A sensor-based online detection and analysis method and system for turbine lubricating oil

Abstract

The application discloses a kind of based on sensor steam turbine lubricating oil online detection analysis method and system, it is related to steam turbine lubricating oil online detection analysis technical field, to solve the problem that current steam turbine lubricating oil online detection analysis method cannot effectively distinguish normal transient fluctuation and real abnormal fluctuation under the condition that steam turbine variable working condition is operated, leading to false alarm frequently, the method comprises: the multiple oil parameters of steam turbine lubricating oil and the operating parameter of steam turbine are collected;According to operating parameter, the operating condition stage that steam turbine currently is in is identified;When operating condition stage represents the specific state that steam turbine is switched between stable operation and variable working condition operation, the real-time fluctuation of oil parameter is dynamically evaluated, to distinguish normal transient fluctuation caused by operating condition stage switching and real abnormal fluctuation indicating oil deterioration or equipment anomaly.

Description

Technical Field

[0001] This application relates to the field of online detection and analysis technology for turbine lubricating oil, and in particular to a sensor-based online detection and analysis method and system for turbine lubricating oil. Background Technology

[0002] In large-scale industrial production facilities, steam turbines serve as core power equipment, and their continuous and stable operation is crucial. Lubricating oil is the "blood" of a steam turbine's normal operation, undertaking multiple responsibilities such as lubrication, cooling, cleaning, and sealing. To ensure the reliability of steam turbines, sensor-based online lubricating oil monitoring and analysis methods are typically deployed to monitor oil conditions in real time and promptly identify potential problems. This method involves installing various sensors in the lubricating oil circulation pipeline, such as viscosity sensors, moisture sensors, particle counters, and temperature and pressure sensors, to continuously collect various physical and chemical parameters of the lubricating oil. These parameters are then analyzed by a data processing unit to provide early warnings at the initial stages of oil deterioration or equipment malfunction.

[0003] In large thermal power plants or combined cycle power plants, the steam turbine unit is the core power equipment, and its long-term stable operation is crucial. To ensure the safety of the unit, an online lubricating oil monitoring system is deployed in the main oil tank and oil supply pipeline of the steam turbine. This system uses multiple sensors installed in the oil circuit to collect key physicochemical indicators of the lubricating oil in real time, such as viscosity, water content, dielectric constant, and the number and size of metal wear particles. The collected information is transmitted to the monitoring computer in the central control room. The analysis software on the computer continuously evaluates the oil quality based on pre-set standard thresholds under normal operating conditions. When any indicator exceeds the preset warning or danger line, the system automatically issues an alarm, prompting maintenance personnel to pay attention or take measures, such as arranging oil filtration, oil replenishment, or oil replacement. This method can effectively detect the slow deterioration trend of lubricating oil caused by oxidation, contamination, and other factors in advance when the steam turbine is operating under stable, high-load conditions for a long time, providing a reliable basis for implementing condition-based maintenance and preventing unplanned shutdowns.

[0004] However, with the increasing penetration of intermittent energy sources such as wind and solar power in the power grid, the operating mode of thermal power units, which are the main force for peak shaving in the power grid, has undergone fundamental changes. Steam turbine units no longer operate stably under high loads for extended periods; instead, they need to frequently start and stop, and rapidly adjust loads to respond to grid dispatch instructions and balance the volatility of renewable energy generation. This frequent change in operating conditions presents new challenges to lubricating oil condition monitoring. For example, during a cold start-up of the unit, the lubricating oil temperature is low, and its viscosity will temporarily be significantly higher than the value at normal operating temperature. At this time, the high viscosity alarm set based on stable operating conditions may be frequently and falsely triggered. Similarly, after the unit is shut down, due to the temperature difference between the inside and outside of the equipment, a small amount of water vapor may condense on the inner wall of the oil tank and in the pipes. This moisture mixes into the oil during the next startup, causing the moisture sensor reading to rise sharply in a short period, thus triggering a false alarm for excessive moisture. These temporary and recoverable fluctuations in indicators caused by changes in operating status do not indicate irreversible deterioration of the lubricating oil itself. However, existing analytical methods cannot distinguish between these transient changes and actual oil deterioration, leading to maintenance personnel being troubled by a large number of invalid alarms, reducing their trust in the monitoring system, and ignoring real fault warnings. Summary of the Invention

[0005] This application provides a sensor-based online detection and analysis method and system for turbine lubricating oil, aiming to solve the problem that existing online detection and analysis methods for turbine lubricating oil cannot effectively distinguish between normal transient fluctuations and real abnormal fluctuations under variable operating conditions of the turbine, resulting in frequent false alarms and reduced system reliability.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a sensor-based online detection and analysis method for turbine lubricating oil, comprising: collecting multiple oil parameters of the turbine lubricating oil and turbine operating parameters; identifying the current operating condition stage of the turbine based on the operating parameters; and dynamically evaluating the real-time fluctuations of the oil parameters when the operating condition stage represents a specific state of the turbine switching between stable operation and variable operating conditions, in order to distinguish between normal transient fluctuations caused by the switching of operating condition stages and real abnormal fluctuations representing oil deterioration or equipment malfunction; wherein, the dynamic evaluation includes: matching the real-time fluctuations of the oil parameters with a preset expected fluctuation pattern under the current operating condition stage, and determining whether to trigger an alarm for the oil parameters based on the matching result.

[0007] Secondly, this application provides a sensor-based online detection and analysis system for turbine lubricating oil, comprising: a data acquisition unit for acquiring multiple oil parameters of the turbine lubricating oil and the turbine's operating parameters; an identification unit for identifying the current operating condition stage of the turbine based on the operating parameters; and an evaluation unit for dynamically evaluating the real-time fluctuations of the oil parameters when the operating condition stage represents a specific state in which the turbine switches between stable operation and variable operating conditions, in order to distinguish between normal transient fluctuations caused by the switching of operating condition stages and real abnormal fluctuations representing oil deterioration or equipment malfunction; wherein, the dynamic evaluation includes: matching the real-time fluctuations of the oil parameters with a preset expected fluctuation pattern under the current operating condition stage, and determining whether to trigger an alarm for the oil parameters based on the matching result.

[0008] This application offers at least the following advantages: By introducing "operating condition stage identification" and "dynamic evaluation" mechanisms, it can dynamically adjust the judgment criteria for oil parameter fluctuations based on the specific operating conditions of the turbine (e.g., startup, shutdown, load changes, etc.). By matching real-time fluctuations with the "expected fluctuation pattern" under the corresponding operating condition, it can effectively distinguish between normal transient fluctuations caused by operating condition switching and real abnormal fluctuations that truly indicate oil deterioration or equipment malfunction. For example, during cold startup, the system evaluates based on the preset viscosity change pattern under startup conditions, rather than simply comparing it with the stable operating condition threshold, thus avoiding false alarms triggered by normal viscosity increases. Therefore, the method of this application can significantly improve the accuracy and reliability of online detection and analysis of turbine lubricating oil, reduce false alarms, and allow maintenance personnel to focus more on handling real oil deterioration or equipment malfunctions, thereby improving the safety and economy of turbine operation. Attached Figure Description

[0009] Figure 1 This is a schematic flowchart of a sensor-based online detection and analysis method for turbine lubricating oil provided in this application. Detailed Implementation

[0010] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0011] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0012] Traditional online monitoring and analysis methods for turbine lubricating oil often struggle to accurately distinguish between normal transient fluctuations caused by changes in operating conditions and genuine oil deterioration or equipment malfunctions when turbine operating conditions change frequently. For example, during cold starts of the unit, the viscosity of the lubricating oil may temporarily increase, or during startup after shutdown, the moisture sensor reading may rise sharply due to water vapor condensation. These are not necessarily signs of oil deterioration, but they can trigger false alarms, causing maintenance personnel to be overwhelmed by numerous invalid alerts and potentially overlooking genuine fault warnings.

[0013] In view of the above problems, this application provides a sensor-based online detection and analysis method for turbine lubricating oil. By introducing an operating condition stage identification and dynamic evaluation mechanism, it can effectively distinguish between normal transient fluctuations and real abnormal fluctuations, thereby significantly improving the accuracy of alarms, reducing false alarms, and enhancing the reliability of online detection and analysis of turbine lubricating oil.

[0014] The sensor-based online detection and analysis method and system for turbine lubricating oil provided in this application will be described in detail and explained through the following specific embodiments.

[0015] Reference Figure 1 This application provides a sensor-based online detection and analysis method for turbine lubricating oil, which may include the following steps: S1. Collect multiple oil parameters of the turbine lubricating oil and the turbine's operating parameters.

[0016] Oil parameters refer to various physical and chemical indicators used to characterize the condition of turbine lubricating oil, such as viscosity, water content, temperature, pressure, particulate matter content, acid value, and dielectric constant. Changes in these parameters can reflect the degree of deterioration or contamination of the lubricating oil.

[0017] Operating parameters refer to various indicators that characterize the current operating status of a steam turbine, such as speed, load, bearing temperature, steam pressure, and steam temperature. Changes in these parameters directly reflect the operating conditions of the steam turbine.

[0018] Specifically, when collecting multiple oil parameters of the turbine lubricating oil and the turbine's operating parameters, various sensors can be deployed in the turbine's lubricating oil circulation pipeline to collect oil parameters. For example, viscosity sensors, moisture sensors, temperature sensors, and pressure sensors can be installed to obtain key indicators such as lubricating oil viscosity, moisture content, temperature, and pressure in real time. Simultaneously, the turbine's operating parameters can be obtained from the turbine control system via a data interface. For example, operating status information such as turbine speed, load, and bearing temperature can be acquired. The collection of these parameters forms the basis for subsequent analysis and evaluation.

[0019] S2. Identify the current operating condition stage of the steam turbine based on the operating parameters.

[0020] The operating condition stage refers to the specific state of a steam turbine under different operating modes, such as the stable operation stage, cold start stage, rapid load change stage, and shutdown stage. Different operating condition stages have a significant impact on the performance requirements and parameter fluctuation characteristics of the lubricating oil.

[0021] Various methods can be used to identify the current operating condition stage of a steam turbine based on its operating parameters. For example, the collected operating parameters can be analyzed according to preset rules. These rules can be defined based on thresholds, rates of change, or combinations of operating parameters. For instance, when the turbine speed and load remain within a set range for an extended period, it can be identified as a stable operating stage; when the speed rises rapidly from zero accompanied by an increase in load, it can be identified as a startup stage. By analyzing the operating parameters, the current operating condition stage of the steam turbine can be accurately determined, such as a stable operating stage, a cold start stage, or a stage with rapid load changes.

[0022] S3. When the operating condition stage represents a specific state in which the steam turbine switches between stable operation and variable operating condition, the real-time fluctuation of oil parameters is dynamically evaluated to distinguish between normal transient fluctuations caused by the switching of operating condition stages and real abnormal fluctuations that represent oil deterioration or equipment abnormality.

[0023] The dynamic assessment includes matching the real-time fluctuations of oil parameters with the preset expected fluctuation patterns under the current operating conditions, and determining whether to trigger an alarm for the oil parameters based on the matching results.

[0024] Expected fluctuation patterns refer to the typical patterns or trends in oil parameters under normal conditions during a specific operating phase. These patterns can be based on historical data statistics, physical model simulations, or expert experience.

[0025] Dynamic evaluation refers to adaptive analysis and judgment of real-time fluctuations in oil parameters based on the current operating conditions of the steam turbine, rather than simply comparing them with fixed thresholds.

[0026] When the operating condition phase characterizes the specific state of the steam turbine switching between stable operation and variable operating conditions, it is necessary to dynamically assess the real-time fluctuations of oil parameters. These specific states may include startup, shutdown, and rapid load increases or decreases. Under these states, the normal fluctuation range and pattern of oil parameters differ significantly from those during stable operation. The purpose of dynamic assessment is to distinguish between normal transient fluctuations caused by the switching of operating conditions and true abnormal fluctuations characterizing oil deterioration or equipment malfunctions. For example, during steam turbine startup, the lubricating oil temperature gradually increases, and the viscosity decreases accordingly—a normal physical change. If the viscosity sensor reading fluctuates at this time, it is necessary to determine whether this fluctuation conforms to the expected change pattern under startup conditions.

[0027] The specific implementation of dynamic assessment involves matching real-time fluctuations in oil parameters with preset expected fluctuation patterns for the current operating condition. These expected fluctuation patterns can be pre-stored in a transient feature library containing patterns corresponding to multiple operating conditions. For example, for the cold start phase, a typical curve showing the decrease in lubricating oil viscosity with increasing temperature can be stored as an expected fluctuation pattern. During real-time monitoring, the real-time change sequence of oil parameters is compared with the expected fluctuation patterns for the corresponding operating conditions in the transient feature library. Similarity calculations can employ various algorithms, such as Euclidean distance, correlation coefficient, or dynamic time warping (DTW). If the similarity result is higher than a first preset threshold, the current oil parameter fluctuation is determined to be a normal transient fluctuation, and no alarm is triggered. Conversely, if the similarity is lower than the threshold, it may indicate actual oil deterioration or equipment malfunction; in this case, the system will decide whether to trigger an alarm for the oil parameters based on the matching result.

[0028] The aforementioned method first collects multiple lubricating oil parameters and turbine operating parameters. These parameters form the basis for a comprehensive assessment of the lubricating oil condition and turbine operating status. Subsequently, based on the collected operating parameters, the system can accurately identify the current operating condition stage of the turbine. This step is one of the key innovations of this application, enabling subsequent lubricating oil parameter assessments to be adjusted according to the actual operating state of the turbine, rather than using a fixed threshold. When the turbine is identified as switching between stable and variable operating conditions, such as during startup, shutdown, or rapid load changes, the system dynamically assesses the real-time fluctuations of the lubricating oil parameters. This dynamic assessment process is the core technical contribution of this application. It no longer simply compares real-time lubricating oil parameters with fixed thresholds, but rather matches them with a preset expected fluctuation pattern under the current operating condition stage. For example, during cold startup, the lubricating oil viscosity gradually decreases due to temperature increases, while the moisture content may temporarily increase due to condensation. These changes are normal under specific operating conditions. By matching real-time fluctuations with these expected patterns, the system can intelligently distinguish between normal transient fluctuations caused by operating condition switching and real abnormal fluctuations that truly represent oil deterioration or equipment malfunction.

[0029] Therefore, the method of this application can effectively avoid false alarms caused by normal physical phenomena during turbine operation under varying conditions, such as temporary fluctuations in viscosity or moisture during startup. By accurately identifying operating conditions and performing contextualized dynamic assessments, the system can significantly reduce the false alarm rate and improve the accuracy and reliability of alarms. This not only reduces the burden on maintenance personnel, allowing them to focus more on handling actual faults, but also increases trust in the online monitoring system, ensuring the effectiveness of turbine lubricating oil monitoring and thus guaranteeing the safe and stable operation of the turbine.

[0030] In some embodiments, the above-mentioned collection of multiple oil parameters of turbine lubricating oil and turbine operating parameters can be further refined into the following steps: collecting oil parameters through multiple sensors deployed in the turbine lubricating oil circulation pipeline, the oil parameters including at least viscosity, moisture content, temperature and pressure; and obtaining operating parameters from the turbine control system through a data interface, the operating parameters including at least speed, load and bearing temperature.

[0031] The acquisition of oil parameters is achieved through the strategic deployment of multiple sensors in the turbine lubricating oil circulation pipeline. These sensors are configured to monitor the key physical and chemical properties of the lubricating oil in real time. Specifically, viscosity sensors measure the flow resistance of the lubricating oil, moisture content sensors detect the concentration of water in the oil, temperature sensors monitor the real-time temperature of the lubricating oil, and pressure sensors measure the pressure of the lubricating oil in the pipeline. These sensors provide continuous and accurate oil condition data, laying the foundation for subsequent oil degradation analysis.

[0032] Furthermore, the turbine's operating parameters are obtained from the turbine control system via a data interface. This data interface can be a standard industrial communication interface, such as Modbus, PROFIBUS, or OPC UA, used to exchange data between the control system and the monitoring and analysis system. Operating parameters include at least the turbine's speed, load, and the temperature of critical bearings. Speed ​​and load directly reflect the turbine's operating intensity and power output, while bearing temperature is an important indicator for assessing the operating status of mechanical components. By directly obtaining these parameters from the control system, it can be ensured that the collected operating parameters are highly synchronized and consistent with the actual operating status of the turbine.

[0033] This application's solution utilizes dedicated sensors directly deployed in the lubricating oil circulation pipeline to acquire key oil parameters such as viscosity, moisture content, temperature, and pressure in real time and accurately. Simultaneously, by establishing a data interface with the turbine control system, it can synchronously acquire operating parameters closely related to the turbine's operating status, such as speed, load, and bearing temperature. This data acquisition method ensures the comprehensiveness, real-time nature, and accuracy of oil and operating parameters, providing a reliable data foundation for subsequent identification of operating condition stages and dynamic evaluation of oil parameters. Therefore, it enables more precise analysis of the correlation between oil parameter fluctuations and turbine operating status, effectively distinguishing between normal transient fluctuations and genuine abnormal fluctuations.

[0034] In some embodiments, this application further proposes a step for identifying the current operating condition stage of a steam turbine based on operating parameters, including: analyzing the operating parameters according to preset rules; identifying the current operating condition stage of the steam turbine based on the analysis results of the operating parameters; the operating condition stage includes at least one or more of the following: stable operation stage, cold start stage, and rapid load change stage.

[0035] Specifically, the preset rules can be a series of logical judgment conditions, threshold ranges, state machine models, or machine learning models. Their purpose is to determine the current operating state based on the combination and changing trends of the turbine's operating parameters. For example, when the turbine's speed, load, and bearing temperature remain stable within a certain range for a long period, the turbine can be determined to be in a stable operating phase. When the turbine starts from a stopped state, with the speed and load gradually increasing, accompanied by temperature changes, it can be determined to be in the cold start phase. When the turbine's load changes significantly within a short period, it can be determined to be in the rapid load change phase.

[0036] The operating parameters can include, but are not limited to, turbine speed, load, bearing temperature, steam pressure, and steam temperature. These parameters comprehensively reflect the real-time operating status of the turbine. By monitoring and analyzing these operating parameters in real time, changes in the turbine's operating conditions can be accurately captured.

[0037] The identification of operating condition stages aims to divide the complex turbine operation process into several distinct phases, such as the stable operation phase, the cold start phase, and the rapid load change phase. The stable operation phase typically refers to the turbine's long-term stable operation under rated or specific loads; the cold start phase refers to the process of the turbine starting from a shutdown state, gradually increasing speed and temperature, and then assuming load; the rapid load change phase refers to the state where the turbine load increases or decreases significantly within a short period. This stage division helps in the subsequent accurate assessment of fluctuations in fuel parameters.

[0038] The proposed solution analyzes operating parameters according to preset rules, enabling the precise division of the complex operation of a steam turbine into different operating condition stages. This in-depth analysis of operating parameters allows the system to accurately identify the specific operating condition of the steam turbine, such as whether it is in stable operation, cold start, or rapid load change. This refined operating condition identification forms the basis for subsequent dynamic evaluation of oil parameter fluctuations, ensuring that fluctuations in oil parameters under different operating conditions can be correctly attributed, thus avoiding misjudgments.

[0039] In some embodiments, this application further proposes matching the real-time fluctuations of oil parameters with the preset expected fluctuation patterns under the current operating condition stage, specifically including: when the operating condition stage represents a specific state in which the turbine switches between stable operation and variable operating condition operation, the real-time change sequence of oil parameters is compared with the expected fluctuation patterns of the corresponding operating condition stage in the transient feature library for similarity calculation; the transient feature library contains expected fluctuation patterns corresponding to multiple operating condition stages; when the similarity calculation result is higher than a first preset threshold, the current oil parameter fluctuation is determined to be a normal transient fluctuation.

[0040] Specifically, the real-time change sequence of oil parameters refers to the set of values ​​continuously collected within a specific time window, showing the changes of oil parameters (such as viscosity, moisture content, temperature, and pressure) over time. This sequence reflects the dynamic trend and amplitude of oil parameters during operating condition switching. The transient feature library can be understood as a pre-established database storing typical, normal, and expected fluctuation patterns of oil parameters under different operating conditions of the turbine (especially the specific state of switching between stable and variable operating conditions). These expected fluctuation patterns can be obtained and verified through historical data analysis, expert experience, or simulation models. Similarity calculation refers to quantifying the degree of similarity between the real-time change sequence of oil parameters and the corresponding expected fluctuation patterns in the transient feature library using mathematical or statistical methods. Commonly used similarity calculation methods include, but are not limited to, correlation coefficient, Euclidean distance, and dynamic time warping (DTW). The first preset threshold is a pre-set value used as a standard to judge whether the fluctuation of oil parameters is a normal transient fluctuation. When the similarity calculation result between the real-time change sequence and the expected fluctuation pattern exceeds the threshold, it indicates that the two are highly similar, and thus it can be determined that the current fluctuation is a normal phenomenon caused by the switching of operating conditions.

[0041] This application's solution addresses the potential accuracy limitations of traditional matching methods by introducing a transient feature library and similarity calculation. Specifically, the transient feature library provides validated normal oil parameter fluctuation patterns under different operating conditions as a reference benchmark, eliminating reliance on vague empirical judgments for dynamic evaluation. When the turbine enters a specific operating condition switching state, by calculating the similarity between the real-time collected oil parameter change sequence and the corresponding expected fluctuation pattern in the library, the degree of agreement between the current fluctuation and the normal pattern can be objectively and quantitatively assessed. Furthermore, by setting a first preset threshold, a clear quantitative standard is provided for determining normal transient fluctuations, avoiding biases from subjective judgments. Therefore, this solution can more accurately identify and distinguish between normal transient fluctuations caused by operating condition switching and true abnormal fluctuations that genuinely indicate oil deterioration or equipment malfunction.

[0042] In some embodiments, this application further proposes a dynamic assessment of the real-time fluctuations of oil parameters when the operating condition stage characterizes a specific state in which the turbine switches between stable operation and variable operating conditions. Specifically, this includes: before turbine startup, refining the startup condition into different startup sub-stages based on the initial thermal state of the lubrication system; calculating the contextualized expected change path of oil parameters during the subsequent startup process based on the startup sub-stages and the physical properties of the lubricating oil; the contextualized expected change path is used to reflect the pattern of operating condition parameters changing with the context; and matching the real-time change sequence of oil parameters with the contextualized expected change path for dynamic assessment.

[0043] Specifically, before starting a steam turbine, the initial thermal state of the lubrication system needs to be assessed. This initial thermal state can be understood as the temperature distribution of the lubricating oil at startup, the temperature of equipment components, and the overall heat dissipation of the system. By assessing the initial thermal state, the originally general startup conditions can be further subdivided into multiple startup sub-stages. For example, based on the initial temperature, it can be distinguished as cold startup, warm startup, or hot startup. These startup sub-stage divisions aim to more accurately reflect the specific environmental conditions during startup. The physical properties of the lubricating oil refer to its inherent characteristics exhibited at different temperatures and pressures, such as viscosity-temperature profiles, density, specific heat capacity, and thermal conductivity. These physical properties are the intrinsic basis for predicting how oil parameters change over time during startup. Contextualized expected change paths are used to reflect the patterns of change in operating parameters with changing conditions. This refers to the predicted trajectory of oil parameters (such as viscosity, temperature, pressure, and moisture content) throughout the startup process, using physical models, empirical formulas, or historical data analysis, under specific startup sub-stage conditions and known lubricating oil physical properties. This path is dynamic, reflecting the changing patterns of operating parameters (such as speed and load) under different conditions, rather than a single static threshold or fixed pattern. For example, in the cold start phase, the lubricating oil viscosity gradually decreases as the temperature rises, and its rate of change and final stable value differ from those in the residual heat start phase. In practical applications, matching the real-time change sequence of oil parameters with the contextualized expected change path involves comparing the real-time oil parameter data stream collected by sensors with the pre-calculated contextualized expected change path for the current start phase. This matching can be achieved using various algorithms, such as time series similarity algorithms and Dynamic Time Warping (DTW) algorithms, to assess the degree of deviation between the real-time data and the expected path.

[0044] This application's solution refines the startup process into different sub-stages based on the initial thermal state of the lubrication system before turbine startup, thus providing a more precise characterization of the specific operating conditions during startup. Based on this, and considering the physical properties of the lubricating oil, a more accurate contextualized expected change path is calculated for each startup sub-stage. This path fully considers the differences in the thermal state of the lubrication system during the initial startup phase and the influence of the lubricating oil's own characteristics, making the expected oil parameter change patterns more closely resemble reality. Therefore, when the real-time change sequence of oil parameters is matched with this contextualized expected change path, it can more accurately distinguish between normal transient fluctuations caused by the switching of startup operation stages and true abnormal fluctuations that genuinely characterize oil degradation or equipment malfunctions. This meticulous evaluation mechanism effectively avoids the misjudgment problems caused by the inability of a single expected pattern to adapt to complex startup scenarios in traditional methods.

[0045] In some embodiments described above in this application, it is proposed to refine the startup conditions into different startup sub-stages based on the initial thermal state of the lubrication system before turbine startup, in order to calculate the contextualized expected change path of oil parameters. Specifically, refining the startup conditions into different startup sub-stages based on the initial thermal state of the lubrication system before turbine startup may include the following steps: monitoring the temperature distribution and temperature change rate of the lubricating oil using temperature sensors deployed at multiple key locations in the lubrication system; comprehensively judging the initial thermal state of the lubrication system in conjunction with the turbine shutdown duration; outputting corresponding startup sub-stage identifiers based on the initial thermal state, and using these identifiers to represent different startup sub-stages, wherein the startup sub-stages include at least fully cooled startup and startup with residual heat.

[0046] Specifically, key locations in the lubrication system may include, but are not limited to, the bottom of the oil tank, the oil pump inlet, the bearing oil supply line, and the return line. By deploying temperature sensors at these locations, real-time temperature data of the lubricating oil in different areas can be obtained. Monitoring the temperature distribution of the lubricating oil aims to understand the heat transfer and distribution of the lubricating oil throughout the system; monitoring the rate of temperature change is used to assess the dynamic process of system cooling or heating. For example, after the turbine is shut down, the temperature of the lubricating oil will gradually decrease, and the rate of decrease can reflect the system's heat dissipation efficiency.

[0047] Furthermore, the turbine shutdown duration is a key factor in determining the initial thermal state. A longer shutdown duration usually means the system has sufficient time to cool down, while a shorter shutdown duration may indicate that the system still retains a significant amount of residual heat. By comprehensively analyzing real-time temperature data (including distribution and rate of change) monitored by temperature sensors in conjunction with the shutdown duration, the initial thermal state of the lubrication system can be accurately determined. For example, if the shutdown duration is long and the temperature at all critical locations has dropped to near ambient temperature, it can be determined as a fully cooled start-up state; if the shutdown duration is short and some critical locations still have relatively high temperatures or the rate of temperature decrease is slowing down, it can be determined as a start-up state with residual heat.

[0048] Therefore, based on the comprehensive assessment of the initial thermal state, the system will output corresponding start-up sub-stage identifiers. These identifiers are used to clarify the specific start-up condition of the turbine, such as "fully cooled start-up" or "start-up with residual heat." The purpose of these start-up sub-stage identifiers is to provide accurate operating condition inputs for subsequent calculations of the contextualized expected change path of oil parameters, ensuring that the expected change path can truly reflect the dynamic behavior of oil parameters under different initial thermal states.

[0049] This application's solution, by meticulously monitoring the temperature distribution and rate of change of the lubricating oil and combining this with the turbine's downtime, enables a comprehensive and accurate assessment of the initial thermal state of the lubrication system. It is precisely this precise judgment of the initial thermal state that allows the startup process to be refined into different sub-stages, such as fully cooled startup and startup with residual heat. This detailed division allows for a thorough consideration of the impact of different initial thermal states on the dynamic behavior of oil parameters when subsequently calculating the contextualized expected change paths. For example, during a fully cooled startup, the viscosity and temperature of the lubricating oil undergo a relatively long transition from low temperature and high viscosity to high temperature and low viscosity; while during a startup with residual heat, the change process of these parameters may be more rapid and the starting temperature may be higher. In this way, this application ensures the accuracy of the contextualized expected change paths, thereby providing a more reliable benchmark for dynamically assessing real-time fluctuations in oil parameters.

[0050] In some embodiments, this application further proposes that after dynamically evaluating the real-time fluctuations of oil parameters, the above method also includes: periodically calibrating the sensor used to collect oil parameters to correct sensor drift, the calibration being based on a comparison of oil parameter data collected when the turbine is under a preset stable operating condition with historical benchmark values.

[0051] Specifically, "periodic calibration" refers to testing and adjusting the sensor at predetermined time intervals or under specific conditions to ensure its measurement accuracy meets requirements. This calibration can be set according to actual needs, such as every certain operating time, after each major overhaul, or triggered when the system detects an abnormal trend in sensor readings. Its purpose is to maintain the accuracy and reliability of sensor measurements. "Correcting sensor drift" refers to eliminating or reducing measurement errors caused by long-term use, environmental changes, or aging of internal components through the calibration process. Sensor drift manifests as a continuous, systematic deviation between its output value and the true value. Through correction, the sensor's output value can be made closer to the true value. In practical applications, "preset stable operating conditions" refer to the state in which the steam turbine operates stably within its design parameter range. For example, the steam turbine operates stably at rated speed and load for a long time, and the fluctuations of various operating parameters (such as speed, load, and bearing temperature) are minimal. Under this condition, the parameters of the lubricating oil are usually also in a relatively stable state, and their variation patterns are known and predictable, making it very suitable as a reference condition for sensor calibration. Furthermore, "historical benchmark values" can be understood as reference values ​​or ranges of oil parameters obtained through multiple measurements and statistical analysis when the sensors are in normal working condition and the turbine is in a preset stable operating condition. These benchmark values ​​are usually established at the initial stage of system operation or after professional calibration and stored in a database as a reference standard for subsequent sensor calibration.

[0052] The proposed solution effectively addresses the impact of sensor drift on the accuracy of oil parameter detection by introducing a periodic calibration mechanism. Specifically, when the turbine is under preset stable operating conditions, the fluctuations in various lubricating oil parameters are small and predictable, making the collected oil parameter data considered relatively reliable. By comparing these real-time collected oil parameter data with pre-stored historical benchmark values, the presence and extent of sensor drift can be accurately identified. Once drift is detected, the corresponding calibration offset can be calculated and applied to subsequent sensor readings, thereby correcting the sensor's measurement error. It is precisely this calibration method based on stable operating conditions and historical benchmark values ​​that enables the sensor to maintain its measurement accuracy over a long period, providing an accurate and reliable data foundation for subsequent dynamic evaluation of oil parameters.

[0053] In some embodiments, periodic calibration of the sensors used to collect oil parameters is proposed to correct sensor drift. Specifically, the periodic calibration of the sensors used to collect oil parameters includes the following steps: when the turbine is detected to have entered a preset stable operating condition and maintained for a first preset duration, a calibration process is triggered; after the calibration is triggered, data of oil parameters are collected within a second preset duration, and their average value is calculated as the current calibration reference value; the current calibration reference value is compared with the stored historical normal values ​​to calculate the parameter deviation; when the parameter deviation shows a continuous unidirectional change and the cumulative amount exceeds a preset deviation threshold, it is determined that the corresponding sensor has drifted, and a calibration offset is generated; the calibration offset is applied to the subsequent collected readings of the corresponding sensor.

[0054] Specifically, during turbine operation, the system continuously monitors the turbine's operating status. When the turbine enters a preset stable operating condition, such as stable operation under rated load, and this stable operating condition lasts for a period of time equal to or exceeding a first preset duration, the system automatically triggers the sensor calibration process. The first preset duration can be set according to actual application requirements, such as several hours or several days, to ensure the true stability of the turbine's operating status.

[0055] After the calibration process is triggered, the system continues to collect data on the target oil parameters within a second preset time period. This second preset time period is typically shorter than the first preset time period, for example, several minutes to tens of minutes, to obtain sufficient data samples for accurate averaging. Subsequently, the collected oil parameter data is averaged to obtain the current calibration baseline value. This current calibration baseline value reflects the average level of the oil parameters measured by the sensor under the current stable operating conditions.

[0056] Next, the calculated current calibration baseline value is compared with the historical normal values ​​pre-stored in the system. Historical normal values ​​can be baseline data measured by the sensor under standard stable operating conditions at the time of manufacture or after professional calibration, or average values ​​of oil parameters considered normal accumulated over long-term operation. By comparing these values, the parameter deviation between the current sensor reading and the historical normal values ​​can be calculated.

[0057] Furthermore, the system continuously tracks this parameter deviation. When the parameter deviation exhibits a sustained unidirectional trend, such as consistently high or consistently low, and its cumulative amount exceeds a preset deviation threshold, the system determines that the corresponding sensor is drifting. The preset deviation threshold can be set based on the sensor's accuracy requirements and the system's tolerance for oil parameter detection. Once sensor drift is determined, the system generates a calibration offset based on the degree and direction of the deviation.

[0058] Finally, the generated calibration offset will be applied to all subsequent readings acquired by the sensor. This means that before the sensor is physically calibrated, the system can correct the sensor readings in real time via software, thereby ensuring the accuracy of subsequent oil parameter data and avoiding misjudgments or omissions caused by sensor drift.

[0059] This application's solution ensures that the calibration process is conducted under the most stable operating conditions by setting clear calibration trigger conditions, namely, the turbine entering a preset stable operating condition and maintaining it for a first preset duration, thereby improving the accuracy and reliability of the calibration. By collecting data within a second preset duration and calculating the average value as the current calibration reference value, instantaneous fluctuations are effectively smoothed, making the reference value more representative. Comparing the current calibration reference value with historical normal values ​​allows for the quantification of sensor deviation. More importantly, by monitoring the continuous unidirectional change and accumulation of parameter deviation, sensor drift can be effectively distinguished from normal random fluctuations, avoiding misjudgments. Once drift is identified, a calibration offset is generated and applied, enabling real-time software correction of the sensor readings. This provides accurate oil parameter data continuously without interrupting turbine operation, effectively solving the drift problem that may occur during long-term sensor operation.

[0060] In some embodiments, this application further proposes that, during periodic calibration, the above method further includes: simultaneously acquiring local environmental parameters and oil flow parameters of the lubrication system during periodic calibration; calculating the contextualized expected values ​​of oil parameters at each sensor installation location based on the local environmental parameters and oil flow parameters, combined with the physical properties of the lubricating oil; separating the deviation into sensor drift components and local oil state fluctuation components by analyzing the deviation between the sensor readings and the contextualized expected values, and combining the local environmental parameters, oil flow parameters, and sensor internal diagnostic signals; and using the sensor drift components to calculate the calibration offset.

[0061] Specifically, during the periodic calibration of the sensors used to collect oil parameters, local environmental parameters and oil flow parameters of the lubrication system are simultaneously acquired. Local environmental parameters may include, but are not limited to, local temperature, local pressure, and local humidity near the sensor installation location; these parameters reflect the actual conditions of the microenvironment in which the sensor is located. Oil flow parameters may include, but are not limited to, oil velocity, flow rate, and turbulence intensity at the sensor installation location; these parameters reflect the dynamic characteristics of the oil as it passes through the sensor. The acquisition of these parameters can be achieved by deploying additional micro-sensors at key locations in the lubrication system or by using existing sensor networks for data fusion.

[0062] Furthermore, based on the collected local environmental parameters and oil flow parameters, combined with the physical properties of the lubricating oil, such as viscosity-temperature curves and density-temperature curves, the contextualized expected values ​​of oil parameters at each sensor installation location are calculated. Contextualized expected values ​​refer to the theoretical or model-predicted values ​​of oil parameters at that location under the current local environmental and oil flow conditions, such as viscosity, moisture content, temperature, and pressure. For example, if the local temperature is high, the expected viscosity of the lubricating oil will decrease; if the flow rate increases, the expected pressure will change. By establishing a physical model or a machine learning model trained based on historical data, the theoretical expected values ​​of oil parameters at the sensor installation location can be accurately calculated based on real-time local environmental parameters and oil flow parameters.

[0063] Based on this, by analyzing the deviation between sensor readings and contextualized expected values, and combining local environmental parameters, oil flow parameters, and internal sensor diagnostic signals, the deviation is separated into sensor-specific drift components and local oil condition fluctuation components. The deviation between sensor readings and contextualized expected values ​​reflects the difference between actual measured values ​​and theoretical expected values. To accurately identify sensor drift, this deviation needs to be further decomposed. Sensor-specific drift components refer to measurement errors caused by aging, contamination, or mechanical wear of internal electronic components, typically manifesting as a systematic shift in readings. Local oil condition fluctuation components refer to the actual fluctuations in oil parameters within a local area caused by changes in local environment or oil flow conditions, such as local hot spots, bubbles, or impurity accumulation, rather than sensor malfunctions. This separation can be achieved through signal processing techniques, statistical analysis methods, or rule engines based on expert knowledge. Simultaneously, internal sensor diagnostic signals, such as sensor self-test results and power supply voltage stability, are used as auxiliary judgment criteria to improve the accuracy of the separation.

[0064] The sensor's own drift component is used to calculate the calibration offset. Once the sensor's own drift component is accurately isolated, a precise calibration offset can be calculated based on this component and applied to the corresponding sensor readings acquired subsequently, thereby effectively correcting the sensor's drift.

[0065] This application's solution, by simultaneously considering the local environmental parameters and oil flow parameters of the lubrication system during periodic calibration, and combining them with the physical properties of the lubricating oil, can construct more accurate contextualized expected values ​​for oil parameters. It is precisely because of the introduction of these contextualized expected values ​​that the deviation between sensor readings and expected values ​​can more accurately reflect the actual situation. Based on this, by comprehensively analyzing this deviation along with local environmental parameters, oil flow parameters, and internal diagnostic signals from the sensor, this application can effectively decompose the total deviation into a sensor drift component and a component representing local fluctuations in the oil's condition. This decomposition mechanism ensures that the calibration offset only corrects for the sensor's own drift, and does not misjudge the actual fluctuations in oil parameters caused by changes in local operating conditions as sensor malfunctions, thus avoiding the misjudgment problems that may exist in traditional calibration methods.

[0066] This application also discloses a sensor-based online detection and analysis system for turbine lubricating oil. The system includes: a data acquisition unit for acquiring multiple oil parameters of the turbine lubricating oil and the turbine's operating parameters; an identification unit for identifying the current operating condition stage of the turbine based on the operating parameters; and an evaluation unit for dynamically evaluating the real-time fluctuations of the oil parameters when the operating condition stage represents a specific state where the turbine is switching between stable operation and variable operating conditions, in order to distinguish between normal transient fluctuations caused by the switching of operating condition stages and real abnormal fluctuations representing oil deterioration or equipment malfunction. The dynamic evaluation includes: matching the real-time fluctuations of the oil parameters with a preset expected fluctuation pattern under the current operating condition stage, and determining whether to trigger an alarm for the oil parameters based on the matching result.

[0067] The aforementioned system aims to address the challenge of traditional online detection and analysis methods for turbine lubricating oil, which struggle to accurately distinguish between normal transient fluctuations caused by operating condition changes and genuine oil degradation or equipment malfunctions when turbine operating conditions frequently change. For example, during cold starts, lubricating oil viscosity may temporarily increase, or during restarts after shutdown, moisture sensor readings may rise sharply due to water vapor condensation. These are not necessarily signs of oil degradation, but they can trigger false alarms, overwhelming maintenance personnel with numerous invalid alerts and potentially causing them to overlook genuine fault warnings. This application, by incorporating acquisition, identification, and evaluation units, achieves intelligent dynamic evaluation of oil parameters, thereby significantly improving alarm accuracy, reducing false alarms, and enhancing the reliability of online detection and analysis of turbine lubricating oil.

[0068] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A sensor-based online detection and analysis method for turbine lubricating oil, characterized in that, include: Collect multiple oil parameters of the turbine lubricating oil and the operating parameters of the turbine; Based on the operating parameters, identify the current operating condition stage of the steam turbine; When the operating condition stage represents a specific state in which the steam turbine switches between stable operation and variable operating condition, the real-time fluctuation of the oil parameters is dynamically evaluated to distinguish between normal transient fluctuations caused by the switching of operating condition stages and real abnormal fluctuations that represent oil deterioration or equipment malfunction. The dynamic assessment includes: matching the real-time fluctuations of the oil parameters with the preset expected fluctuation patterns under the current operating conditions, and determining whether to trigger an alarm for the oil parameters based on the matching results. The step of matching the real-time fluctuations of the oil parameters with the preset expected fluctuation pattern under the current operating conditions includes: When the operating condition stage represents a specific state in which the steam turbine switches between stable operation and variable operating condition operation, the real-time change sequence of the oil parameters is compared with the expected fluctuation pattern of the corresponding operating condition stage in the transient feature library for similarity calculation; the transient feature library contains expected fluctuation patterns corresponding to multiple operating condition stages. When the similarity calculation result is higher than the first preset threshold, the current oil parameter fluctuation is determined to be a normal transient fluctuation. When the operating condition stage represents a specific state in which the steam turbine switches between stable operation and variable operating conditions, the real-time fluctuation of the oil parameters is dynamically evaluated, including: Before the turbine is started, the starting conditions are broken down into different starting sub-stages based on the initial thermal state of the lubrication system. Based on the aforementioned start-up sub-stage and the physical properties of the lubricating oil, the contextualized expected change path of the oil parameters during the subsequent start-up process is calculated; the contextualized expected change path is used to reflect the pattern of operating parameters changing with the context. The real-time change sequence of the oil parameters is matched with the contextualized expected change path for use in the dynamic assessment. Before the turbine starts, the starting conditions are further subdivided into different starting sub-stages based on the initial thermal state of the lubrication system, including: Temperature sensors deployed at multiple key locations in the lubrication system are used to monitor the temperature distribution and rate of temperature change of the lubricating oil. Based on the turbine's downtime, the initial thermal state of the lubrication system is comprehensively assessed. Based on the initial thermal state, a corresponding start-up sub-stage identifier is output, and different start-up sub-stages are represented by the start-up sub-stage identifier. The start-up sub-stage includes at least fully cooled start-up and start-up with residual heat.

2. The method according to claim 1, characterized in that, The collection of multiple oil parameters of the turbine lubricating oil and the operating parameters of the turbine includes: Oil parameters are collected by multiple sensors deployed in the turbine lubricating oil circulation pipeline. These oil parameters include at least viscosity, moisture content, temperature, and pressure. Operating parameters are obtained from the turbine control system via a data interface, including at least speed, load, and bearing temperature.

3. The method according to claim 1, characterized in that, The step of identifying the current operating condition stage of the steam turbine based on the operating parameters includes: The operating parameters are analyzed according to preset rules; Based on the analysis results of the operating parameters, the current operating condition stage of the steam turbine is identified; the operating condition stage includes one or more of the following: stable operation stage, cold start stage, and rapid load change stage.

4. The method according to claim 1, characterized in that, After dynamically evaluating the real-time fluctuations of the oil parameters, the method further includes: The sensors used to collect the oil parameters are periodically calibrated to correct sensor drift. The calibration is based on a comparison of oil parameter data collected when the turbine is in a preset stable operating condition with historical benchmark values.

5. The method according to claim 4, characterized in that, The periodic calibration of the sensors used to collect the oil parameters includes: When the turbine is detected to have entered a preset stable operating condition and maintained for a first preset duration, the calibration process is triggered. After calibration is triggered, data of the oil parameters are collected within a second preset time period, and their average value is calculated as the current calibration reference value. The current calibration reference value is compared with the stored historical normal values ​​to calculate the parameter deviation; When the parameter deviation shows a continuous unidirectional change and the cumulative amount exceeds the preset deviation threshold, it is determined that the corresponding sensor has drifted, and a calibration offset is generated; The calibration offset is applied to the corresponding sensor readings acquired subsequently.

6. The method according to claim 4, characterized in that, The method further includes: During periodic calibration, local environmental parameters and oil flow parameters of the lubrication system are collected simultaneously. Based on the local environmental parameters and oil flow parameters, combined with the physical properties of the lubricating oil, the contextualized expected values ​​of the oil parameters at each sensor installation location are calculated. By analyzing the deviation between the sensor readings and the contextualized expected values, and combining the local environmental parameters, oil flow parameters, and sensor internal diagnostic signals, the deviation is separated into a sensor drift component and a local oil condition fluctuation component; the sensor drift component is used to calculate the calibration offset.

7. A sensor-based online detection and analysis system for turbine lubricating oil, characterized in that, The system applied to the sensor-based online detection and analysis method for turbine lubricating oil as described in claim 1 includes: The data acquisition unit is used to acquire multiple oil parameters of the turbine lubricating oil and the operating parameters of the turbine. The identification unit is used to identify the current operating condition stage of the steam turbine based on the operating parameters. The evaluation unit is used to dynamically evaluate the real-time fluctuations of the oil parameters when the operating condition stage represents a specific state in which the steam turbine switches between stable operation and variable operating condition, so as to distinguish between normal transient fluctuations caused by the switching of operating condition stages and real abnormal fluctuations that represent oil deterioration or equipment abnormality. The dynamic assessment includes matching the real-time fluctuations of the oil parameters with the preset expected fluctuation patterns under the current operating conditions, and determining whether to trigger an alarm for the oil parameters based on the matching results.

Images