A construction site energy consumption monitoring system

Abstract

The application discloses a construction site energy consumption monitoring system, comprising: an energy consumption monitoring unit collects electric energy parameters and water consumption parameters of a construction site; a presence sensing unit detects the number of personnel in each area of the construction site in real time; an edge gateway determines the personnel situation in each area according to the number of personnel in each area, and when a certain area is unoccupied and the unoccupied duration reaches a threshold value, generates a power-off instruction and delivers the power-off instruction to the corresponding execution unit of the area; the execution unit is used for executing power-off operation of the corresponding area according to the power-off instruction; a cloud platform predicts the future energy consumption of each area according to the energy consumption data, personnel data, construction plan and weather data of the construction site; according to the energy consumption data, threshold value anomaly detection, model prediction deviation anomaly detection, electrical feature anomaly detection and water consumption anomaly detection are carried out, and then energy-saving optimization suggestions are generated. The application improves the energy consumption data collection efficiency, reduces energy consumption waste, and reduces construction safety hazards through multi-dimensional anomaly detection means.

Description

Technical Field

[0001] This application relates to the field of Internet of Things sensing and intelligent electrical control technology, and in particular to a construction site energy consumption monitoring system. Background Technology

[0002] Currently, most construction sites still rely on traditional, extensive energy management models using manual labor and simple hardware assistance. A full-chain digital energy management system has not yet been established. Existing mainstream energy management methods can be mainly divided into four categories. The first is the manual meter reading and inspection model, where dedicated water and electricity management personnel are assigned to the construction site to read the original readings of electricity and water meters distributed in various areas on a fixed daily or weekly basis. Energy consumption data is recorded manually in paper ledgers, completing the basic energy consumption data collection. The second is the passive current limiting protection scheme, which simply installs basic protection devices such as circuit breakers in the distribution box at the construction site. Relying on the inherent overload protection function of the circuit breaker, it automatically cuts off the power supply circuit when the circuit current exceeds the rated threshold. This only avoids the safety risk of circuit overload and does not have any active energy consumption regulation or saving capabilities. The first aspect is the ability to manage and control; the second is the timed on / off control method, which involves installing mechanical timers for equipment with fixed working conditions such as conventional temporary lighting at construction sites. This method relies on preset fixed time intervals to automatically turn the lighting equipment on and off, but it is only suitable for a single scenario and cannot adapt to the dynamic changes in construction conditions at the site; the third aspect is the monthly bill review management, where the construction unit only uses the monthly payment bills issued by the power company and water company as the sole basis for energy consumption statistics. The total electricity and water consumption for the whole month are then summarized afterward. If there is a monthly energy consumption overrun, management personnel are arranged to manually investigate the energy consumption anomalies throughout the entire area.

[0003] The aforementioned traditional energy management methods are low-cost and simple to deploy, and can meet the most basic energy metering and circuit safety protection needs of construction sites. However, they have many inherent defects in adapting to the current dynamic and large-scale construction site scenarios, and are difficult to support refined energy-saving management and comprehensive energy consumption risk control. The specific problems are as follows:

[0004] First, energy consumption data collection is lagging and coarse-grained, making it extremely difficult to trace the source of energy consumption anomalies. Current solutions rely on manual, periodic meter readings, with data collection frequencies ranging from once a week to once a day, failing to achieve high-frequency data collection at the second or hourly level. Management personnel cannot grasp the detailed energy consumption dynamics of different areas and equipment on the construction site in real time. All energy consumption anomalies are only passively discovered after the monthly official bill is issued, by which time energy waste has often continued for weeks or even an entire calendar month, resulting in irreversible economic losses. Furthermore, the system only provides overall total energy consumption data, lacking independent metering capabilities for different areas and equipment. Once energy consumption exceeds limits, it is impossible to accurately pinpoint the responsible area or specific electrical and water-using equipment corresponding to the abnormal energy consumption, leading to extremely low efficiency in tracing and investigating the source.

[0005] Secondly, the proportion of ineffective energy consumption in unmanned areas remains high, and energy waste is a persistent problem. Construction sites encompass multiple work zones, including civil engineering, installation, and decoration. Work shifts and hours vary significantly across these zones, resulting in many areas being unmanned during nighttime, lunch breaks, and downtime. However, temporary lighting, equipment charging ports, small power tools, and standby distribution boxes remain powered and supplied with water, leading to long-term ineffective energy consumption during idle and standby periods. Based on measured energy consumption data from construction sites, ineffective energy consumption under unmanned conditions accounts for 20%-35% of the total project energy consumption. Calculations show that a single medium-sized building construction site incurs direct economic losses exceeding 150,000 yuan annually due to ineffective energy consumption, significantly increasing the overall construction cost over the long term.

[0006] Third, the lack of real-time monitoring capabilities for energy-related faults leads to a double burden of economic losses and construction safety risks. Construction sites involve complex pipeline layouts, with potential leakage risks in power supply lines and numerous water pipe joints. Construction machinery operations and earthmoving disturbances easily cause pipe damage and line aging, leading to faults such as electrical leakage and continuous water leakage. Current technology lacks online real-time monitoring and automatic fault alarm functions; all equipment and pipeline anomalies are discovered incidentally during manual on-site inspections, resulting in extremely delayed warnings. Water leakage is the most prominent issue, with a single large leak point wasting tens of tons of water daily. Such faults typically take several days or even weeks to be manually detected, causing significant water resource and fee losses. Furthermore, electrical leakage can lead to electric shocks and fires, while water leakage can cause water accumulation in foundation pits and slippery roads, directly threatening the personal safety of on-site construction workers.

[0007] Fourth, the energy management model is rudimentary and lacks data-driven energy-saving optimization strategies. Traditional management models only focus on controlling the total monthly energy consumption, failing to break down electricity and water usage structures and unable to analyze core data such as the energy consumption ratio of high-energy-consuming equipment and peak-valley energy consumption distribution. There is no intelligent coordination mechanism for the start-up and shutdown sequence of high-power construction equipment such as tower cranes, concrete pumps, and construction elevators on construction sites. When multiple high-power devices start simultaneously, it generates instantaneous high-power load impacts. On the one hand, this can easily cause transformers on the construction site to operate under overload conditions, affecting the stable operation of construction equipment; on the other hand, it can trigger the power company's demand-based billing rules, generating high additional demand charges and further increasing the project's electricity costs. Overall energy management lacks data analysis capabilities and cannot output intelligent energy-saving scheduling solutions tailored to on-site working conditions.

[0008] Fifth, the lack of a unified energy consumption management view across multiple projects within the group, and the absence of a horizontal benchmarking and management mechanism. For construction groups with multiple construction sites, energy consumption data from each site is stored independently and in a scattered manner, without unified cloud-based aggregation and coordinated management. This prevents the group's management from horizontally comparing core energy intensity indicators such as energy consumption per unit output value and energy consumption per unit construction area across different projects. It also hinders the rapid identification of high-energy-consuming and inefficient projects for targeted rectification and optimization, and prevents the accumulation of standardized energy-saving management experience for nationwide implementation. Consequently, the overall energy management level within the group is inconsistent, making it difficult to achieve a unified low-carbon and energy-saving management goal across all projects.

[0009] In summary, the current traditional energy management model for construction sites suffers from a series of pain points, such as data lag, serious energy waste, untimely fault warnings, extensive management and control, and insufficient overall coordination capabilities of the group. It can no longer meet the actual needs of green construction, cost reduction and efficiency improvement, and full-domain digital management and control at the current stage of construction sites. Summary of the Invention

[0010] To address the aforementioned issues, this application provides a construction site energy consumption monitoring system.

[0011] In view of this, the first aspect of this application provides a construction site energy consumption monitoring system, comprising:

[0012] The energy consumption monitoring unit is used to collect the electrical energy parameters and water consumption parameters of the construction site, obtain energy consumption data, and upload the energy consumption data to the edge gateway;

[0013] The presence sensing unit is used to detect the presence of personnel in various areas of the construction site in real time, output the number of personnel in each area, and upload the personnel data to the edge gateway.

[0014] The edge gateway is used to upload the energy consumption data and personnel data to the cloud platform, and to determine the personnel situation in each area based on the number of personnel in each area. When a certain area is unoccupied and the duration of unoccupancy reaches a threshold, a power outage command is generated and sent to the execution unit corresponding to that area.

[0015] An execution unit is used to perform a power-off operation on the corresponding area according to the power-off command;

[0016] The cloud platform includes:

[0017] The energy consumption prediction module is used to predict the future energy consumption of each area based on the energy consumption data of the construction site, personnel data of each area, construction plan and weather data, and obtain the energy consumption prediction results.

[0018] Anomaly detection module is used to perform statistical threshold anomaly detection, model prediction deviation anomaly detection, electrical characteristic anomaly detection, and water usage anomaly detection based on the energy consumption data, and obtain anomaly detection results;

[0019] The energy-saving strategy optimization module generates energy-saving optimization suggestions based on the energy consumption prediction results and the anomaly detection results.

[0020] Optionally, the energy consumption monitoring unit includes three-phase smart meters installed on the incoming side of the main distribution box at the construction site and in each distribution box, and current recorders for high-power equipment; the high-power equipment is equipment whose rated power exceeds the target power; the energy parameters include voltage data, current data, power data, energy data, power factor, and total harmonic distortion rate;

[0021] The energy consumption monitoring unit also includes an electromagnetic flow meter installed on the main water inlet pipe, remote water meters in each area, and ground water sensors deployed below pipe joints, in areas where valve groups are concentrated, and in low-lying areas.

[0022] Optionally, the edge gateway is specifically used to determine the personnel situation in each area based on the number of people in each area;

[0023] If an area is unoccupied and the duration of unoccupancy reaches a threshold, determine whether the area is a non-power-off zone. If so, skip the power-off process.

[0024] If not, the system will determine whether there are ongoing construction tasks in the area based on the construction plan. If there are ongoing construction tasks in the area, the power outage process will be skipped.

[0025] If there is no construction task in the area, it checks whether there is any equipment in the area that needs to be continuously powered on for cooling. If so, the power-off process is skipped; otherwise, a power-off command is generated and sent to the corresponding execution unit in the area.

[0026] Optionally, the edge gateway is further configured to generate a power restoration command to the execution unit corresponding to the power outage area when it detects that personnel have entered the power outage area, so that the execution unit performs the power restoration operation of the power outage area according to the power restoration command.

[0027] Optionally, the energy consumption prediction module is specifically used to input historical energy consumption data sequences, weather forecast data, construction plans and holiday data into the energy consumption prediction model to predict energy consumption and obtain the hourly electricity consumption of each region and the total daily electricity consumption within the future preset days.

[0028] Optionally, the energy consumption prediction module is further configured to extract features from personnel data, construction plans, and weather data of each area within a preset time period after the start of construction, to obtain time features, construction features, personnel features, and weather features; train a random forest regression model using the time features, construction features, personnel features, and weather features as input data, and using the daily energy consumption data within the preset time period as the training target, to obtain a baseline model; and predict the expected energy consumption of each area under normal construction conditions through the baseline model.

[0029] Optionally, the model prediction bias anomaly detection process includes:

[0030] Calculate the deviation between the actual energy consumption and the expected energy consumption predicted by the baseline model;

[0031] When the deviation value exceeds the first deviation threshold but does not exceed the second deviation threshold, and continues for a first preset duration, a first-level alarm is triggered;

[0032] When the deviation value exceeds the second deviation threshold but does not exceed the third deviation threshold, and continues for a second preset duration, a second-level alarm is triggered;

[0033] When the deviation value exceeds the third deviation threshold, a third-level alarm is immediately triggered;

[0034] The deviation threshold is a preset percentage of the expected energy consumption, the first preset duration is greater than the second preset duration, and the levels of the first-level alarm, the second-level alarm, and the third-level alarm increase sequentially.

[0035] Optionally, the water usage anomaly detection process includes:

[0036] Pipeline leak detection is performed based on the instantaneous flow rate of water pipes in each zone, including:

[0037] The instantaneous flow rate of water pipes in each zone during the nighttime shutdown period is obtained from the energy consumption data. When the instantaneous flow rate continuously exceeds the preset nighttime baseline flow rate threshold, it is determined to be a suspected pipe leak.

[0038] Alternatively, the instantaneous flow rate and network pressure of the water supply pipeline can be acquired simultaneously. If the instantaneous flow rate does not increase but the network pressure fluctuates abnormally, it is determined to be a suspected pipeline leak.

[0039] Alternatively, calculate the difference between the total water inflow at the construction site and the sum of the water flow in each zone. If the difference exceeds a preset percentage threshold of the total water inflow, it is determined to be a suspected pipe leak.

[0040] Water waste detection is performed based on the cumulative water consumption data and construction status information at the construction site, including:

[0041] If water continues to be used in a certain area after construction is completed and the duration exceeds the preset time threshold, it is judged as a wasteful behavior of leaving the valve open.

[0042] Alternatively, when the single water consumption in a certain area exceeds a preset multiple threshold of the historical average water consumption of similar processes, it is judged as a water waste behavior.

[0043] Optionally, the electrical anomaly detection process includes:

[0044] The imbalance of the three-phase current is calculated based on the current data. When the imbalance exceeds the preset imbalance threshold and continues for a third preset time, it is determined to be a suspected leakage current and a leakage current alarm is output. At the same time, the neutral current is acquired. When the zero-sequence current exceeds the preset zero-sequence current threshold, the leakage current protector is triggered to check and remind.

[0045] The operating power and rated power of each monitoring device are obtained from the energy consumption data. When the operating power of the monitoring device is lower than a preset percentage of the rated power and exceeds the idle time threshold corresponding to the monitoring device, the device is determined to be in an idle operating state.

[0046] Harmonic detection is performed based on the total harmonic distortion rate: when the total harmonic distortion rate exceeds the preset harmonic threshold, it is determined that the harmonics exceed the standard and a harmonic alarm is output.

[0047] Optionally, the system further includes:

[0048] The mobile terminal is used to obtain energy consumption data, anomaly detection results and energy-saving optimization suggestions from the cloud platform, and automatically generate reports according to a preset time period. The reports include data on electricity consumption, water consumption, abnormal events and savings.

[0049] The generated report is exported as a preset format file and connected to external business systems via an API interface.

[0050] As can be seen from the above technical solutions, this application has the following advantages:

[0051] The construction site energy consumption monitoring system provided in this application comprises an energy consumption monitoring unit and a multi-technology integrated on-site sensing unit forming the front-end sensing layer. An edge gateway enables local real-time calculation and millisecond-level power outage control. An execution unit performs physical power outage operations. The cloud platform integrates an energy consumption prediction module, a multi-dimensional anomaly detection module, and an energy-saving strategy optimization module, collaboratively forming a complete system from data acquisition and local closed-loop control to cloud-based intelligent optimization. This solution achieves a shift from extensive total energy consumption statistics to three-level, progressively refined traceability, reducing anomaly detection time from days or even weeks to less than one hour. Through a "power supply by personnel, power outage by unmanned personnel" intelligent linkage system, it systematically eliminates 20% of total energy consumption. This solution addresses 35% to 35% of ineffective energy consumption in unmanned areas. Utilizing AI prediction, it supports peak-valley electricity pricing optimization, demand management, and monthly budget early warning, shifting from reactive post-event response to proactive pre-event prevention. Simultaneously, a tiered strategy balances automation efficiency with human safety, and multi-level dashboards cover the entire management scenario, from single construction sites to cross-site comparisons across multiple sites within a group. Real-world testing has verified that this solution can reduce monthly electricity consumption by 25% to 30% and water consumption by 15% to 20% for medium-sized construction sites, resulting in monthly savings of approximately RMB 24,600 and annualized savings exceeding RMB 250,000. With a payback period of 6 to 12 months, it demonstrates significant economic benefits and broad prospects for widespread application. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A schematic diagram of a construction site energy consumption monitoring system provided in this application embodiment;

[0054] Figure 2 This is a project manager APP interface provided for an embodiment of this application. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0056] For easier understanding, please refer to Figure 1 This application provides a construction site energy consumption monitoring system, including:

[0057] The energy consumption monitoring unit is used to collect the electrical and water consumption parameters of the construction site, obtain energy consumption data, and upload the energy consumption data to the edge gateway;

[0058] The presence sensing unit is used to detect the presence of personnel in various areas of the construction site in real time, output the number of personnel in each area, and upload the personnel data to the edge gateway.

[0059] Edge gateways are used to upload energy consumption data and personnel data to the cloud platform, and determine the personnel situation in each area based on the number of personnel in each area. When a certain area is unoccupied and the duration of vacancy reaches a threshold, a power outage command is generated and sent to the corresponding execution unit in that area.

[0060] The execution unit is used to perform power-off operations in the corresponding area according to the power-off command.

[0061] Cloud platforms, including:

[0062] The energy consumption prediction module is used to predict the future energy consumption of each area based on the energy consumption data of the construction site, personnel data of each area, construction plan and weather data, and obtain the energy consumption prediction results.

[0063] The anomaly detection module is used to perform statistical threshold anomaly detection, model prediction deviation anomaly detection, electrical characteristic anomaly detection, and water usage anomaly detection based on energy consumption data, and obtain anomaly detection results.

[0064] The energy-saving strategy optimization module generates energy-saving optimization suggestions based on energy consumption prediction results and anomaly detection results.

[0065] The energy consumption monitoring unit in this embodiment includes energy monitoring equipment (smart meters and current recorders). A three-level monitoring system is used to deploy the energy monitoring equipment, enabling the tiered collection of energy consumption data for the main power line, each distribution area, and key high-power equipment. The first level is the main power line monitoring, where a three-phase smart meter is installed on the incoming side of the main distribution box at the construction site. A non-contact connection method using current transformers is adopted to avoid altering the existing power distribution lines. This level of electricity meter monitors the following complete set of electrical parameters: voltage parameters (three-phase voltage Va, Vb, Vc), current parameters (three-phase current Ia, Ib, Ic), power parameters (active power P, reactive power Q, apparent power S), energy parameters (cumulative active power (kWh), cumulative reactive power (kVarh), power factor (three-phase and total power factor cosφ), maximum demand in fifteen minutes (directly affecting basic electricity charges), and total harmonic distortion (THD) (normally should not exceed 5%). The data acquisition frequency is set to once per minute, and can be automatically increased to once per second when the system detects abnormal fluctuations to capture transient fault characteristics.

[0066] The second level is branch or area monitoring, which involves embedding single-phase or three-phase smart energy meters in each distribution box on the construction site. Coverage includes the main construction area (including tower cranes, concrete pumps, construction elevators, etc.), temporary living areas (dormitories, canteens, bathhouses, etc.), office areas (project management, supervision, client's on-site staff, etc.), processing areas (steel reinforcement processing, carpentry processing, etc.), and fire and safety facilities (monitored independently, not included in energy-saving control). The monitoring parameters for this level are exactly the same as for the first level, collecting a full set of data including voltage, current, power, energy consumption, power factor, demand, and harmonics. However, the collection frequency is reduced to once every five minutes to decrease the data upload load. The core function of the second-level monitoring is to pinpoint the specific area where energy consumption anomalies occur, such as determining whether a sudden increase in electricity consumption originates from the main construction area or the living area.

[0067] The third level is equipment-level monitoring, targeting critical equipment with high rated power or high energy consumption, including tower cranes (rated power 30-80kW, key monitoring), concrete pumps (rated power 45-90kW), construction elevators (rated power 11-22kW), large welding machines (rated power 20-60kW), and air compressors (rated power 7.5-37kW). Monitoring equipment uses smart sockets or clamp-type current recorders, installed on the equipment's power supply line. This level of monitoring also collects the same complete set of electrical parameters as the first and second levels, but the collection frequency is further increased to once every thirty seconds to finely analyze the equipment's start-up and shutdown times, no-load and load states, operating efficiency, and current harmonic characteristics. High-frequency acquisition enables the system to accurately determine whether the equipment is operating under no-load conditions (e.g., tower cranes in standby mode), whether leakage has occurred (calculated through three-phase current imbalance), and whether there are abnormalities such as excessive harmonics.

[0068] The data flow of the three-tier monitoring system is as follows: all smart meters communicate with the edge computing gateway via RS485 bus or LoRaWAN wirelessly. The wired solution uses the Modbus RTU protocol, with a maximum of 32 devices connected to a single RS485 bus; the wireless solution uses LoRaWAN, with a single gateway covering a radius of up to 500 meters, suitable for large, dispersed construction sites. After aggregating all meter data, the edge computing gateway first stores and preprocesses the data locally, then encrypts and uploads it to the cloud platform via 4G or 5G network. To address unstable network conditions at construction sites, the gateway has a built-in local storage function that can cache 72 hours of raw data, automatically re-uploading it in chronological order after network recovery to ensure no data loss. The entire system maintains consistent parameter definitions from tier 1 to tier 3, with higher tiers including all parameters from lower tiers, refining only the acquisition frequency and deployment granularity at each level, thus achieving end-to-end energy consumption traceability from the main power distribution room to individual devices.

[0069] The core advantages of Level 3 equipment-level monitoring are reflected in six aspects: First, with high-frequency data acquisition every 30 seconds, the system can more accurately capture the start-up and shutdown moments, no-load status, and load fluctuations of high-power equipment, significantly improving data accuracy. Second, monitoring reaches individual equipment directly, enabling precise traceability of energy consumption to specific machines, rather than remaining at the regional level. Third, through fine-grained current and power analysis, the system can accurately distinguish between the operating, no-load, and standby states of equipment, thereby achieving precise control over no-load power consumption. Fourth, based on high-frequency data, the system can quickly identify abnormal current, excessive harmonics, and potential leakage current in individual equipment, achieving early fault diagnosis. Fifth, the system can generate personalized energy-saving strategies for each piece of equipment, such as staggered start-up and shutdown or load reduction operation, ensuring the implementation of energy-saving measures. Finally, by constructing a complete three-level data hierarchy structure of main incoming line, zone, and equipment, energy consumption accounting and cost breakdown are clearer, providing a reliable data foundation for refined energy management.

[0070] The energy consumption monitoring unit also includes water monitoring equipment, which is used to monitor total water intake, zoned water use, and leak detection, enabling real-time metering of water consumption and highly sensitive identification of abnormal leaks at the construction site. For total water intake monitoring, an electromagnetic flow meter is installed on the main water intake pipe (behind the water meter), with the pipe diameter selected from DN50 to DN100 based on the size of the main pipeline at the construction site. This electromagnetic flow meter measures instantaneous flow rate (m³ / s). 3 / h) and cumulative water consumption (m 3The system achieves an accuracy of ±0.5% of full scale, with a data acquisition frequency of once per minute, ensuring real-time monitoring of total water usage at the construction site. For zoned water usage monitoring, the system deploys remote water meters (pulse output type) in conjunction with LoRa wireless transmission modules, covering water usage for concrete mixing and curing (typically the area with the highest water consumption), water usage in worker living areas (dormitories, canteens, and bathhouses), fire-fighting water storage (monitored independently, not included in energy-saving control), and water usage for dust control spraying in green construction. The zoned water meters acquire data every ten minutes, sufficient to capture water usage fluctuations in typical processes. For leak detection, the system deploys surface water sensors below main pipe joints, in areas with concentrated valve installations, and in low-lying areas prone to water accumulation. These sensors utilize a resistive principle and have an IP68 protection rating, allowing for long-term immersion operation. The sensor's trigger logic is as follows: when water accumulation is detected and weather data indicates non-rainy weather, the system determines it as a suspected pipe leak and generates an alarm. In addition to water accumulation sensors, the system also employs flow anomaly detection methods (details in the subsequent anomaly detection section), including the minimum nighttime flow method, flow-pressure correlation analysis method, and regional flow balance method, thereby elevating leak detection from point-based sensing to system-wide monitoring. Data from all water meters and flow meters is also aggregated to the edge computing gateway via LoRa wireless or wired RS485 bus, sharing the same transmission channel and offline caching mechanism as electricity meter data, and is ultimately uploaded to the cloud platform for water usage anomaly analysis and water-saving effect evaluation.

[0071] The presence sensing unit includes personnel presence sensing and equipment presence sensing, employing a multi-technology fusion scheme to accurately determine the status of personnel and equipment in each area. The primary solution utilizes UWB (Ultra-Wideband) positioning technology. This involves deploying UWB base stations at 100-meter intervals (adjustable according to actual conditions) across various construction areas, and leveraging UWB tags already widely used in workers' safety helmets to achieve real-time positioning accuracy of ±30 centimeters, outputting the number of people present in each area. The auxiliary solution uses facial recognition or card-swipe gates installed in dormitory areas and main entrances to record personnel entering and exiting each area. A supplementary solution is video AI personnel detection, reusing existing video surveillance cameras on the construction site and running the YOLOv8 personnel detection algorithm to count the number of people in peripheral areas without UWB coverage. In addition, the on-site sensing unit also combines equipment operation status sensing: it judges the equipment operation status through electrical characteristics, that is, when the current is greater than 1.2 times the no-load current, it is determined that the equipment is in operation; when the current is less than or equal to the no-load current, it is determined to be no-load or standby; at the same time, vibration sensors are installed on large mechanical equipment (such as tower cranes and concrete pumps), and smart circuit breakers provide remote control and status feedback.

[0072] The edge computing gateway is the core hub connecting front-end sensors (energy consumption monitoring units, presence sensing units) and the cloud platform, undertaking multiple tasks such as data acquisition, local processing, offline caching, real-time control, and data uploading. In terms of hardware specifications, the gateway uses an ARM Cortex-A72 quad-core processor with a main frequency of 1.8GHz, equipped with 4GB of RAM and a 64GB embedded multimedia storage chip (eMMC), and supports TF card expansion for offline data caching. Communication interfaces include four RS485 ports (for connecting smart meters and water meters), two Ethernet ports (for local network access), a 4G or 5G module (for encrypted data upload to the cloud platform), a LoRa module (for receiving wireless sensor data), and 16 relay outputs (for controlling circuit breakers or solenoid valves). The edge gateway operates in a temperature range of -20℃ to 70℃, with an IP54 protection rating, enabling it to withstand dusty and humid environments on construction sites; it is installed using a standard 35mm DIN rail and can be embedded in a distribution box.

[0073] The edge gateway implements multiple local real-time processing functions. It can collect data from all electricity meters, water meters, and sensors with millisecond-level precision (1000-millisecond timestamps), and perform real-time calculations of active power, reactive power, apparent power, and power factor locally. In the event of a network outage, the gateway uses its local storage capacity to cache at least 72 hours of raw data, automatically re-uploading it to the cloud in chronological order after network recovery, ensuring no data loss. Furthermore, the gateway performs data preprocessing: it aggregates the raw data collected every minute into a five-minute average and maximum value locally, compresses it, and then uploads it to the cloud, saving approximately 80% of 4G traffic; while the raw data corresponding to abnormal events is uploaded completely without compression to retain high-precision details for fault investigation.

[0074] At the data architecture level, the overall data flow is divided into four layers. The bottom layer is the sensor layer, including smart meters, smart water meters, and various sensing sensors; these devices transmit data to the edge gateway layer via RS485 wired bus or LoRa wireless method. After completing local computing, storage, and control, the edge gateway layer uploads the data to the cloud platform layer via 4G or 5G network with TLS1.3 protocol encryption. The cloud platform layer provides big data storage, AI analysis, and management interface, and outputs data to the application layer through application programming interfaces (APIs), including project site dashboards, group management platforms, and mobile apps.

[0075] The database design employs a hybrid architecture. The time-series database uses InfluxDB to store all collected time-series data, including current, voltage, power, and flow. Its retention strategy is as follows: raw one-minute granular data is stored for one year, and aggregated five-minute data is stored for three years. Query performance achieves a retrieval time of less than 200 milliseconds for 1 million data points. The relational database uses PostgreSQL to store equipment ledgers (basic information on electricity meters, water meters, and sensors), alarm records (historical alarm events and handling records), energy-saving measure records (when and what energy-saving actions were performed), and project progress correlations (energy consumption baselines for different construction stages).

[0076] In terms of data security, the system implements multi-layered protection. Transmission uses TLS 1.3 end-to-end encryption; statically stored data uses AES-256 encryption; access control employs role-based access control (RBAC) to differentiate between different permission levels, such as project managers, energy administrators, and read-only roles; the data backup strategy involves daily full backups and off-site storage, meeting the data security and compliance requirements of government projects. Through this design, the edge computing gateway and cloud platform together form a highly reliable, low-latency, and secure construction site energy consumption data collection and analysis system.

[0077] Based on the functional attributes and security requirements of each area, the edge gateway divides all power-consuming areas on the construction site into four categories: A, B, C, and D, and formulates differentiated power outage control strategies for each category.

[0078] Category A areas are those that can be completely de-energized, including temporary lighting, tool charging areas, rest sheds, and temporary fencing areas containing unoccupied and non-critical facilities. A Category A area is de-energized when no one is present for five consecutive minutes or more. Power restoration is automatically completed within three seconds of detecting personnel entering the area, and the equipment used is a remotely controlled intelligent circuit breaker.

[0079] Category B areas are areas where partial power outages are permitted, typically including processing workshops and material warehouses. In Category B areas, a selective power outage strategy is implemented when no one is present: processing equipment and work lighting are completely powered off, while independent circuits such as safety monitoring power and emergency lighting are kept powered. The power outage condition is when no one is present for ten consecutive minutes or more.

[0080] Category C areas are areas requiring continuous power supply, including fire alarm systems, security cameras, tower crane obstruction lights, construction elevator safety systems, perimeter safety warning lights, project office servers, and emergency lighting main circuits—facilities requiring 24-hour continuous power. Category C areas are physically locked by the system and are not included in any automatic power-off zones.

[0081] Category D targets energy conservation under no-load conditions for large equipment, including tower cranes, concrete pumps, and construction elevators, which are in operation but in a no-load state. The strategy involves the edge gateway sending a notification to the operator to pause the equipment when it remains idle for a set time (e.g., a tower crane idle for more than fifteen minutes). After operator confirmation, the equipment is reduced to idle or standby mode. It should be noted that, for safety reasons, direct remote power-off is not implemented for large equipment; only suggestions are sent for operator decision-making.

[0082] At the power outage control execution level, intelligent circuit breakers with a rated current selectable from 16A to 400A are used as the execution unit. The control method involves the edge gateway driving the intelligent circuit breaker to open or close via relay output, with a response time of less than 500 milliseconds from command issuance to circuit breaker action. To ensure safety, the system has a triple safety interlock mechanism: First, before power outage, it checks whether the target circuit belongs to Class C protection circuit; if so, it skips the execution. Second, the power outage command must receive execution confirmation within one second; otherwise, it will retries three times before alarming and requiring manual intervention. Third, any circuit breaker can be manually forced to close on-site, reflecting the principle of safety priority and allowing manual overriding of system decisions.

[0083] The automatic power-off decision-making process cycles every 60 seconds. Step 1: The edge gateway collects the presence status of each area, with data sources including UWB positioning, video detection, and access control records, outputting the number of people in each area. Step 2: By comparing timestamps (e.g., comparing the timestamp of the last time no one was detected to the timestamp of the current time no one was detected), the duration of no one in each area is determined: if the duration is less than a set time threshold, observation continues without power interruption; if the time threshold is reached or exceeded, the power-off process begins. Step 3: Pre-power-off safety checks are performed, including verifying whether the target circuit is a Class C protection circuit (skip if so), verifying whether any procedures in the construction plan are underway and the time has not yet expired (skip if so), and verifying whether any equipment requires continuous power for cooling or temperature reduction (e.g., welding machines or motors that have just stopped running within 30 minutes are skipped). Step 4: The power-off is formally executed: the edge gateway sends a power-off command to the smart circuit breaker in the corresponding area, controlling its tripping, and recording the operation log, including time, area, circuit, and trigger reason. The fifth step is to restore power. There are three triggering conditions: First, personnel are detected entering the area (either through UWB or video sensing), and the system immediately restores power within three seconds (the edge gateway sends a power restoration command to the smart circuit breaker in the corresponding area to control its closing); Second, the construction plan shows that the next process in the area will begin 15 minutes in advance, and the system restores power in advance to provide preheating or preparation time; Third, power is immediately restored through manual operation of the on-site circuit breaker or remote operation via a mobile APP, and the manual intervention event is recorded at the same time.

[0084] The following two typical scenarios illustrate the energy-saving effect. In a construction site dormitory area (with 200 people), on weekdays, after workers leave at 7:00 AM, the system detects that no one is in the dormitory area. At 7:10 AM (after a ten-minute delay for confirmation), the dormitory area's lighting and charging sockets are automatically powered off. After workers gradually return at 5:30 PM, the system automatically restores power. The daily power outage duration is approximately 10.5 hours, saving approximately 60kW × 10.5h = 630kWh / day. At a price of 0.85 yuan / kWh, this translates to a saving of approximately 535 yuan / day in electricity costs: 630kWh × 0.85 yuan / kWh.

[0085] During off-duty hours at night, when the last worker leaves the rebar processing area at 8:00 PM, the processing equipment and lighting are automatically powered off ten minutes later. The next morning at 7:50 AM, the construction schedule shows a processing task scheduled for 8:00 AM, and the edge gateway automatically restores power ten minutes in advance. The daily power outage duration is approximately 11.75 hours, resulting in energy savings of approximately 120kW × 11.75h = 1410kWh / day for the processing area equipment. In summary, the automatic power-off system for unmanned areas significantly reduces ineffective energy consumption on the construction site.

[0086] The edge gateway uploads energy consumption and sensing data collected by the sensor layer to the cloud platform, which then performs AI predictions and anomaly detection based on this data. The cloud platform includes an energy consumption prediction module, which consists of two main modules: energy consumption baseline establishment and energy consumption time-series prediction. Establishing the energy consumption baseline requires a data collection period, specifically the first thirty days after construction begins. During this period, hourly electricity consumption (kWh) and daily water consumption (m³) in each area are collected. 3 The data collected included the day's construction content (obtained through the construction log), the number of workers on site, and weather data (including temperature and rainfall status, automatically obtained via a weather application programming interface). Feature extraction was performed on this data, extracting four categories of features: time, construction, personnel, and weather. Time features included hour, day of the week, and project progress stage (foundation, main structure, decoration). Construction features included the main work processes of the day (earthwork, concrete, masonry, installation). Personnel features included the number of workers on site, categorized by job type. Weather features included temperature, humidity, and rainfall status. The baseline model employed an ensemble algorithm of Random Forest Regression and Gradient Boosting Tree (GBDT) to predict the expected energy consumption (including water and electricity consumption) in each area under normal construction conditions. For model validation, 20% of the data was used as the validation set to validate the trained baseline model, requiring a mean absolute percentage error (MAPE) of less than 8%. To ensure that the baseline always reflects the actual energy consumption patterns of the current construction phase, the system implements a dynamic baseline update mechanism: when the construction phase changes (for example, from the earthwork phase to the foundation phase and then to the main structure phase), the energy consumption baseline for that phase is automatically re-established; at the same time, it is updated weekly to include the latest seven days of data to maintain the timeliness of the baseline.

[0087] The energy consumption prediction module employs a Long Short-Term Memory (LSTM) time-series prediction model. The model's input features include historical energy consumption sequences (e.g., hourly energy consumption data for the past 7 days), future weather forecasts (temperature and precipitation, etc.) obtained from a weather API, future construction plans obtained from a project management system (e.g., work schedules for the next 3 days), and holiday or rest day markings. The model output includes hourly electricity consumption predictions for each region over the next 24 hours, total daily electricity consumption predictions for the next 7 days, and prediction confidence intervals (80% / 95% confidence band). The prediction accuracy of the trained LSTM time-series prediction model is validated using a validation set. The target is set to a 24-hour average absolute percentage error (ARR) of less than 6% and a 7-day ARR of less than 12%.

[0088] The forecast results support three core application scenarios. The first is peak-valley electricity price optimization: The cloud platform identifies the electricity consumption periods of high-energy-consuming processes based on the forecast results and automatically suggests adjusting high-energy-consuming operations to be performed during off-peak hours, which can save 30-40% of electricity costs. Taking a concrete pouring as an example, the peak electricity price is 1.2 yuan / kWh (09:00-11:00, 18:00-21:00), and the off-peak electricity price is 0.4 yuan / kWh (23:00-07:00 the next day). The cloud platform predicts that the electricity consumption for concrete pouring tomorrow will be approximately 800kWh, and automatically generates an optimization suggestion to adjust the pumping period from 10:00-14:00 to 23:00-03:00. The original electricity cost was 800kWh × 1.2 yuan = 960 yuan (peak period), and the optimized electricity cost is 800kWh × 0.4 yuan = 320 yuan (off-peak period), saving 640 yuan in electricity costs per pouring. The second feature is demand management: the cloud platform predicts the maximum demand in the next fifteen minutes. If the predicted value exceeds the contracted demand, an early warning is issued, and it is recommended to start high-power equipment during off-peak hours to avoid transformer overload or exceeding the demand and electricity bill limits. The third feature is monthly budget early warning: the cloud platform predicts the total electricity and water consumption for the month and compares it with the monthly budget to identify the risk of overspending in advance so that timely control measures can be taken.

[0089] The cloud platform also includes an anomaly detection module, which adopts a multi-dimensional algorithm parallel detection strategy to comprehensively monitor the energy consumption data of the construction site from four dimensions: statistical characteristics, model prediction deviation, electrical characteristics, and water consumption characteristics, so as to realize hierarchical anomaly identification from rapid response to fine diagnosis.

[0090] Anomaly Type 1: Statistical Threshold Anomaly (Rapid Detection). This method is based on the Z-Score anomaly detection principle, calculating in real time the deviation of current energy consumption data from historical energy consumption statistical characteristics. The Z-Score calculation formula is Z = (Current Value - Historical Mean) / Historical Standard Deviation. The anomaly detection module maintains a sliding window containing historical data for the same time period within the last thirty days, with the mean and standard deviation updated in real time based on this window. Whenever new minute-level data arrives, the anomaly detection module immediately calculates the anomaly value Z and triggers a tiered alarm based on its absolute value: when 3.0 ≥ |Z| > 2.0, it is a minor anomaly (Warning level), requiring attention; when 4.0 ≥ |Z| > 3.0, it is a moderate anomaly (Alert level), requiring investigation; when |Z| > 4.0, it is a severe anomaly (Critical level), requiring immediate action. This method has a real-time response speed, meaning that calculation and judgment are completed as soon as data arrives every minute.

[0091] Anomaly Type 2: Model Prediction Bias Anomaly (Fine-grained Detection). This method compares the actual energy usage with the predicted values ​​of the AI ​​baseline model to eliminate the interference of differences in construction procedures on anomaly detection. Details are as follows:

[0092] If the baseline model prediction value × 30% ≥ actual energy consumption - baseline model prediction value > baseline model prediction value × 15%, and this continues for more than 30 minutes, a Warning level alarm will be triggered.

[0093] If the baseline model prediction value × 50% ≥ actual energy consumption - baseline model prediction value > baseline model prediction value × 30%, and this continues for more than 15 minutes, an Alert level alarm will be triggered.

[0094] When the actual energy consumption minus the baseline model prediction exceeds 50% of the baseline model prediction, a Critical level alarm is immediately triggered, which is highly likely to indicate equipment failure or electricity theft. The advantage of this method is that by comparing baseline values ​​under the same procedures and conditions, it effectively reduces false alarms caused by normal variations in construction intensity.

[0095] Anomaly Type 3: Electrical Characteristic Anomalies (Equipment-Level Diagnosis). This type includes three sub-detection functions. The first is leakage current detection, which is based on a comprehensive judgment of excessive three-phase current imbalance combined with abnormal neutral current. Imbalance = (Maximum Phase Current - Minimum Phase Current) / Average Current. When the imbalance exceeds 15% and lasts for more than 5 minutes, the anomaly detection model determines it as a suspected leakage current and issues an alarm; if the zero-sequence current exceeds 30mA, it triggers a leakage current protection device (RCD) check and reminder. The second is equipment no-load operation detection. The principle is that when the equipment's operating power is lower than 20% of the rated power and lasts for more than a set time, it is determined to be no-load operation. The system sets differentiated thresholds for different equipment: tower cranes are alerted when the no-load current lasts for more than 15 minutes, concrete pumps are alarmed when the no-load current lasts for more than 10 minutes, and welding machines are advised to be disconnected when the standby time lasts for more than 20 minutes. The third item is harmonic exceedance detection, which addresses the harmonic issues generated by nonlinear loads such as frequency converters and welding machines. When the total harmonic distortion rate exceeds 8%, an alarm is issued, indicating that excessive harmonics will lead to additional transformer losses and shorten equipment lifespan, and suggesting the installation of filters or adjustment of load distribution.

[0096] Anomaly Type 4: Water Usage Anomaly Detection. This type includes two main functions: pipe leak detection and water waste detection. Pipe leak detection uses three methods in parallel. Method 1 is the minimum nighttime flow method. Its logic is that construction work is suspended between 23:00 and 05:00 at night, and water consumption should theoretically be zero. If water consumption is consistently greater than 0.1m during this period... 3 A flow rate of / h is considered a suspected leak. Method two is the flow-pressure correlation analysis method. During normal water use, increased flow is accompanied by a decrease in network pressure, while leaks occur when there is no water demand but abnormal pressure fluctuations, thus identifying leak events. Method three is the regional flow balance method. This involves calculating the total inflow minus the sum of water consumption in each zone to determine unexplained losses. When unexplained losses exceed 5% of the total inflow, a suspected leak is identified. Regarding water waste detection, an alarm is triggered if water continues to be used in a certain area after construction is completed; an alarm is triggered if a single water consumption exceeds 200% of the historical average for similar processes. Through comprehensive detection and diagnosis of these four types of anomalies, the system can detect various energy consumption anomalies on a minute-level timescale, providing accurate evidence for timely handling.

[0097] The anomaly diagnosis module employs a combination of rule-based reasoning and machine learning. After various anomaly detection algorithms trigger alarms, it automatically performs multi-source data correlation and enrichment on the alarm information to pinpoint the root cause and provide actionable handling suggestions. Taking an original alarm, "Electricity consumption in Area A is abnormally high (deviation +65%)", as an example, the system automatically retrieves multi-dimensional data for that period for cross-validation: UWB location data shows zero people present in Area A during that period; the execution record of the automatic power-off strategy for unoccupied areas is retrieved, confirming that Area A should be in a power-off state during that period; and the status feedback from the smart circuit breaker confirms that the circuit breaker is actually closed and supplying power. Based on this evidence, the anomaly diagnosis module uses Bayesian inference or rule-weighting to rank the probabilities of various possible causes: the probability of the circuit breaker's automatic power-off failure is 65%; the probability of someone entering but not being detected (e.g., workers not wearing UWB tags) is 20%; the probability of equipment malfunctioning and starting up is 10%; and the probability of other causes is 5%. The anomaly diagnosis module ultimately generates structured processing suggestions: "Please check the UWB coverage in area A and confirm on-site whether anyone is working; if no one is confirmed, it is recommended to manually disconnect the power supply to area A and check for circuit breaker faults." Through this mechanism, maintenance personnel can obtain accurate diagnostic conclusions and action guidelines without having to piece together scattered data themselves.

[0098] For typical abnormal scenarios, the anomaly diagnosis module embeds a targeted reasoning chain. Scenario 1: Abnormally high electricity consumption late at night. Symptoms include total electricity consumption exceeding the baseline prediction by 40% at 2 AM. The diagnostic logic is as follows: First, compare electricity consumption by zone, pinpointing the anomaly to the processing workshop in Zone B; then, query the UWB sensing data for Zone B, showing that the area is unoccupied, and according to the unoccupied power-off strategy, power should have been cut off; next, check the circuit breaker status, finding that the circuit breaker in Zone B is showing as closed, inconsistent with the expected state, which is a key anomaly signal; finally, compare the recent electricity consumption waveform in Zone B, finding a high degree of consistency with the waveform at the same time yesterday, exhibiting a regular pattern. Based on the above clues, the anomaly diagnosis module infers that someone is highly likely using the construction site equipment for private production or electricity theft. This alarm is classified as Critical, recommending "arranging on-site security to immediately go to Zone B for verification."

[0099] Scenario 2 involves a sudden increase in water consumption on a sunny day. The symptom is a 45% increase in water consumption yesterday compared to the average of the previous seven days, with no rainfall on that day, ruling out the possibility of increased water consumption due to dust control. The diagnostic logic is as follows: First, the construction log was checked to confirm that today's concrete work volume was consistent with normal levels, ruling out factors that could reasonably increase water consumption for construction projects; then, water consumption data for each zone was checked, revealing that water consumption in the living area was normal, while water consumption in the concrete curing area was 80% higher; finally, the minimum nighttime flow method was used for verification, showing a continuous flow of 0.8 m³ / h from 23:00 to 05:00. 3The flow rate was [amount missing] m³ / h, while construction was halted during that period, theoretically the water consumption should have been zero. Based on this, it is inferred that the leak occurred in the pipes or joints within the concrete curing area, with an estimated loss of 0.8 m³ / h × 24h = 19.2 m³ / h. 3 The economic loss is approximately 57 yuan per day. The alarm was classified as Alert level, and it was recommended to "check the water supply pipe joints in the concrete curing area, with a focus on areas recently disturbed by construction."

[0100] Scenario 3: Persistently low power factor. Symptoms include a total incoming power factor consistently below 0.75 for the past week, while the normal value should be above 0.90. A persistently low power factor below 0.90 will trigger a power factor adjustment fee from the power company. The diagnostic logic is as follows: First, a source analysis of the power factor in each area was conducted, identifying the main source of the low power factor as the welding area in Zone C. It was confirmed that a large number of inductive loads, such as welding machines and motors, were concentrated in this area. Then, the status of the reactive power compensation capacitors was checked, revealing that two capacitors were offline and faulty, resulting in insufficient reactive power compensation capacity. It was deduced that the faulty reactive power compensation device was the root cause of the persistently low power factor. The estimated loss is a monthly power factor adjustment fee penalty of approximately 3200 yuan. This alarm was classified as a Warning (and upgraded to an Alert due to its duration of one week), with the recommendation to "immediately repair the reactive power compensation capacitors in Zone C to restore the power factor to above 0.90." Through the automatic reasoning of the above-mentioned anomaly diagnosis module, the system can transform the original alarm signal into a complete diagnostic report with root cause analysis, economic loss estimation and specific handling suggestions, which significantly improves the efficiency of anomaly handling.

[0101] The comprehensive energy-saving strategy library is divided into three priority categories: P1, P2, and P3, based on the degree of automation and the depth of decision-making intervention. Category P1 consists of automatically executed strategies, directly executed by the energy-saving strategy optimization module without manual confirmation. It includes four specific measures: First, automatic power cut-off in unoccupied areas; second, overdue idle equipment alarms, which send a notification to the operator when equipment runs idle for more than a set time, but do not force a power cut-off to ensure safety; third, automatic shutdown of unnecessary power consumption in living areas at night, specifically dimming dormitory public area lighting to 50% of normal brightness after 10 PM; and fourth, automatic shutdown of unnecessary lighting at construction site entrances after sunrise. Category P2 consists of recommended execution strategies, generated by the energy-saving strategy optimization module and pushed to the project manager for one-click confirmation before execution. These include suggestions for adjusting peak-valley electricity pricing for work periods, suggestions for off-peak startup of high-power equipment, suggestions for power factor compensation adjustments, and suggestions for prioritizing leak repairs. P3 is a reference optimization strategy that is not directly pushed to the on-site decision-making level. Instead, it is included in the monthly report for long-term optimization reference. It includes energy intensity benchmarking (comparing with similar construction sites to identify optimization space), suggestions for replacing high-energy-consuming old equipment, and suggestions for optimizing construction organization (such as reducing equipment idle waiting time by optimizing process connection).

[0102] In terms of optimizing the staggered startup of high-power equipment, the system implements precise control over common unreasonable concentrated startup scenarios on construction sites. For example, in a typical scenario:

[0103] At 7:55 AM, the tower crane (rated power 50kW) started, followed by the concrete pump (rated power 75kW) at 7:56 AM, the construction elevator (rated power 15kW) at 7:57 AM, and the rebar bending machine (rated power 11kW) at 7:58 AM. Within four minutes, the total starting power reached 151kW, and the starting inrush current could reach seven times the rated current, easily causing transformer overload tripping or excessive demand, resulting in high electricity bills. The energy-saving strategy optimization module achieves automatic peak-shifting through the following optimization scheme: After the tower crane starts at 7:55 AM, the system monitors and records this high-current impact; from the next day, the system automatically executes the planned start sequence: after the tower crane starts at 7:55 AM, wait two minutes until it stabilizes; at 7:57 AM, start the concrete pump and wait one minute; at 7:58 AM, start the construction elevator and wait one minute; and at 7:59 AM, start the rebar bending machine. This peak-shifting adjustment reduced the maximum inrush current by about 60%, and the maximum demand in 15 minutes dropped from 151kW to about 85kW, effectively saving on demand-related electricity costs.

[0104] Taking monthly data from a medium-sized construction site as an example, the combined effects of various energy-saving measures are statistically analyzed as follows: Automatic power outages in unmanned areas save 12,800 kWh of electricity per month, representing a saving rate of approximately 18%; peak-valley electricity pricing optimization, by shifting high-energy-consuming processes to off-peak hours, saves 8,600 kWh of electricity per month, representing a saving of approximately 22% of the total cost; leak repair saves 320 cubic meters of water. 3 The power factor improvement saved 3,200 yuan in adjustment costs per month; the reduced idle power consumption of equipment decreased by 5,400 kWh, and the idle rate dropped from 35% to 12%. In total, these measures resulted in monthly savings of approximately 24,600 yuan for the construction site, representing a reduction of about 28% in overall energy costs compared to before optimization.

[0105] Furthermore, the system of this application also includes a mobile terminal, which is used to obtain energy consumption data, anomaly detection results and energy-saving optimization suggestions from the cloud platform, and automatically generate reports according to a preset time period. The reports include data on electricity consumption, water consumption, abnormal events and savings. The generated reports are exported as preset format files and connected to external business systems through API interfaces.

[0106] The mobile app enables real-time monitoring of the dashboard, which can be categorized into three levels based on different user roles and usage scenarios: the construction site dashboard, the project manager's app, and the group management platform. These levels serve on-site monitoring, mobile decision-making, and group-level overall management, respectively. The construction site dashboard is deployed on a large-screen display in the project department's duty room, showing real-time total power consumption (kW), today's cumulative power consumption (kWh), total water flow (cubic meters per hour), and today's cumulative water consumption (cubic meters). Energy status for each area is visually represented by green, yellow, and red color blocks: green indicates normal energy consumption, yellow indicates minor anomalies, and red indicates alarms or serious anomalies. The current alarm list displays all unprocessed alarms and highlights them. Today's energy savings are updated in real-time by comparing them with baseline predictions. This dashboard provides duty personnel with a clear overview of the overall energy consumption situation.

[0107] The project manager app, as a mobile management tool, provides real-time data summary push notifications without requiring constant screen monitoring, presenting key metrics concisely on the mobile interface (please refer to...). Figure 2 The alarm push notification system employs a tiered strategy: Critical alarms are sent directly to the project manager via telephone; Warning alarms are sent via app pop-ups; and Info alarms are aggregated into daily briefings and sent uniformly. The app includes a one-click operation function, allowing project managers to directly confirm or reject system-recommended energy-saving strategies on their mobile phones. Furthermore, the app provides real-time summaries of monthly energy-saving achievements, visually displaying savings in electricity and water, enabling project managers to easily monitor progress towards energy-saving targets.

[0108] The group's management platform is geared towards the management of construction units, enabling horizontal comparison and overall monitoring across multiple construction sites. The platform displays the energy consumption intensity ranking of each construction site (measured in kilowatt-hours of electricity consumption per 10,000 yuan of output value), monthly comparisons of the group's overall energy consumption trends, deviations between the energy-saving target completion rate of each construction site and the group's target, and early warnings for high-energy-consuming construction sites—when the energy consumption intensity of a construction site exceeds 50% of the group's average, the system automatically marks it in red as a warning, prompting the group's management to pay close attention and intervene.

[0109] Regarding automatic report generation, the system automatically generates a daily energy consumption report at 8:00 AM every day, covering daily electricity and water consumption, abnormal event records, energy savings, number of unattended power outages, and the number of energy-saving strategies implemented. On the first day of each month, a monthly report for the previous month is automatically generated, including a monthly energy consumption summary, year-on-year and month-on-month comparisons, detailed analysis of abnormal events, effectiveness evaluation of various energy-saving measures, energy-saving recommendations for the following month, and a detailed breakdown of cost savings. The system also supports the generation of specialized reports as needed, including abnormal event investigation reports (specific analyses of events such as water leaks, electricity theft, or equipment malfunctions), energy-saving measure effectiveness evaluation reports, and energy consumption budget execution reports. All reports and data can be exported to multiple formats such as Excel, PDF, CSV, and JSON, and are integrated with enterprise ERP systems and project management platforms via API interfaces to achieve cross-system data sharing and business collaboration.

[0110] The advantages of the construction site energy consumption monitoring system in this application embodiment are as follows:

[0111] 1. Addressing the shortcomings of traditional manual meter reading, such as coarse data granularity, delayed updates, and inability to accurately pinpoint responsible areas and equipment for energy consumption anomalies, this system establishes a three-tiered power consumption monitoring system encompassing the main inlet line, zones, and key equipment, coupled with a two-tiered water consumption monitoring architecture for the main pipeline and zones. Differentiated data collection frequencies allow for a complete departure from the rudimentary daily and weekly manual meter reading model. Simultaneously, leveraging edge gateways for real-time local processing of time-series energy consumption data, and utilizing the InfluxDB time-series database, it enables millisecond-level queries of massive energy consumption data. Management personnel can view real-time energy consumption parameters for every circuit and every high-power device. The system allows for step-by-step tracing of energy consumption anomalies from the main circuit to the zone and then to individual devices. Once energy consumption exceeds limits or power consumption anomalies occur, the fault location, abnormal time period, and corresponding construction equipment can be directly identified without requiring manual full-area investigation. The energy consumption anomaly detection cycle is reduced from monthly reviews to minutes, fundamentally solving the problem of delayed detection of energy waste and irreversible losses. Furthermore, it enables precise breakdown of energy costs by region and equipment, meeting the needs of refined project cost accounting.

[0112] 2. Addressing the industry pain point of asynchronous operations across multiple areas on construction sites and the high energy consumption rate of 20%-35% due to unmanned operation, this system integrates UWB high-precision personnel positioning, video AI human figure recognition, and multimodal perception technology from access control gates. It accurately determines the real-time presence status of each construction area and incorporates an unmanned delay confirmation mechanism to avoid accidental power outages caused by personnel briefly leaving their posts, balancing energy efficiency with construction convenience. The system divides construction sites into four categories based on their functional differences: completely de-energized zones, partially de-energized zones, non-de-energized safety zones, and large idle equipment control zones. Differentiated automatic energy-saving strategies are then applied to each category, improving the problem of routine idle standby energy waste on construction sites.

[0113] 3. Addressing the shortcomings of traditional methods that rely on manual inspections for electrical and water leakage faults, resulting in significant delays in detection and both economic losses and construction safety hazards, this system establishes a four-dimensional energy consumption anomaly detection engine. Combining statistical Z-Score algorithms, AI baseline prediction deviation analysis, electrical feature recognition, and water usage anomaly detection, it comprehensively monitors common energy consumption faults on construction sites. The system can automatically issue alarms and output root cause analysis reports within a short time after a fault occurs, greatly improving fault response efficiency compared to the original manual troubleshooting cycle of several days to weeks. On the one hand, it promptly stops the water and electricity cost losses caused by continuous water and electrical leakage; on the other hand, it proactively identifies electrical hazards such as aging lines and excessive power grid harmonics, avoiding construction safety accidents such as electric shock, fire, and water accumulation in foundation pits, achieving dual protection for energy consumption control and safe production on construction sites.

[0114] 4. Addressing the pain points of traditional models, such as lack of energy consumption data analysis capabilities, excessive electricity costs due to concentrated startup of high-power equipment, and transformer overload, this system builds a random forest energy consumption baseline model and an LSTM time-series energy consumption prediction model based on historical energy consumption data from the 30 days prior to project commencement, combined with multi-dimensional characteristics such as construction procedures, weather, and the number of personnel on site. This allows for accurate prediction of hourly energy consumption over the next 24 hours and overall energy consumption changes over 7 days. The system automatically outputs two core energy-saving optimization strategies: first, intelligent peak-valley pricing for off-peak hours, automatically suggesting that high-energy-consuming processes such as concrete pouring be moved to off-peak hours, significantly saving electricity costs for single high-energy-consuming processes; second, staggered startup of high-power equipment, dispersing the inrush current of tower cranes, concrete pumps, and construction elevators, reducing maximum electricity demand, and effectively preventing transformer overload tripping.

[0115] 5. Addressing the shortcomings of fragmented energy consumption data across multiple construction sites within the group, the inability to benchmark horizontally, and the lack of replicable energy-saving experiences, this system establishes a unified cloud-based energy management platform. It aggregates electricity, water, equipment energy consumption, and energy-saving data from all ongoing construction sites, automatically generating energy intensity rankings for each project, monthly year-on-year and month-on-month energy consumption reports, and energy-saving target completion rate dashboards. Group managers can intuitively compare energy consumption per unit of output value and per unit of construction area across different projects, quickly identifying high-energy-consuming and inefficient projects and issuing rectification requirements. Simultaneously, it comprehensively accumulates mature energy-saving strategies and abnormal fault handling solutions, enabling one-click replication and promotion of standardized energy-saving management experience. This fills the gap in the group-level digital energy management and control, helping the construction company achieve its overall green construction goals.

[0116] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or apparatus.

[0117] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A construction site energy consumption monitoring system, characterized in that, include: The energy consumption monitoring unit is used to collect the electrical energy parameters and water consumption parameters of the construction site, obtain energy consumption data, and upload the energy consumption data to the edge gateway; The presence sensing unit is used to detect the presence of personnel in various areas of the construction site in real time, output the number of personnel in each area, and upload the personnel data to the edge gateway. The edge gateway is used to upload the energy consumption data and personnel data to the cloud platform, and to determine the personnel situation in each area based on the number of personnel in each area. When a certain area is unoccupied and the duration of unoccupancy reaches a threshold, a power outage command is generated and sent to the execution unit corresponding to that area. An execution unit is used to perform a power-off operation on the corresponding area according to the power-off command; The cloud platform includes: The energy consumption prediction module is used to predict the future energy consumption of each area based on the energy consumption data of the construction site, personnel data of each area, construction plan and weather data, and obtain the energy consumption prediction results. Anomaly detection module is used to perform statistical threshold anomaly detection, model prediction deviation anomaly detection, electrical characteristic anomaly detection, and water usage anomaly detection based on the energy consumption data, and obtain anomaly detection results; The energy-saving strategy optimization module generates energy-saving optimization suggestions based on the energy consumption prediction results and the anomaly detection results.

2. The construction site energy consumption monitoring system according to claim 1, characterized in that, The energy consumption monitoring unit includes three-phase smart meters installed on the incoming side of the main distribution box at the construction site and in each distribution box, as well as current recorders for high-power equipment; the high-power equipment is equipment whose rated power exceeds the target power; the energy parameters include voltage data, current data, power data, energy data, power factor, and total harmonic distortion rate; The energy consumption monitoring unit also includes an electromagnetic flow meter installed on the main water inlet pipe, remote water meters in each area, and ground water sensors deployed below pipe joints, in areas where valve groups are concentrated, and in low-lying areas.

3. The construction site energy consumption monitoring system according to claim 1, characterized in that, The edge gateway is specifically used to determine the personnel situation in each area based on the number of people in each area. If an area is unoccupied and the duration of unoccupancy reaches a threshold, determine whether the area is a non-power-off zone. If so, skip the power-off process. If not, the system will determine whether there are ongoing construction tasks in the area based on the construction plan. If there are ongoing construction tasks in the area, the power outage process will be skipped. If there is no construction task in the area, it checks whether there is any equipment in the area that needs to be continuously powered on for cooling. If so, the power-off process is skipped; otherwise, a power-off command is generated and sent to the corresponding execution unit in the area.

4. The construction site energy consumption monitoring system according to claim 1, characterized in that, The edge gateway is also used to generate a power restoration command to the execution unit corresponding to the power outage area when it detects that personnel have entered the power outage area, so that the execution unit performs the power restoration operation of the power outage area according to the power restoration command.

5. The construction site energy consumption monitoring system according to claim 1, characterized in that, The energy consumption prediction module is specifically used to input historical energy consumption data sequences, weather forecast data, construction plans and holiday data into the energy consumption prediction model to predict energy consumption and obtain the hourly electricity consumption of each region and the total daily electricity consumption within the future preset days.

6. The construction site energy consumption monitoring system according to claim 5, characterized in that, The energy consumption prediction module is also used to extract features from personnel data, construction plans, and weather data of each area within a preset time period after the start of construction, to obtain time features, construction features, personnel features, and weather features; using time features, construction features, personnel features, and weather features as input data, and using daily energy consumption data within the preset time period as training targets, a random forest regression model is trained to obtain a baseline model; the baseline model is used to predict the expected energy consumption of each area under normal construction conditions at the construction site.

7. The construction site energy consumption monitoring system according to claim 6, characterized in that, The model prediction bias anomaly detection process includes: Calculate the deviation between the actual energy consumption and the expected energy consumption predicted by the baseline model; When the deviation value exceeds the first deviation threshold but does not exceed the second deviation threshold, and continues for a first preset duration, a first-level alarm is triggered; When the deviation value exceeds the second deviation threshold but does not exceed the third deviation threshold, and continues for a second preset duration, a second-level alarm is triggered; When the deviation value exceeds the third deviation threshold, a third-level alarm is immediately triggered; The deviation threshold is a preset percentage of the expected energy consumption, the first preset duration is greater than the second preset duration, and the levels of the first-level alarm, the second-level alarm, and the third-level alarm increase sequentially.

8. The construction site energy consumption monitoring system according to claim 1, characterized in that, The abnormal water usage detection process includes: Pipeline leak detection is performed based on the instantaneous flow rate of water pipes in each zone, including: The instantaneous flow rate of water pipes in each zone during the nighttime shutdown period is obtained from the energy consumption data. When the instantaneous flow rate continuously exceeds the preset nighttime baseline flow rate threshold, it is determined to be a suspected pipe leak. Alternatively, the instantaneous flow rate and network pressure of the water supply pipeline can be acquired simultaneously. If the instantaneous flow rate does not increase but the network pressure fluctuates abnormally, it is determined to be a suspected pipeline leak. Alternatively, calculate the difference between the total water inflow at the construction site and the sum of the water flow in each zone. If the difference exceeds a preset percentage threshold of the total water inflow, it is determined to be a suspected pipe leak. Water waste detection is performed based on the cumulative water consumption data and construction status information at the construction site, including: If water continues to be used in a certain area after construction is completed and the duration exceeds the preset time threshold, it is judged as a wasteful behavior of leaving the valve open. Alternatively, when the single water consumption in a certain area exceeds a preset multiple threshold of the historical average water consumption of similar processes, it is judged as a water waste behavior.

9. The construction site energy consumption monitoring system according to claim 2, characterized in that, The electrical anomaly detection process includes: The imbalance of the three-phase current is calculated based on the current data. When the imbalance exceeds the preset imbalance threshold and continues for a third preset time, it is determined to be a suspected leakage current and a leakage current alarm is output. At the same time, the neutral current is acquired. When the zero-sequence current exceeds the preset zero-sequence current threshold, the leakage current protector is triggered to check and remind. The operating power and rated power of each monitoring device are obtained from the energy consumption data. When the operating power of the monitoring device is lower than a preset percentage of the rated power and exceeds the idle time threshold corresponding to the monitoring device, the device is determined to be in an idle operating state. Harmonic detection is performed based on the total harmonic distortion rate: when the total harmonic distortion rate exceeds the preset harmonic threshold, it is determined that the harmonics exceed the standard and a harmonic alarm is output.

10. The construction site energy consumption monitoring system according to claim 1, characterized in that, The system also includes: The mobile terminal is used to obtain energy consumption data, anomaly detection results and energy-saving optimization suggestions from the cloud platform, and automatically generate reports according to a preset time period. The reports include data on electricity consumption, water consumption, abnormal events and savings. The generated report is exported as a preset format file and then connected to an external business system via an API interface.

Images