An intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling

Abstract

The present application relates to the field of hazardous waste recycling and Internet of Things technology, and discloses a kind of intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling, which is composed of Internet of Things intelligent recycling terminal deployed in waste production site, cloud AI scheduling platform responsible for data processing and intelligent decision-making and mobile terminal execution subsystem supporting field operation.The present application enables stores to save labor hours and improve labor efficiency every month and every operating personnel without manual liquid level inspection, phone call collection and manual report of joint order.The average daily coverage rate of single vehicle is increased by 4 times through AI scheduling, transportation response time is shortened, electronic joint order is automatically generated and reported, the report rate is increased from less than 30% of the industry average to 100%, there is no environmental protection penalty record, and the store space occupation and communication cost are reduced.The procurement cost of detection equipment is lower than that of similar products in the market.In addition, the system automatically calculates carbon reduction data, the annual carbon reduction amount of single store is high, and the carbon emission reduction effect is significant.

Description

Technical Field

[0001] This invention relates to the fields of hazardous waste recycling and Internet of Things (IoT) technology, specifically to an intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling. Background Technology

[0002] Waste mineral oil refers to mineral oil products that have been phased out due to performance deterioration, pollution, or no longer meeting specific application requirements during use. These oils typically contain a certain amount of contaminants, including moisture, mechanical impurities, and chemical contaminants. Waste mineral oil, hereinafter referred to as "waste oil," is a hazardous waste (HW08 category) clearly defined in the National Hazardous Waste List. HW08 category includes waste mineral oil and mineral oil-containing waste listed in the National Hazardous Waste List. Large quantities of waste mineral oil are generated daily in waste-generating units such as auto repair shops and 4S stores. Waste mineral oil generated by auto repair shops and 4S stores includes waste engine oil, waste transmission oil, waste brake fluid, waste power steering fluid, and waste coolant. Currently, in waste mineral oil recycling scenarios, the separate deployment of liquid level monitoring and component detection in existing technologies leads to asynchronous data timestamps and a lack of spatial correlation, resulting in data silos. Furthermore, existing liquid level prediction algorithms do not consider the impact of oil composition changes on the rate of liquid level change. Traditional recycling methods also suffer from the following pain points: Manual inspection of liquid levels is inefficient and prone to overflow. The composition of oil (water content, impurities) cannot be quantified, leading to chaotic pricing. Illegal mixing (such as mixing with other hazardous waste) is difficult to detect. Recycling and dispatch rely on telephone, resulting in unreasonable route planning and high logistics costs. Manually filling out hazardous waste transfer forms leads to low reporting rates, many errors and omissions, and great difficulty in environmental supervision. Existing patents disclose liquid level monitoring devices based on the Internet of Things (application number 202020826025.8). Current technologies are limited to liquid level alarms and do not form a closed-loop system with component detection, AI scheduling, and automatic order processing. In particular, in the field of hazardous waste recycling, there is a lack of a comprehensive solution that can simultaneously solve the four major problems of safety monitoring, component evaluation, intelligent scheduling, and compliant reporting. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling. This system solves the problems of existing technologies, such as reliance on manual labor, unclear composition, low scheduling efficiency, and high compliance risks in waste mineral oil recycling, as well as the problems of low efficiency of manual inspection, inability to quantify oil composition, unreasonable scheduling path planning, and numerous errors and omissions in manual form filling.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a waste oil intelligent recycling system based on multi-source sensor fusion and AI scheduling, comprising: The Internet of Things (IoT) smart recycling terminal is deployed at the waste mineral oil generation point. It is configured to enable an optical liquid level sensor and a near-infrared spectral detection sensor to collect waste mineral oil liquid level data and composition data through the same edge computing unit and mark them with a unified timestamp. The composition data includes water content and impurity ratio. The cloud-based AI scheduling platform is connected to the IoT smart recycling terminal and is configured to receive liquid level data and composition data with a unified timestamp. Based on the time series prediction algorithm, it predicts the overflow time according to the liquid level data change trend, and determines the waste mineral oil value level according to the water content and impurity ratio in the composition data, and generates scheduling instructions that include recycling priority and disposal channel recommendations. The mobile execution subsystem is connected to the cloud-based AI scheduling platform and is configured to receive the scheduling instructions and display the optimal recycling path, scan the QR code on the oil drum to complete the handover confirmation and trigger the generation of an electronic manifest. The edge computing unit is configured to perform Kalman filtering smoothing on the liquid level data, remove outliers from the component data, and trigger a local alarm when the liquid level exceeds a threshold or the component is abnormal.

[0005] As a preferred technical solution of the present invention, the hardware architecture of the IoT smart recycling terminal consists of a sensor array unit, a data processing unit, a communication transmission unit, and a human-computer interaction unit. The sensor array unit integrates an optical liquid level sensor and a near-infrared spectroscopy detection sensor to collect data on the liquid level and composition of waste mineral oil. The data processing unit uses a low-power embedded processor to process the sensor data. The communication transmission unit has a built-in wireless communication module that uploads the collected data to the cloud server in real time and receives instructions. The human-machine interaction unit includes a QR code label and a status indicator light.

[0006] As a preferred technical solution of the present invention, the IoT smart recycling terminal includes a multi-source sensor fusion module, an edge computing and data preprocessing module, and a wireless communication and secure transmission module; The multi-source sensor fusion module integrates an optical liquid level sensor and a near-infrared spectral detection sensor to achieve dual-dimensional perception of the quantity and quality of waste mineral oil. The optical liquid level sensor calculates the liquid level height by emitting an infrared beam and receiving the reflected signal, while the near-infrared spectral detection sensor analyzes multiple component indicators of waste mineral oil in real time and quickly determines the quantitative indicators of waste mineral oil. The edge computing and data preprocessing module uses a low-power processor, runs a streamlined embedded operating system, and has local data processing capabilities to achieve data filtering and outlier handling, data compression and breakpoint resume, as well as local alarm triggering. Data filtering and outlier handling: Kalman filtering algorithm is used to smooth the liquid level data, and box plot method is used to remove outliers from the component detection data; Data compression and breakpoint resume: When the network connection is unstable or the bandwidth is limited, the collected data is cached and compressed locally, and automatically resumed after the network is restored; Local alarm triggering: In case of emergency, it can respond quickly independently of the cloud, triggering local audible and visual alarms and executing preset emergency operations; The wireless communication and secure transmission module supports mainstream communication standards and automatically switches according to network coverage. It uses the MQTT protocol to connect to the cloud platform. In terms of data transmission security, it uses the TLS encryption protocol to achieve end-to-end encryption, adopts a two-way authentication mechanism based on digital certificates, and performs irreversible encryption processing on sensitive information.

[0007] As a preferred technical solution of the present invention, the cloud AI scheduling platform includes four storage engines: time-series database, relational database, object storage and graph database. The platform includes a predictive scheduling module, a component assessment and value grading module, a carbon footprint accounting module and a compliance management module. The predictive scheduling module uses historical liquid level change curves and time series prediction algorithms to estimate the specific time when each oil drum will reach the overflow threshold and generate recovery tasks in advance. The system employs two prediction models, ARIMA and LSTM, and automatically selects the optimal model or a combination of models based on the actual liquid level change characteristics of each waste-generating unit. It includes four key steps: historical data accumulation, trend analysis and prediction, recycling task generation, and scheduling instruction issuance.

[0008] As a preferred technical solution of the present invention, the component evaluation and value grading module transforms the near-infrared spectral data collected by the Internet of Things terminal into a value indicator with business significance, while identifying illegal mixing behavior, triggering a warning of hazardous waste mixing, and undertaking three major functions: waste mineral oil quality detection, value grading assessment, and abnormal component warning. In terms of waste mineral oil quality testing, it receives raw near-infrared spectral data, performs spectral preprocessing, feature extraction and quantitative analysis, and outputs quantitative indicators such as water content, impurity ratio and oil purity. In terms of value grading, waste mineral oil is divided into three value grades: high, medium, and low. High-value waste mineral oil: water content <1%, impurities <2%, purity >98%, matched with high-grade treatment channels; medium-value waste mineral oil: water content 1%-5%, impurities 2%-10%, requires pretreatment before use; low-value waste mineral oil: water content >5%, impurities >10%, requires special treatment. Regarding the early warning of abnormal components, a normal fluctuation range model for waste mineral oil components is established, and an alarm is triggered when the detected component index deviates from the normal range.

[0009] As a preferred technical solution of the present invention, the component evaluation and value grading module dynamically adjusts the weight of each quality indicator in the value evaluation by constructing an adaptive value evaluation model. The value assessment model is a component-value mapping model based on machine learning. The input of the model is various quality indicators of waste mineral oil, and the output is the value grade of waste mineral oil or the predicted value of recycling revenue. The value assessment model adopts a multi-input neural network architecture with attention mechanism, which includes three parts: feature embedding layer, interaction learning layer and output prediction layer; The feature embedding layer converts the original values ​​of each quality indicator into a dense vector representation suitable for neural network processing, while learning the preliminary correlation between each indicator. The interactive learning layer employs a self-attention mechanism to adaptively learn the complex interactive effects between different quality indicators; The output prediction layer summarizes the results of interactive learning and outputs the value level classification results or the value amount regression prediction.

[0010] As a preferred technical solution of the present invention, the value assessment model training adopts an incremental learning strategy. When a new recycling transaction is completed, the component data of this transaction is used as training samples, and the actual disposal revenue fed back by the disposal factory is used as labels, which are added to the historical training dataset. Every fixed period, the model will be incrementally fine-tuned on the newly added data to update the network parameters. The results output by the value assessment model and the original rule-based assessment system form a dual-track parallel mechanism: The component detection data uploaded by the terminal is first quickly assessed by the rule evaluation system to give a preliminary value level conclusion; Subsequently, the data is fed into a neural network model for in-depth evaluation, providing the model's predicted value level or profit forecast. When the two conclusions are consistent, the conclusion is adopted; when the two conclusions are inconsistent, the discrepancies are recorded for manual review, and the final evaluation is given by combining the results of the two.

[0011] As a preferred technical solution of the present invention, the carbon footprint accounting module automatically calculates the carbon emission reduction of each batch of waste mineral oil recycling based on the carbon emission factors of each link in the entire recycling chain and combined with actual operation data, and generates a verifiable carbon reduction report. Carbon accounting methods follow international standard frameworks, classifying carbon emission sources into two main categories: direct emissions and indirect emissions; The carbon emission reduction accounting logic is based on a comparative analysis of waste mineral oil recycling and direct disposal, calculating the amount of primary resources replaced by the recycling of this batch of waste mineral oil and the corresponding carbon emission reduction. The generated carbon footprint report can serve as the data basis for waste-generating entities to prepare ESG reports and apply for carbon incentive rewards; The compliance management module achieves a 100% form completion rate and undertakes three major functions: automatic generation of electronic forms, automatic submission to the regulatory system, and automatic maintenance of electronic ledgers. Regarding the automatic generation of electronic manifests: information on waste-generating units, drivers, vehicles, waste mineral oil, and disposal plants is automatically extracted to generate electronic manifests; Regarding the automatic reporting of regulatory systems: By connecting with the provincial solid waste supervision and management information system through data interfaces, the online application and approval process for hazardous waste transfer is automatically completed; Regarding the automatic maintenance of electronic ledgers: an electronic hazardous waste management ledger is established for each waste-generating unit, and it is automatically updated every time a recycling or transfer is completed.

[0012] As a preferred technical solution of the present invention, the mobile execution subsystem includes a driver-side APP module and a waste-generating mini-program module; The driver-side APP module is a mobile operation platform for recycling workers, integrating core functional units including task reception, route navigation, QR code scanning, and handover confirmation. The task receiving unit is responsible for receiving recycling tasks issued by the cloud platform. The path navigation unit is deeply integrated with mainstream map navigation services and accesses the optimal recycling route calculated by the cloud AI scheduling platform. The QR code scanning and handover confirmation unit scans the QR code of the oil drum to automatically identify relevant information and takes a photo of the replaced drum as a handover certificate after the empty drum is replaced.

[0013] As a preferred technical solution of the present invention, the waste-generating end mini-program module is a mini-program application for waste mineral oil generating units, providing the following core functions; The liquid level monitoring unit displays the current liquid level height and liquid level change trend of each oil drum in the store in real time, and pushes a reminder notification when the liquid level is close to the overflow threshold; The history query section supports querying the store's historical recycling records by time range; The manifest management unit displays the status of all electronic manifests for hazardous waste transfer in this store, and supports viewing manifest details and downloading electronic archives; The carbon reduction contribution unit displays the store's cumulative waste mineral oil recovery volume and corresponding carbon emission reduction contribution in the form of visual charts, supporting the preparation of ESG reports.

[0014] Compared with existing technologies, this invention provides a smart waste oil recycling system based on multi-source sensor fusion and AI scheduling, which has the following advantages: 1. By integrating optical liquid level sensors and spectral composition detection sensors into the same smart terminal, the quantity and quality of waste mineral oil can be obtained simultaneously through multi-source sensor fusion. This not only determines when the tank is full, but also what is full. High-purity, low-impurity waste mineral oil is prioritized for dispatching and matched with higher-value disposal channels. Based on component-based differentiated scheduling and value assessment, the overall profitability of the recycling network is improved. Furthermore, by using machine learning to predict liquid level change trends, predictive scheduling algorithms are applied in the field of hazardous waste recycling, transforming passive response into proactive prevention and minimizing environmental risks such as overflow and leakage. The system integrates liquid level, composition, logistics trajectory, manifest filling, and carbon footprint accounting into a closed-loop system, achieving full closed-loop compliance automation. It realizes digital supervision of the entire process from waste generation to disposal, solves regulatory pain points, and links the carbon emission factors of each link in the recycling process with real operational data collected by the Internet of Things. By combining the carbon footprint accounting algorithm with the hazardous waste recycling chain, it generates verifiable carbon reduction certificates for each waste-generating unit and supports ESG reporting. This method acquires liquid level and oil composition data through multi-source sensor fusion, uses AI algorithms to achieve predictive scheduling, and automatically generates electronic manifests and ledgers. Ultimately, it achieves safe, efficient, and fully compliant waste mineral oil recycling, solving the problems of existing technologies such as reliance on manual labor, unclear composition, low scheduling efficiency, and high compliance risks in waste mineral oil recycling. Furthermore, through the synergy of spatiotemporal synchronous acquisition by multi-source sensors and AI prediction models, it solves the problems of overflow risk and inefficient scheduling caused by the lag in traditional manual inspection and component detection.

[0015] 2. The predictive scheduling module can predict the overflow time in advance, so as to complete the recycling operation before the oil drum is actually full, completely eliminating the waste mineral oil leakage and environmental pollution accidents caused by overflow. From the perspective of operational efficiency, predictive scheduling enables recycling vehicles to complete multiple recycling tasks in the optimal route sequence, reducing empty driving mileage and waiting time, and shortening transportation response time. From the perspective of resource optimization, batch scheduling based on prediction results can improve vehicle loading rate and reduce the marginal cost of a single recycling. Through an adaptive value assessment model, in terms of pricing decisions, the dynamic model can capture value-influencing factors that fixed rules cannot identify, providing more accurate pricing suggestions and avoiding the problem of undervaluing high-quality waste mineral oil or overvaluing low-quality waste mineral oil. In terms of scheduling optimization, the waste mineral oil value predicted by the model can serve as an important basis for prioritizing recycling, giving priority to dispatching orders for high-value waste mineral oil and matching it with high-grade disposal channels, thereby improving the overall profitability of the recycling network. In terms of anomaly detection, the model can learn the composition-value mapping relationship of normal waste mineral oil. When the actual transaction price deviates significantly from the model prediction, it triggers an abnormal transaction warning, helping to detect fraudulent transactions or profit transfer behaviors.

[0016] 3. Through the collaborative application of IoT smart recycling terminals, cloud-based AI scheduling platforms, and mobile execution subsystems, stores no longer need to manually inspect liquid levels, make phone calls to collect waste, or manually fill out forms. This saves time per month per operator, improving labor efficiency. AI scheduling increases the daily coverage of recycling points by four times, shortens transportation response time, and improves recycling efficiency. The automatic generation and submission of electronic forms increases the reporting rate from the industry average of less than 30% to 100%, resulting in no environmental penalties and improved compliance. At the same time, it reduces store space occupancy, communication costs, and the procurement cost of testing equipment is 60% lower than similar products on the market. In addition, the system automatically calculates carbon reduction data, with a single store achieving an annual carbon reduction of up to 50 tons of CO2 equivalent, demonstrating significant carbon emission reduction effects. Attached Figure Description

[0017] Figure 1 This is an architecture diagram of the intelligent waste oil recycling system of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. However, it should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0019] In the description of this invention, it should be noted that when an element is referred to as being "fixed to" or "set on" another element, it can be directly on or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to or indirectly connected to the other element.

[0020] In the description of this invention, it should be noted that the terms "center," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. "Several" means one or more, unless otherwise explicitly specified.

[0021] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0022] Please see Figure 1 The present invention provides the following technical solution: a waste oil intelligent recycling system based on multi-source sensor fusion and AI scheduling, which is applied to intelligent monitoring, predictive scheduling and full-process compliance management of waste mineral oil recycling in the automotive aftermarket. It consists of three parts: an IoT intelligent recycling terminal deployed at the waste generation site, a cloud AI scheduling platform that undertakes data processing and intelligent decision-making, and a mobile terminal execution subsystem that supports on-site operations. The IoT smart recycling terminal, cloud AI scheduling platform, and mobile execution subsystem achieve real-time data interaction and collaborative work through 4G / 5G wireless communication network, forming a full-chain digital closed-loop management system from waste source to end disposal; The IoT smart recycling terminal, as the sensing front end of the system, is deployed at waste mineral oil generation points, including below auto repair bays and dedicated recycling areas. It serves as the data acquisition entry point for the entire waste mineral oil recycling IoT system. It adopts an integrated design, integrating sensors, communication modules, and processing units into a protective shell. They are connected through an internal bus to achieve intelligent monitoring and management of waste mineral oil recycling. Its hardware architecture consists of a sensor array unit, a data processing unit, a communication transmission unit, and a human-machine interaction unit. The sensor array unit integrates an optical liquid level sensor and a near-infrared spectroscopy detection sensor to collect data on the liquid level and oil composition of waste mineral oil, enabling dual-dimensional perception of the quantity and quality of waste mineral oil. The data processing unit uses a low-power embedded processor to perform sensor data preprocessing, outlier filtering, and data compression. The communication transmission unit has a built-in 5G wireless communication module, which uploads the collected data to the cloud server in real time and receives instructions from the cloud. The human-computer interaction unit includes a QR code label and status indicator lights, which facilitates on-site operators to quickly identify the terminal status; The IoT smart recycling terminal includes a multi-source sensor fusion module, an edge computing and data preprocessing module, and a wireless communication and secure transmission module; The multi-source sensor fusion module integrates an optical liquid level sensor and a near-infrared spectral detection sensor to achieve dual-dimensional perception of the quantity and quality of waste mineral oil. It integrates the originally separate liquid level monitoring and component detection functions into the same terminal, realizing the spatiotemporal consistency of waste mineral oil quantity and quality data acquisition, and achieving unified timestamp labeling and spatial positioning association. This provides a high-quality data foundation for subsequent component-value mapping analysis and differentiated scheduling decisions. Spatiotemporal synchronization is specifically achieved through a GPS positioning module and the Time Synchronization Protocol (NTP) to ensure that the timestamp error between liquid level data and component data is less than 1 second and the spatial positioning error is less than 5 meters. The optical liquid level sensor adopts a non-contact measurement principle and is installed on the inside of the top of the oil drum. It calculates the liquid level by emitting an infrared beam and receiving the reflected signal. The sensor automatically collects liquid level data every 2 hours with a measurement accuracy of millimeters, accurately sensing the filling degree of the oil drum. When the liquid level reaches the preset alarm threshold, i.e., 90% liquid level, the collection frequency is increased to once every 15 minutes. Near-infrared spectroscopy sensors analyze multiple component indicators of waste mineral oil in real time. The wavelength range of the near-infrared spectroscopy sensor is 900-1700nm, the resolution is 10nm, and the detection cycle is 30 seconds. By emitting near-infrared light of a specific wavelength into the waste mineral oil sample, the position and intensity of the characteristic peaks of the absorption spectrum are analyzed to quickly determine quantitative indicators such as water content, impurity ratio, and oil purity of the waste mineral oil. Near-infrared spectroscopy detection technology has the advantages of fast detection speed, no need for chemical reagents, and no secondary pollution, making it particularly suitable for rapid on-site detection scenarios. The edge computing and data preprocessing module is responsible for data preprocessing and preliminary analysis. It adopts a low-power ARM Cortex-M series processor, runs a streamlined embedded Linux operating system, and has local data processing capabilities, realizing data filtering and outlier handling, data compression and breakpoint resume, as well as local alarm triggering. Data filtering and outlier handling: Noise and outliers may occur in the raw sensor data due to equipment vibration and signal interference. Kalman filtering algorithm is used to smooth the liquid level data, and box plot method is used to remove outliers from the component detection data to ensure the quality of the uploaded data. Data compression and breakpoint resume: When the network connection is unstable, the collected data is cached and compressed locally, and automatically retransmitted after the network is restored, ensuring the integrity and continuity of the data; Local alarm triggering: In case of emergencies such as excessive liquid level or abnormal composition, the system can respond quickly and independently of the cloud, triggering local audible and visual alarms and executing preset emergency operations, including closing the air intake valve. The wireless communication and secure transmission module supports mainstream 5G communication standards and automatically switches according to network coverage to ensure reliable connection in various network environments. It uses the MQTT protocol to connect to the cloud platform. In terms of data transmission security, the TLS encryption protocol is adopted, and all data uploaded to the cloud is encrypted end-to-end to prevent data from being stolen and tampered with during transmission. A two-way authentication mechanism based on digital certificates is also adopted, where the cloud platform verifies the identity of the terminal while the terminal also verifies the legitimacy of the cloud platform to prevent man-in-the-middle attacks. At the same time, data desensitization processing is adopted, and sensitive information is irreversibly encrypted before being uploaded. The SHA-256 hash algorithm is used to desensitize the device serial number, and the AES-256 algorithm is used to encrypt the transmitted data. Sensitive information includes the device serial number and location information, ensuring that user privacy will not be exposed in the event of data leakage in the cloud. The cloud-based AI scheduling platform undertakes the core functions of data aggregation, storage and computing, intelligent analysis and scheduling decision-making. It adopts a distributed data storage architecture and includes four storage engines: time-series database, relational database, object storage and graph database. The time-series database stores time-series data on liquid level and composition collected by sensors; the relational database stores structured business data including waste-generating units, recycling vehicles, disposal plants, and user accounts; the object storage is used to store unstructured files including electronic manifest images and barrel-changing photos; and the graph database stores the relationships between waste-generating units, recycling routes, and disposal plants, supporting complex path queries and scheduling optimization. The cloud-based AI scheduling platform includes a predictive scheduling module, a component assessment and value grading module, a carbon footprint accounting module, and a compliance management module. The predictive scheduling module uses historical liquid level change curves and time series prediction algorithms to estimate the specific time when each oil drum will reach the overflow threshold, which is 90% of the liquid level in each oil drum. This allows for the generation of recovery tasks in advance, transforming passive response into proactive prevention and minimizing the environmental risks of overflow and leakage. The trigger threshold for predictive scheduling is: when the prediction model outputs that the oil drum will reach 90% of the liquid level within 48 hours, a recovery task is generated. The time series prediction algorithms used include two models: ARIMA and LSTM. ARIMA is an autoregressive integral moving average model, and LSTM is a long short-term memory network. The ARIMA model is suitable for scenarios where the liquid level change pattern is relatively stable and the seasonality is obvious. It has high computational efficiency and strong model interpretability. The long short-term memory network LSTM model is suitable for scenarios where the liquid level change is affected by multiple factors and there are complex nonlinear relationships. It has higher prediction accuracy but consumes more computational resources. A model ensemble strategy is adopted to automatically select the optimal prediction model based on the actual liquid level change characteristics of each waste-generating unit in order to achieve the best prediction results. The specific workflow includes four key steps: historical data accumulation, trend analysis and prediction, task generation, and scheduling instruction issuance. Historical data accumulation: The system continuously receives liquid level data uploaded by each terminal and constructs a historical curve of liquid level change for each oil drum. The longer the data accumulation time, the higher the accuracy of the prediction model. Trend Analysis and Forecasting: Based on a forecasting model trained on historical data, input the current liquid level data and the latest sensor readings, and output the predicted liquid level change curve for the next few days and the expected time of overflow. Recycling task generation: When the prediction model determines that a certain oil drum will reach the overflow threshold within the next 24-48 hours, the system automatically generates a recycling task and prioritizes the tasks based on multiple factors, including the urgency of the task, the current location of the vehicle, and the capacity of the disposal plant. Dispatch instruction issuance: The task is issued to the mobile execution subsystem via API interface. After receiving the task, the driver's APP proceeds to execute it according to the optimal route planned by AI. The component assessment and value grading module transforms near-infrared spectral data collected by IoT terminals into business-meaning value indicators, providing pricing basis for downstream disposal plants. It also identifies illegal mixing behavior, triggers hazardous waste mixing warnings, and undertakes three major functions: waste mineral oil quality testing, value grading assessment, and abnormal component warning. In the quality testing of waste mineral oil: raw near-infrared spectral data is received, and after spectral preprocessing, feature extraction and quantitative analysis, the spectral preprocessing includes smoothing, noise reduction and baseline correction, feature extraction includes principal component analysis and characteristic wavelength selection, and quantitative analysis includes partial least squares regression. The output is quantitative indicators including water content, impurity ratio and oil purity. Compared with traditional laboratory testing methods, these indicators have the advantages of fast detection speed, low cost and continuous online monitoring, and are suitable for large-scale application scenarios. Traditional methods include distillation and Karl Fischer method. Regarding value rating: A value assessment model is constructed based on oil quality indicators, classifying waste mineral oil into three value levels: high, medium, and low. The training sample size of the value assessment model is no less than 1,000 sets of historical transaction data, and the mean squared error loss function is adopted with a learning rate of 0.001. High-value waste mineral oil is characterized by a water content of less than 1%, an impurity ratio of less than 2%, and an oil purity of more than 98%. It is suitable for high-grade disposal channels and can be processed using high-grade regeneration processes to produce high-quality regenerated lubricating oil base oil with high economic value. Medium-value waste mineral oil has a water content of 1%-5% and an impurity ratio of 2%-10%. After pretreatment, it can be used for lower-grade regeneration products. Low-value waste mineral oil has a water content of more than 5%, an impurity ratio of more than 10%, and contains other harmful impurities. It requires special processing. The value rating results not only affect the recycling price but also determine the downstream disposal channels and transportation priorities that the batch of waste mineral oil should be matched with. Regarding abnormal component early warning: The module establishes a normal fluctuation range model for waste mineral oil components. When the detected component index deviates from the normal range, an alarm is triggered. When the water content of an oil suddenly increases from the normal level of 0.3% to 15%, it means that the waste mineral oil has been illegally mixed with other water-containing substances, including waste emulsion. The module will immediately trigger a suspected illegal mixing warning, notify the person in charge of the waste generating unit and the environmental protection regulatory department to intervene and inspect in advance to avoid the risk of illegal discharge. Through the abnormal detection mechanism based on component changes, the problem of traditional manual inspections being unable to detect illegal mixing behavior is effectively solved. The component assessment and value grading module constructs an adaptive value assessment model and dynamically adjusts the weight of each quality indicator in the value assessment to achieve accurate prediction and dynamic grading of the value of waste mineral oil. The value assessment model adopts a multi-input neural network architecture with attention mechanism, which includes three parts: feature embedding layer, interaction learning layer and output prediction layer; Feature embedding layer: Converts the original values ​​of each quality indicator into a dense vector representation suitable for neural network processing, and learns the preliminary correlation between each indicator; Interactive learning layer: Employs a self-attention mechanism, which can adaptively learn complex interactive effects between different quality metrics; Output prediction layer: Summarizes the results of interactive learning and outputs the value level classification results and value amount regression prediction; The value assessment model is a component-value mapping model based on machine learning. The input of the model is various quality indicators of waste mineral oil, and the output is the value grade of waste mineral oil and the predicted value of recycling revenue. The parameters of the model, especially the weight coefficients of each indicator, are not preset by experts, but are automatically adjusted by learning historical transaction data and disposal feedback data. The various quality indicators of waste mineral oil constitute the input feature vector of the classification model, including: xi1: water content (%), xi2: impurity ratio (%), xi3: oil purity (%), xi4: acid value (mgKOH / g), xi5: flash point (°C), xi6: metal content (mg / kg). All input features need to be standardized before being fed into the neural network. The interactive learning layer employs a multi-head self-attention mechanism, which adaptively learns the complex interactive effects between different quality indicators. It was found that when the water content is high, the impact of the impurity ratio on the final value is significantly amplified, revealing a nonlinear relationship. The calculation process for the k-th attention head is as follows: First, calculate the Query, Key, and Value matrix: Q(k)=H(l)·WQ(k), K(k)=H(l)·WK(k), V(k)=H(l)·WV(k); Attention weights are calculated using a scaled dot product attention mechanism; The output of multi-head attention is obtained through concatenation and linear transformation: H(l+1)=Concat(a(1),a(2),...,a(H))•WO+bO; The interactive learning layers are stacked in L layers, usually 3 layers, and each layer learns the feature interaction relationships at different levels of abstraction through a self-attention mechanism; For value level classification tasks, with three categories—high, medium, and low—the output layer uses the softmax activation function. The output predicts the value level category c∈{H,M,L}, which represents the three levels: High, Medium, and Low, respectively. For the task of predicting recovery returns, the output layer uses a linear activation function: Output the predicted recovery amount r In addition, the model training adopts an incremental learning strategy. When a new recycling transaction is completed, the component data of this transaction is used as training samples, and the actual disposal revenue reported by the disposal factory is used as labels. These are added to the historical training dataset. Every fixed period, which is one week, the model will be incrementally fine-tuned on the new data to update the network parameters, so that the model can learn the latest market conditions and changes in disposal technology. Meanwhile, the results output by the value assessment model and the original rule-based assessment system form a dual-track parallel mechanism: The component detection data uploaded by the terminal is first quickly assessed by the rule evaluation system to give a preliminary value level conclusion; Subsequently, the data is fed into a neural network model for in-depth evaluation, which provides the value level of the model's predictions. When the two conclusions are consistent, the conclusion shall be adopted; when the two conclusions are inconsistent, the discrepancies shall be recorded for manual review, and the final evaluation shall be given by combining the results of the two. The carbon footprint accounting module is based on the carbon emission factors of each link in the entire recycling chain, including pickup, transportation and disposal. It combines operational data such as actual mileage and equipment energy consumption to automatically calculate the carbon emission reduction of each batch of waste mineral oil recycling and generate verifiable carbon reduction reports. It undertakes the quantitative accounting function of carbon emissions and carbon emission reduction of the entire waste mineral oil recycling chain. The carbon accounting method follows the ISO 14064 international standard framework, classifying carbon emission sources in the waste mineral oil recycling process into two main categories: direct emissions and indirect emissions. Direct emissions mainly include CO2 emissions from diesel / gasoline combustion during the operation of transport vehicles and grid emission factors corresponding to the electricity consumption generated by equipment operation during the recycling process. Indirect emissions mainly include greenhouse gas emissions, including methane, generated during the storage of waste mineral oil due to volatilization and leakage, expressed in CO2 equivalent. The carbon emission reduction accounting logic is based on a comparative analysis of waste mineral oil recycling and direct disposal. After formal recycling and regeneration, waste mineral oil can replace part of the production of virgin lubricating oil base oil, thereby reducing carbon emissions during crude oil extraction, transportation and refining. Based on the composition grade of waste mineral oil and the downstream treatment process, the corresponding emission reduction coefficients are matched from the carbon emission reduction factor database to calculate the amount of primary resources replaced by the recycling of this batch of waste mineral oil and the corresponding carbon emission reduction. The generated carbon footprint report can serve as the data basis for waste-generating entities to prepare ESG reports and apply for carbon incentive rewards; Taking the carbon footprint calculation of a certain batch of waste mineral oil as an example: the transportation distance is 22 kilometers. According to the carbon emission factor of the transportation vehicle, the carbon emission of the transportation link is 5.2 kg CO2. In the waste mineral oil recycling and disposal link, the carbon emission reduction is calculated to be 1.44 tons CO2 based on the substitution effect. The net carbon emission reduction contribution of this batch of waste mineral oil is 1.435 tons CO2. The carbon footprint report generated by the system contains complete carbon emission and carbon emission reduction details of this batch of waste mineral oil, and is accompanied by a third-party verification statement. It can be used as the data basis for waste-generating units to prepare ESG reports and apply for carbon inclusive rewards. The compliance management module achieves a 100% completion rate for joint forms, completely solving the problems of low completion rate and many errors and omissions in traditional manual joint form filling. It undertakes three major functions: automatic generation of electronic joint forms, automatic reporting to the regulatory system, and automatic maintenance of electronic ledgers. Regarding the automatic generation of electronic manifests, after receiving the signal that the recycling task is completed, the system automatically extracts information on the waste-generating unit, driver, vehicle, waste mineral oil, and disposal plant, and automatically generates an electronic manifest in accordance with the format requirements of the national hazardous waste transfer manifest. The waste-generating unit information includes name, address, and license number; the driver information includes name, ID number, and professional qualification certificate number; the vehicle information includes license plate number and transportation permit number; the waste mineral oil information includes quantity, composition, and value grade; and the disposal plant information includes name, address, and business license number. Regarding automatic reporting to the regulatory system, it connects with the provincial solid waste supervision and management information system through a data interface. After the electronic manifest is generated, the manifest data is automatically pushed to the regulatory platform to complete the online application and approval process for hazardous waste transfer. It supports connection with the regulatory systems of multiple provinces and can automatically select the corresponding regulatory platform for reporting based on the geographical location of the waste generating unit and the disposal plant. Regarding the automatic maintenance of electronic ledgers, an electronic hazardous waste management ledger is established for each waste-generating unit, recording ledger elements including the generation, transfer, disposal, and inventory of all waste mineral oils in the unit. The electronic ledger and electronic manifest are linked in real time. The ledger is automatically updated every time a recycling or transfer is completed, without the need for manual maintenance. Waste-generating units can view and export the complete electronic ledger at any time to meet the data requirements of environmental protection departments for on-site inspections and annual declarations. The mobile execution subsystem includes a driver-side APP module and a waste-generating mini-program module; The driver-side APP module is a mobile operation platform for recycling workers, including drivers. It integrates core functional units such as task reception, route navigation, QR code scanning, and handover confirmation, and is the key carrier for the implementation of dispatch instructions. The task receiving unit is responsible for receiving recycling tasks issued by the cloud platform. Task details include the location information of the oil drums to be recycled, the expected number of drums to be recycled, the recycling priority, and special requirements. The task list supports automatic sorting by priority and distance, and drivers can check their work schedule for the day. The route navigation unit is deeply integrated with mainstream map navigation services, including Gaode Map and Baidu Map. It accesses the cloud AI scheduling platform to calculate the optimal recycling route. The route planning algorithm comprehensively considers multiple factors, including the geographical location of each recycling point, traffic conditions, time window requirements, and vehicle loading capacity, and outputs the route plan with the shortest overall driving distance and the least total time. The driver arrives at each recycling point in sequence according to the navigation route. The system records the actual driving trajectory in real time, providing data support for subsequent carbon footprint accounting. The QR code scanning and handover confirmation unit is a key link in realizing closed-loop management of waste mineral oil recycling. After arriving at the recycling point, the driver uses the APP to scan the QR code on the oil drum. The system automatically identifies the waste-generating unit, oil drum number, and historical recycling records corresponding to the oil drum, and retrieves the latest component test data of the batch of waste mineral oil. After the driver completes the replacement of the empty drum, he uses the APP to take a photo of the drum replacement as a handover certificate. The photo is automatically stamped with a timestamp and GPS location watermark to ensure the authenticity and traceability of the handover process. After the handover confirmation is completed, the system automatically triggers the electronic manifest generation and push process. The waste generation end mini-program module is a mini-program application for waste mineral oil generating units, including auto repair shops and 4S shops, providing core functions such as liquid level monitoring, historical query, manifest management and carbon reduction contribution. The liquid level monitoring unit displays the current liquid level height and liquid level change trend of each oil drum in the store in real time. The person in charge of the waste generating unit can keep track of the oil drum status at any time without on-site inspection. When the liquid level is close to the overflow threshold, the module pushes a reminder notification to remind the waste generating unit to prepare for changing the drum. The historical query unit supports querying the store's historical recycling records by time range, including recycling date, recycling quantity, oil quality grade, transport vehicle information, and electronic manifest number; The manifest management unit displays the status of all electronic manifests for hazardous waste transfer in the store, including pending confirmation, confirmed, submitted, and completed. It supports viewing manifest details and downloading electronic archives. Waste-generating units can retrieve the complete information of any manifest at any time to meet the data needs of environmental inspections and internal audits. The carbon reduction contribution unit displays the store's cumulative waste mineral oil recycling volume and corresponding carbon emission reduction contribution in the form of visual charts. The data source is the calculation results of the carbon footprint accounting module. Waste-generating units can use carbon reduction data for external publicity and ESG report preparation.

[0023] Example 1: Taking a car maintenance shop as an example, one smart oil tank of the present invention is deployed; Initial liquid level 0%, component detection module output: water content 0.3%, impurities 0.1%, oil purity 99.6%, determined to be high-value waste mineral oil; After the system accumulates liquid level data for 7 days, the predictive model calculates that the waste mineral oil production cycle of the store reaches 90% liquid level every 5 days. On the morning of the 5th day, the cloud platform automatically dispatches an order to a nearby idle transport vehicle. The driver navigates to the location via the APP, scans the QR code on the oil drum, and the system automatically generates an electronic manifest. The electronic manifest includes the waste generating unit, driver, vehicle, and destination disposal plant. After the driver replaces the empty drum, the full drum information is automatically updated, and the manifest is pushed to the provincial solid waste system for filing. After the disposal plant receives the goods, the system automatically completes the settlement and generates a carbon footprint report for the batch of waste mineral oil: transportation distance 22 kilometers, carbon emissions 5.2 kg, and CO2 emission reduction of 1.44 tons in the disposal process. Store managers can view the cumulative recycling volume, manifest archives and carbon reduction contribution for the month at any time through the mini program. When the system detects a sudden increase in the water content of oil to 15%, it triggers an alert for suspected illegal mixing and notifies the environmental protection department to intervene and conduct an inspection in advance to prevent illegal discharge.

[0024] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling, characterized in that: include: The Internet of Things (IoT) smart recycling terminal is deployed at the waste mineral oil generation point. It is configured to enable an optical liquid level sensor and a near-infrared spectral detection sensor to collect waste mineral oil liquid level data and composition data through the same edge computing unit and mark them with a unified timestamp. The composition data includes water content and impurity ratio. The cloud-based AI scheduling platform is connected to the IoT smart recycling terminal and is configured to receive liquid level data and composition data with a unified timestamp. Based on the time series prediction algorithm, it predicts the overflow time according to the liquid level data change trend, and determines the waste mineral oil value level according to the water content and impurity ratio in the composition data, and generates scheduling instructions that include recycling priority and disposal channel recommendations. The mobile execution subsystem is connected to the cloud-based AI scheduling platform and is configured to receive the scheduling instructions and display the optimal recycling path, scan the QR code on the oil drum to complete the handover confirmation and trigger the generation of an electronic manifest. The edge computing unit is configured to perform Kalman filtering smoothing on the liquid level data, remove outliers from the component data, and trigger a local alarm when the liquid level exceeds a threshold or the component is abnormal.

2. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 1, characterized in that: The hardware architecture of the IoT smart recycling terminal consists of a sensor array unit, a data processing unit, a communication transmission unit, and a human-computer interaction unit. The sensor array unit integrates an optical liquid level sensor and a near-infrared spectroscopy detection sensor to collect data on the liquid level and composition of waste mineral oil. The data processing unit uses a low-power embedded processor to process the sensor data. The communication transmission unit has a built-in wireless communication module that uploads the collected data to the cloud server in real time and receives instructions. The human-machine interaction unit includes a QR code label and a status indicator light.

3. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 2, characterized in that: The IoT smart recycling terminal includes a multi-source sensor fusion module, an edge computing and data preprocessing module, and a wireless communication and secure transmission module. The multi-source sensor fusion module integrates an optical liquid level sensor and a near-infrared spectral detection sensor to achieve dual-dimensional perception of the quantity and quality of waste mineral oil. The optical liquid level sensor calculates the liquid level height by emitting an infrared beam and receiving the reflected signal, while the near-infrared spectral detection sensor analyzes multiple component indicators of waste mineral oil in real time and quickly determines the quantitative indicators of waste mineral oil. The edge computing and data preprocessing module uses a low-power processor, runs a streamlined embedded operating system, and has local data processing capabilities to achieve data filtering and outlier handling, data compression and breakpoint resume, as well as local alarm triggering. Data filtering and outlier handling: Kalman filtering algorithm is used to smooth the liquid level data, and box plot method is used to remove outliers from the component detection data; Data compression and breakpoint resume: When the network connection is unstable or the bandwidth is limited, the collected data is cached and compressed locally, and automatically resumed after the network is restored; Local alarm triggering: In case of emergency, it can respond quickly independently of the cloud, triggering local audible and visual alarms and executing preset emergency operations; The wireless communication and secure transmission module supports mainstream communication standards and automatically switches according to network coverage. It uses the MQTT protocol to connect to the cloud platform. In terms of data transmission security, it uses the TLS encryption protocol to achieve end-to-end encryption, adopts a two-way authentication mechanism based on digital certificates, and performs irreversible encryption processing on sensitive information.

4. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 1, characterized in that: The cloud-based AI scheduling platform includes four storage engines: time-series database, relational database, object storage, and graph database. The platform also includes a predictive scheduling module, a component assessment and value grading module, a carbon footprint accounting module, and a compliance management module. The predictive scheduling module uses historical liquid level change curves and time series prediction algorithms to estimate the specific time when each oil drum will reach the overflow threshold and generate recovery tasks in advance. The system employs two prediction models, ARIMA and LSTM, and automatically selects the optimal model or a combination of models based on the actual liquid level change characteristics of each waste-generating unit. It includes four key steps: historical data accumulation, trend analysis and prediction, recycling task generation, and scheduling instruction issuance.

5. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 4, characterized in that: The component assessment and value grading module transforms the near-infrared spectral data collected by the IoT terminal into business-meaning value indicators, while identifying illegal mixing behavior, triggering hazardous waste mixing warnings, and undertaking three major functions: waste mineral oil quality detection, value grading assessment, and abnormal component warnings. In terms of waste mineral oil quality testing, it receives raw near-infrared spectral data, performs spectral preprocessing, feature extraction and quantitative analysis, and outputs quantitative indicators such as water content, impurity ratio and oil purity. In terms of value grading, waste mineral oil is divided into three value grades: high, medium, and low. High-value waste mineral oil: water content <1%, impurities <2%, purity >98%, matched with high-grade treatment channels; medium-value waste mineral oil: water content 1%-5%, impurities 2%-10%, requires pretreatment before use; low-value waste mineral oil: water content >5%, impurities >10%, requires special treatment. Regarding the early warning of abnormal components, a normal fluctuation range model for waste mineral oil components is established, and an alarm is triggered when the detected component index deviates from the normal range.

6. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 5, characterized in that: The component assessment and value grading module dynamically adjusts the weight of each quality indicator in the value assessment by constructing an adaptive value assessment model. The value assessment model is a component-value mapping model based on machine learning. The input of the model is various quality indicators of waste mineral oil, and the output is the value grade of waste mineral oil or the predicted value of recycling revenue. The value assessment model employs a multi-input neural network architecture with an attention mechanism. It consists of three parts: a feature embedding layer, an interactive learning layer, and an output prediction layer. The feature embedding layer converts the original values ​​of each quality indicator into a dense vector representation suitable for neural network processing, while learning the preliminary correlation between each indicator. The interactive learning layer employs a self-attention mechanism to adaptively learn the complex interactive effects between different quality indicators; The output prediction layer summarizes the results of interactive learning and outputs the value level classification results or the value amount regression prediction.

7. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 6, characterized in that: The value assessment model is trained using an incremental learning strategy. When a new recycling transaction is completed, the component data of this transaction is used as a training sample, and the actual disposal revenue reported by the disposal factory is used as a label. These are then added to the historical training dataset. Every fixed period, the model will be incrementally fine-tuned on the newly added data to update the network parameters. The results output by the value assessment model and the original rule-based assessment system form a dual-track parallel mechanism: The component detection data uploaded by the terminal is first quickly assessed by the rule evaluation system to give a preliminary value level conclusion; Subsequently, the data is fed into a neural network model for in-depth evaluation, providing the model's predicted value level or profit forecast. When the two conclusions are consistent, the conclusion is adopted; when the two conclusions are inconsistent, the discrepancies are recorded for manual review, and the final evaluation is given by combining the results of the two.

8. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 4, characterized in that: The carbon footprint accounting module automatically calculates the carbon emission reduction for each batch of waste mineral oil recycling based on the carbon emission factors of each link in the entire recycling chain and combined with actual operation data, and generates a verifiable carbon reduction report. Carbon accounting methods follow international standard frameworks, classifying carbon emission sources into two main categories: direct emissions and indirect emissions; The carbon emission reduction accounting logic is based on a comparative analysis of waste mineral oil recycling and direct disposal, calculating the amount of primary resources replaced by the recycling of this batch of waste mineral oil and the corresponding carbon emission reduction. The generated carbon footprint report can serve as the data basis for waste-generating entities to prepare ESG reports and apply for carbon incentive rewards; The compliance management module achieves a 100% form completion rate and undertakes three major functions: automatic generation of electronic forms, automatic submission to the regulatory system, and automatic maintenance of electronic ledgers. Regarding the automatic generation of electronic manifests: information on waste-generating units, drivers, vehicles, waste mineral oil, and disposal plants is automatically extracted to generate electronic manifests; Regarding the automatic reporting of regulatory systems: By connecting with the provincial solid waste supervision and management information system through data interfaces, the online application and approval process for hazardous waste transfer is automatically completed; Regarding the automatic maintenance of electronic ledgers: an electronic hazardous waste management ledger is established for each waste-generating unit, and it is automatically updated every time a recycling or transfer is completed.

9. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 1, characterized in that: The mobile execution subsystem includes a driver-side APP module and a waste-generating mini-program module. The driver-side APP module is a mobile operation platform for recycling workers, integrating core functional units including task reception, route navigation, QR code scanning, and handover confirmation. The task receiving unit is responsible for receiving recycling tasks issued by the cloud platform. The path navigation unit is deeply integrated with mainstream map navigation services and accesses the optimal recycling route calculated by the cloud AI scheduling platform. The QR code scanning and handover confirmation unit scans the QR code of the oil drum to automatically identify relevant information and takes a photo of the replaced drum as a handover certificate after the empty drum is replaced.

10. The intelligent waste oil recycling system based on multi-source sensor fusion and AI scheduling according to claim 9, characterized in that: The waste-generating end mini-program module is a mini-program application for waste mineral oil generating units, providing the following core functions; The liquid level monitoring unit displays the current liquid level height and liquid level change trend of each oil drum in the store in real time, and pushes a reminder notification when the liquid level is close to the overflow threshold; The history query section supports querying the store's historical recycling records by time range; The manifest management unit displays the status of all electronic manifests for hazardous waste transfer in this store, and supports viewing manifest details and downloading electronic archives; The carbon reduction contribution unit displays the store's cumulative waste mineral oil recovery volume and corresponding carbon emission reduction contribution in the form of visual charts, supporting the preparation of ESG reports.

Images