific prediction model; the prediction model is only used to output the active load quantile and / or reactive load quantile for the same metering object, the same sampling granularity, and the same prediction period within a future configuration cycle. For ease of verification, this embodiment uses a histogram-based gradient boosting regressor, training three models with quantile loss parameters of 0.50, 0.90, and 0.95 respectively; continuous prediction and historical median prediction for the construction phase—week—15-minute period are also set as controls. If quantile crossover occurs at the same time, it is calculated according to P... 50 P 90 P 95 The cumulative maximum value correction is performed sequentially.

[0104] Among them, P 50 This represents the 50th percentile predicted value of active power load within the future configuration period, used to characterize the normal load level; P 90 This represents the 90th percentile forecast value, used to characterize higher load levels; P 95 This represents the 95th percentile predicted value, used to characterize capacity pressure near the high-load range. P 95 Not equivalent to the maximum load value, the actual peak value can be higher than P. 95 The above percentile values ​​are used as capacity verification inputs and are not directly equivalent to the target available capacity or recommended configuration capacity.

[0105] To ensure the consistency of quantile prediction results in engineering applications, the active load quantile values ​​under the same measurement object, the same sampling granularity, and the same prediction period should meet the requirement of P. 50 ≤P 90 ≤P 95 When the predicted output does not satisfy this monotonic relationship, perform monotonicity correction. If P appears in the validation data... 50 If the value is consistently higher than the actual peak value in the same period, it is determined to be an anomaly in metering caliber, inverse normalization, or time alignment, and needs to be rechecked before entering the capacity verification process.

[0106] S104: Generate device scenario load and perform conservation allocation.

[0107] The system extracts equipment operating probabilities, load factors, and simultaneity factors from the construction phase planning data. It then combines this with data on weekdays, day / night shifts, rainfall, and high-concurrency scenarios to determine the equipment deployment set, forming the uncalibrated expected load for each device. For each prediction time and each quantile, the equipment weight is calculated, scaling the total predicted load down to the device level. This process ensures that the sum of the equipment's active power aligns with the corresponding quantile prediction value, while retaining the name, category, power factor, and topology path of each device.

[0108] S105: Aggregate radial topology and verify capacity.

[0109] like Figure 3 As shown, parent-child relationships are constructed based on the parent node field of the node table, and the branch direction is determined based on the upstream and downstream nodes of the branch table. Node paths and branch paths from the switch box to the transformer are generated for each device.

[0110] For each device i, the active load corresponding to the quantile q When known, based on the equipment power factor Determine the reactive component When the reactive power has already been provided by on-site metering, the reactive power component is directly used. The system will... and The equivalent active load of any node or branch is obtained by accumulating the loads step by step along the corresponding node path and branch path. and equivalent reactive load and according to Calculate apparent capacity requirements. For scenarios where only the equivalent power factor is obtained, it can also be calculated as follows: Conversion; of which, This is the equivalent power factor of the power supply and distribution link; This is the equivalent power factor of the power supply and distribution link at the quantile q. From this, the load factor and safety margin of the transformer, each level of distribution box, and all branch cables can be obtained simultaneously in P50, P90, and P95 scenarios.

[0111] The load factor is calculated based on apparent capacity, i.e. The maximum load rate during the test period is ,in, The apparent capacity requirement at time t within the test period. This refers to the rated capacity of the power supply and distribution system. All formulas above use apparent capacity as the calculation caliber, and do not directly divide the active power (kW) by the rated capacity (kVA). Risk level thresholds are set in the configuration file. For example, this embodiment verifies the algorithm, defining a load rate below 85% as normal, 85% to less than 100% as a warning, and reaching or exceeding 100% as exceeding limits. These thresholds are only used for risk classification in this embodiment; actual projects should determine them according to approved temporary power supply designs and current standards.

[0112] S106: Generate capacity configuration results and update configuration status.

[0113] For each warning or bottleneck exceeding the limit, according to and Calculate the power reduction required to return to the target load rate, iterate through downstream equipment of the asset, prioritizing equipment with low power supply guarantee priority, then selecting equipment with medium power supply guarantee priority. For candidate circuits of the same level, determine the target circuit and transfer capacity only when the voltage level, distribution level, and remaining capacity at the target load rate meet the set conditions; if a capacity gap still exists, proceed according to... Determine the target available capacity; if the warning status still cannot be adjusted after the above judgment, manual intervention by the on-site safety officer is required. The system generates a structured capacity configuration result, but does not automatically change the formal topology. High power supply priority devices are not included in the first round of peak shaving by default.

[0114] The temporary power management terminal displays the bottleneck location, calculation basis, and capacity configuration results. After confirmation by on-site management personnel, the device writes the set of peak-shifting equipment and the target time period into the equipment plan correction field, and writes the target circuit and transfer capacity into the candidate path constraint field, or writes the target available capacity into the capacity verification field, thereby forming the configuration status for the next forecast cycle. In the next forecast cycle, the updated construction stage, equipment status, metering data, and the configuration status are read, and forecasting and capacity verification are re-executed.

[0115] S107, initiate additional impact verification.

[0116] S107 is an optional step for motor-driven equipment such as tower cranes, construction hoists, and water pumps. It allows for the activation of an additional starting impact check. The branch peak current is the sum of the stable operating current and the maximum starting increment of the downstream critical equipment in that branch. ; in, branch road l peak current, branch road l Stable operating current, For equipment i The starting impact factor (usually 4 to 8 times the rated current, the specific value is determined according to the parameters on the equipment nameplate). For equipment i The rated current. This implementation method assumes only the startup of one critical device in the same branch and does not simultaneously sum the startup currents of all devices. The startup impact verification result is output as a capacity verification prompt, separately from the stable operating load rate.

[0117] S108, System Deployment and Result Traceability.

[0118] The data acquisition and standardization module can receive data from smart meters and the project management system via file interfaces, database interfaces, or message interfaces. Each functional module can be deployed on field edge computing terminals, project servers, or cloud servers. The configuration decision and status management module sends structured results to web terminals, mobile terminals, or on-site large screens and receives confirmation results. Each risk record must include at least the prediction time, quantile, asset type, asset number, load rate, safety margin, target load rate, required power reduction, set of peak-shifting equipment, target loop, transfer capacity, target available capacity, confirmation status, and project boundary indications to enable result traceability.

[0119] Example 1: Simulation Verification Based on Public Datasets

[0120] This embodiment uses construction site temporary power simulation data based on a publicly available dataset for verification, including publicly available industrial load, non-residential building auxiliary load, and weather data. This data is then combined with construction stage, equipment rated parameters, commissioning rules, and radial temporary power distribution topology to generate a verification dataset. The data does not represent the actual measured load at a real construction site.

[0121] In this embodiment, the construction phase involves the intersection of the main structure and secondary structure. The power distribution topology has a single transformer root node, including 26 devices, 18 branches, and 19 nodes. The details of each branch are shown in the table below. The capacity status threshold is set as: predicted P 95 Capacity strain is indicated when the load factor reaches 90% or higher, or the maximum load factor reaches 95% or higher. (Predicted P) 95 A warning is issued when the load rate reaches 80% or higher but not 90%, or when the maximum load rate reaches 90% or higher but not 95%.

[0122] Table 1 Branch parameters of the embodiment

[0123] The data sampling interval is 15 minutes, the prediction window is 1 hour, and the test set is the data from the last two days of the dataset. The operational sets within the test period are: L1 includes tower cranes, construction hoists, and temporary pump sets; L2 includes D1 to D5 steel bar processing and auxiliary equipment; L3 includes concrete pumping equipment, mixing auxiliary equipment, and vibration equipment; and L4 includes temporary lighting, office area loads, and candidate loads for small machinery.

[0124] For each branch, the n=192 active load samples within two days are sorted in ascending order as x(1), x(2), ..., x(n), and the quantile position is determined by h=1+(n-1)q; if h=j+γ, then Pq=(1-γ)x(j)+γx(j+1). Therefore, when q=0.50, h=96.5, when q=0.90, h=172.9, and when q=0.95, h=182.45.

[0125] Table 2. Actual statistical results for the last two days of the test set.

[0126] As shown in Table 2, each power supply branch satisfies the condition that the actual peak value is ≥ the actual P. 95 ≥ Actual P 90 ≥ Actual P 50 The statistical relationship indicates that the quantile statistics in the test data conform to the general law of load distribution and there is no P-value. 50 The caliber is abnormally high, consistently exceeding the actual peak value.

[0127] Taking the branch road in the L2 rebar processing area as an example, the sorted key sampling values ​​are x(96)=42.3kW, x(97)=42.9kW, x(172)=68.4kW, x(173)=68.95kW, x(182)=75.9kW, and x(183)=77.0kW. From this, the actual P is obtained. 50 =0.5×42.3+0.5×42.9=42.6kW; Actual P 90 =0.1×68.4+0.9×68.95≈68.9kW; Actual P 95 =0.55×75.9+0.45×77.0≈76.4kW. The actual peak value of L2 during the test period was 82.1kW, which is consistent with the actual situation.

[0128] Table 3. Consistency Evaluation Results of Quantile Predictions

[0129] In Table 3, P 50 Relative error is used to evaluate the ability of prediction results to characterize normal construction loads, P 95 Relative error is used to evaluate the ability of prediction results to characterize high-load regions; P 95 Coverage representation predicts P 95 Not lower than the actual P within the same test cycle 95 This allows it to serve as a conservative input for capacity warning calculations. Predicted P for each branch. 50 The relative errors are all less than 4%, and the predicted P 95 All are not lower than the corresponding actual P 95 This can meet the input requirements for subsequent capacity status determination.

[0130] Taking L2 branch as an example, predict P 50 The predicted power is 43.8 kW. 95 It is 78.1 kW. P 50 The absolute error is |43.8-42.6|=1.2kW, and the relative error is 1.2 / 42.6×100%=2.82%; P95 The absolute error is |78.1-76.4|=1.7kW, and the relative error is 1.7 / 76.4×100%=2.23%. Predicted P 50 Used to reflect normal load levels and predict P 95 Used to reflect capacity pressure in high-load areas, neither of which is directly output as the target available capacity.

[0131] Table 4. Capacity Conversion and Configuration Status Determination Results

[0132] In Table 4, the predicted P 95 The conversion represents the prediction of P. 95 / cos The apparent capacity and its proportion of the rated capacity are obtained; the actual peak value is converted to the actual peak value / cos²t during the test period. The peak apparent capacity and maximum load factor are obtained. Taking L2 as an example, the predicted P... 95 The corresponding apparent capacity is 78.1 / 0.84 = 92.98 kVA, and the predicted P 95 The load factor is 92.98%; the peak apparent capacity corresponding to the actual peak of 82.1kW is 82.1 / 0.84=97.74kVA, and the maximum load factor is 97.74 / 100×100%=97.74%, which is 97.7% to one decimal place.

[0133] In this embodiment, P is predicted. 95 When the load rate reaches 90% or higher, or the maximum load rate during the test period reaches 95% or higher, the corresponding branch will be included in the capacity-constrained category; predict P 95 When the load rate reaches 80% or higher but less than 90%, or the maximum load rate reaches 90% or higher but less than 95%, the system is included in the early warning list. This threshold is only for engineering demonstration purposes in this embodiment and can be adjusted according to the safety margin, power supply level, and management requirements of the construction site. As shown in Table 4, L1 and L2 enter a capacity shortage state, L3 enters an early warning state, and L4 remains in a normal state. Subsequent embodiment 2 will select one of the capacity-strained branches to illustrate the dynamic configuration handling process.

[0134] Example 2: Dynamic configuration handling of local branch capacity shortages

[0135] This embodiment follows the capacity determination results of Embodiment 1, selecting the L2 rebar processing area branch to illustrate the dynamic configuration process when local capacity is tight. To avoid merely providing a disposal conclusion, this embodiment further provides the load decomposition of downstream equipment at the L2 level, as shown in Table 5. Equipment D1 to D5 are initially connected to the L2 branch, and their active and reactive load components are accumulated step by step along the path of "equipment access point—L2 secondary distribution box—upper-level distribution node". The system forms equivalent active and reactive loads at each node, and then calculates the corresponding apparent capacity demand.

[0136] The initial commissioning set is {D1, D2, D3, D4, D5}, where D1 and D2 are high-priority power supply devices and are not included in the first round of peak-shaving; D4 is capable of peak-shaving operation, and D5 is capable of being transferred to L4 or a backup branch. The rated capacity of the candidate receiving branch L4 is 80kVA, and the predicted P before adjustment is... 95 With an apparent capacity of 49.2 kVA and a load factor of 61.5%, L2 has sufficient margin to handle D5. Therefore, L2 is marked as a capacity-critical device and enters the configuration processing flow.

[0137] Table 5. Downstream Equipment Level P95 Load and Configuration Strategy for L2 Branch

[0138] Table 5 shows that the predicted P of L2 before adjustment 95 The active load is P L2 =21.0+18.5+15.6+16.0+7.0=78.1kW, reactive load is Q L2 =13.2+12.2+10.0+12.0+3.1=50.5kvar. Corresponding apparent capacity S L2 = The load factor is approximately 93.0 kVA, and the load factor is 93.0 / 100×100%=93.0%, which exceeds the 90% capacity stress threshold set in this embodiment.

[0139] The system does not prioritize cutting power supply protection for higher-priority devices D1 and D2. Instead, it searches for devices that can be shifted or transferred during off-peak hours among lower-priority devices. D4 consists of threading and welding auxiliary equipment, which can be adjusted to operate during off-peak hours; D5 consists of small tools and temporary lighting, with a smaller load, and can be transferred to L4 or the backup branch. This selection reflects the constraint effect of power supply protection priority on the devices that can be shifted or transferred during off-peak hours.

[0140] After configuration, L2 retains D1, D2, and D3, and the adjusted active load P' L2 =21.0+18.5+15.6=55.1kW, reactive load Q' L2=13.2+12.2+10.0=35.4kvar, apparent capacity S' L2 = The load factor drops to 65.5%, approximately 65.5 kVA. After D5 is switched to L4, the predicted active load of L4 (P95) increases from 45.3 kW to 52.3 kW. After adding the reactive power of D5 (3.1 kvar), the corresponding apparent capacity is approximately 56.9 kVA. The load factor of L4 is 56.9 / 80×100%=71.1%, which is still below the warning threshold.

[0141] Table 6 Comparison of branch capacity status before and after dynamic configuration

[0142] As shown in Table 6, through peak shifting at D4 and transfer at D5, the load rate of the originally strained L2 branch decreased from 93.0% to 65.5%; the load rate of the L4 branch, which took over the transferred load, increased from 61.5% to 71.1%, remaining within the normal range. This process did not transfer the risk to other branches, nor did it reduce the supply of critical processing equipment with high power supply priority.

[0143] The capacity configuration results include trigger branch L2, triggering basis, equipment set {D4, D5}, target off-peak period for D4, target transfer branch for D5, and adjusted capacity status. After the temporary power management terminal displays the results, on-site management personnel confirm them in conjunction with construction procedures, equipment safety requirements, and the approved temporary power organization design. After confirmation, the target period, target branch, and equipment set are written into the configuration status for the next forecast cycle. If candidate equipment cannot be off-peaked or candidate transfer branches do not meet the remaining capacity requirements, the system further outputs the target available capacity and a manual review prompt, without automatically changing the formal temporary power topology.

[0144] This invention is used for dynamic screening and configuration assistance of temporary power capacity risks at construction sites. It does not replace formal temporary power supply organization design, short-circuit current calculation, protection settings and coordination, grounding, leakage current protection device selection, and on-site acceptance. In practical applications, equipment nameplate parameters, construction phase plans, power distribution single-line diagrams, cable specifications, and on-site metering data can be replaced by configuration files or data tables without changing the prediction, conservation allocation, radial aggregation, capacity verification, and closed-loop handling of this invention.

Claims

1. A method for dynamically configuring temporary power capacity at construction sites based on construction progress coupled with load prediction, characterized in that, Includes the following steps: S1. Obtain multi-source data from the construction site, including construction schedule data, equipment parameter data, operating status data, historical electricity consumption data, on-site operating condition data, and power supply and distribution topology data; based on the power supply and distribution topology data, determine the power supply path information from each electrical device to the power supply source, and perform time alignment processing on the multi-source data to form a temporary electricity consumption dataset arranged according to a unified time scale. S2. Based on the current time and its historical time series information in the temporary power consumption dataset, determine the information for prediction, and predict multiple active load quantile prediction values ​​reflecting different levels of conservatism within a future set time window according to the information for prediction. S3. Based on the construction schedule data and on-site operating condition data, determine the operating status of each electrical device and the uncalibrated expected active power load in the future time period, and perform a conservation allocation of the uncalibrated expected active power load according to the total load constraints corresponding to the multiple active power load quantile prediction values, to obtain the allocated load of each electrical device; wherein: equipment i The uncalibrated expected active power load within the prediction window corresponding to time t is: ； in, For equipment i Rated power; The commissioning coefficient is used to characterize the equipment. i Is it in operation at time t? The load factor is used to characterize the equipment. i The expected load level at time t; The step of conserving and allocating the uncalibrated expected active load according to the total load constraints corresponding to the multiple active load quantile prediction values ​​includes: When the sum of the uncalibrated expected active loads is greater than a preset positive number ε, the allocation weight of each electrical device is as follows for each prediction time t and each quantile q: ； in, This represents the sum of the unspecified expected active power loads of all electrical equipment at time t. The load distribution of each electrical device at the quantile q is as follows: ； in, Indicates device i Distributed active load at time t and quantile q; S4. Based on the electrical parameters of each electrical device, the allocated load is converted into capacity demand, and the equivalent active load and equivalent reactive load of each power supply link in the power supply and distribution topology data are determined by accumulating the load step by step from the end device side to the power supply side along the power supply path information. S5. Calculate the load status of each power supply and distribution link based on the rated capacity or allowable current carrying capacity of each power supply and distribution link, and identify bottleneck links with capacity risks based on the comparison results between the load status and the preset capacity conditions. S6. Calculate the required adjustment capacity based on the current load status and target load rate of the bottleneck link, and generate a capacity adjustment strategy based on the power supply guarantee priority and adjustability of the downstream electrical equipment of the bottleneck link. The capacity adjustment strategy includes at least one or more of the following: load shifting measures, load transfer measures, or capacity expansion prompts. S7. Output the capacity adjustment strategy, write the confirmed capacity adjustment strategy into the configuration status of the next prediction cycle, and execute subsequent predictions, verifications and configurations in a rolling manner as construction progress and operation data are updated.

2. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The construction schedule data includes the division of construction phases, the start and end times of each phase, the progress ratio, and the equipment entry and exit schedule; the equipment parameter data includes rated power, power factor, rated current, power supply priority, and start-up impact coefficient; the operating status data includes equipment start-up and shutdown status, operating current, active power, and meter readings; and the on-site operating condition data includes shift information, weekday / holiday information, weather information, and concurrent operation scenario information.

3. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The power supply path information is an ordered set of nodes from the power device access node through each level of parent nodes to the power root node, and an ordered set of branch cables connecting adjacent nodes; the power supply path information is updated as the access location of the power device or the power supply and distribution topology changes.

4. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The multiple active load quantile prediction values ​​include: the 50th percentile prediction value P, which characterizes the normal load level. 50 The 90th percentile predicted value P, representing a higher load level 90 And the 95th percentile predicted value P, which characterizes the capacity pressure in the high-load zone. 95 ;P 50 P 90 and P 95 satisfy Monotonicity constraints.

5. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 4, characterized in that, The prediction of the plurality of active load quantile prediction values ​​includes: Let the sampling interval be Δt, and the future set time window contain H sampling periods. The future prediction label is defined as y(t)=max{P(t+1),P(t+2),...,P(t+H)}, where H is the number of sampling periods contained in the future set time window. The label does not include the load at time t. The training set, validation set, and test set are divided in chronological order, and isolation intervals of no less than H sampling periods are set between adjacent data subsets; The uncertainty prediction model outputs active load quantile predictions with quantiles q of 0.50, 0.90, and 0.95, respectively, and performs cumulative maximum value correction according to the quantile from low to high to ensure... .

6. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1 or 2, characterized in that, The information used for prediction includes: The current time t includes the construction stage, progress ratio, workday attributes, day and night shifts, temperature, rainfall, total active power, apparent power, transformer load rate, maximum branch load rate, maximum distribution node load rate, number of active devices, and number of risk assets; as well as the historical load values, historical load rate values, and historical number of active devices at several sampling times prior to the current time t, and / or rolling statistics calculated based on the above historical time series data, wherein the rolling statistics include at least the mean, maximum value, and standard deviation within a previously set time window.

7. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The stepwise accumulation of information along the power supply path includes: For any distribution node n, the active power, reactive power of the equipment directly connected to the node, and the aggregated results of each sub-node are summed to obtain the equivalent active load of node n. and equivalent reactive load ; For any branch cable l The equivalent active load of cable l is obtained by summing the active and reactive power of the devices along the branch path that include the cable. and equivalent reactive load ; The step-by-step accumulation is performed in the order of the switch box to the secondary distribution box, the primary distribution box, and the transformer.

8. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The preset capacity conditions include at least two risk level thresholds, and the load status is divided into normal status, warning status and over-limit status according to the risk level thresholds; the bottleneck link is the power distribution node or branch cable where the load status is in the warning status or over-limit status.

9. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The required adjustment capacity for calculation includes: calculating the required reduction of apparent power and the required reduction of active power; The required reduction in apparent power is calculated using the following formula: ； in, The apparent capacity requirement of bottleneck component b at quantile q; The available capacity of bottleneck link b Target load rate; when At that time, the required reduction in active power for: ； in, Let be the current active power of bottleneck component b at quantile q, when hour, .

10. The method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in claim 1, characterized in that, The adjustable capability includes the ability of the device to reduce power, the device i Power reduction Assign active load to it at present With minimum sustaining power difference.

11. A dynamic configuration system for temporary power capacity at construction sites based on construction progress coupled with load forecasting, used to execute the dynamic configuration method for temporary power capacity at construction sites based on construction progress coupled with load forecasting as described in any one of claims 1 to 10, characterized in that, include: The data acquisition and standardization module is used to perform step S1. The feature and quantile prediction module is used to perform step S2. The equipment scenario and load allocation module is used to execute the S3 step; A radial topology aggregation module is used to perform step S4; The capacity verification module is used to perform step S5; and Configure the decision and status management module to execute steps S6 and S7.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method for dynamic configuration of temporary power capacity at construction sites based on construction progress coupled with load prediction as described in any one of claims 1 to 10.

Images