Adaptive observation method and device, storage medium and electronic device

Abstract

The application provides an adaptive observation method and device, a storage medium and an electronic device, and relates to the field of meteorological observation. The observation device can obtain real-time working condition feature information from multiple sensors, and use the information to obtain a search result from a frozen soil observation knowledge base, so that the strategy generation has domain knowledge constraints rather than experience settings. Then, a locally configured task reasoning model is called to process reasoning information containing the feature information and the search result, and an observation strategy matching the current working condition is output. In this way, the generation process of the observation strategy is always based on the real-time perception of the frozen soil state itself, so that it can respond to phase change dynamics and support high-fidelity, high-resolution and high-credibility long-term observation data output.

Description

Technical Field

[0001] This application relates to the field of meteorological observation, and more specifically, to an adaptive observation method, apparatus, storage medium, and electronic device. Background Technology

[0002] High-altitude permafrost regions are sensitive indicators of global climate change, with their freeze-thaw processes directly impacting surface water cycles, carbon and nitrogen release, and infrastructure stability. Because these areas are often high-altitude continuous permafrost zones, polar permafrost zones, or plateau seasonal permafrost zones, transportation is extremely inconvenient, the climate is harsh, and manual inspections are costly and pose safety risks. Therefore, deploying high-altitude permafrost dynamic observation stations is often adopted to achieve long-term, continuous, unmanned field observations.

[0003] Current mainstream dynamic monitoring stations for high-altitude permafrost generally adopt a fixed-period data acquisition strategy. This means that regardless of whether the soil is in a frozen and stable state or whether the power supply is sufficient, sensor sampling is triggered at a uniform time interval, and the raw data is stored locally or uploaded as needed. Meanwhile, the operation and maintenance functions of peripheral devices such as photovoltaic panel cleaning, equipment compartment temperature control, and sound-based bird deterrence operate independently, each determining its start and stop based on a single threshold value from its own sensor. There is no state sharing or collaboration between them.

[0004] Therefore, current observation methods lack the basis for responding to the physical processes of permafrost itself, making it difficult to support the production of high-fidelity, high-resolution, and high-reliability long-term observation data required for scientific research. Summary of the Invention

[0005] To overcome at least one deficiency in the prior art, this application provides an adaptive observation method, apparatus, storage medium, and electronic device, which enables the generation process of the observation strategy to always be based on the real-time sensed state of the permafrost itself, thus responding to phase change dynamics and supporting the production of high-fidelity, high-resolution, and high-reliability long-term observation data.

[0006] Firstly, this application provides an adaptive observation method, the method comprising: Acquire real-time operating condition characteristic information from multiple sensors; The search results are obtained from the permafrost observation knowledge base using the real-time operating condition feature information. The locally configured task inference model is invoked to process the inference information, including the real-time operating condition feature information and the retrieval results, to obtain an observation strategy that matches the current operating condition.

[0007] Secondly, this application provides an adaptive observation device, the device comprising: The operating condition monitoring module is used to acquire real-time operating condition characteristic information from multiple sensors; The knowledge enhancement module is used to obtain retrieval results from the permafrost observation knowledge base using the real-time working condition feature information; The strategy execution module is used to call the locally configured task inference model to process the inference information, including the real-time operating condition feature information and the retrieval results, to obtain an observation strategy that matches the current operating condition.

[0008] Thirdly, this application provides a storage medium storing a computer program that, when executed by a processor, implements the adaptive observation method.

[0009] Fourthly, this application provides an electronic device, which includes a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the adaptive observation method.

[0010] Compared with the prior art, this application has the following beneficial effects: The adaptive observation method, apparatus, storage medium, and electronic equipment provided in this application enable the observation equipment to acquire real-time operating condition characteristic information from multiple sensors and use this information to retrieve search results from a permafrost observation knowledge base, allowing strategy generation to be constrained by domain knowledge rather than empirical settings. Then, a locally configured task inference model is invoked to process the inference information containing the characteristic information and search results, outputting an observation strategy that matches the current operating conditions. Thus, the generation process of the observation strategy is always based on the real-time perceived permafrost state itself, thereby responding to phase change dynamics and supporting the production of high-fidelity, high-resolution, and high-reliability long-term observation data. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 One of the flowcharts of the adaptive observation method provided in the embodiments of this application; Figure 2 A schematic diagram illustrating the principle of data quality control provided in this application embodiment; Figure 3 A second schematic flowchart illustrating the adaptive observation method provided in this application embodiment; Figure 4 The third schematic flowchart of the adaptive observation method provided in the embodiments of this application; Figure 5This is a schematic diagram of the adaptive observation device provided in the embodiments of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of this application (hereinafter referred to as "the embodiments") clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0014] 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 the 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.

[0015] It should be noted that similar labels 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.

[0016] In the description of this application, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0017] Based on the above statement, as introduced in the background section, current observation methods lack the basis for responding to the physical processes of permafrost itself, making it difficult to support the production of high-fidelity, high-resolution, and high-reliability long-term observation data required for scientific research.

[0018] Specifically, regardless of whether the soil is in a frozen and stable state, or whether the power supply is sufficient, the sensor is triggered to sample at a uniform time interval or index, and the raw data is stored locally or uploaded at an opportune time.

[0019] For example, in a fixed-period permafrost monitoring scheme, data is collected using a preset sampling period and uploaded periodically via local storage on a data logger or a remote communication link. Because the sampling strategy is decoupled from the actual state of the permafrost, rapid changes are easily missed during the critical period of freeze-thaw phase transition due to insufficient sampling frequency, while a large amount of redundant data is generated during the deep stable freezing period due to excessively high sampling frequency, resulting in wasted storage space and ineffective power consumption. The freeze-thaw phase transition refers to the physical state jump process of water in permafrost near 0℃, involving freezing or thawing.

[0020] For example, in adaptive sampling schemes based on a single threshold, the sampling frequency is adjusted according to the power consumption, communication status, or a single environmental parameter threshold. Because the identification of the freeze-thaw phase change window in permafrost, the coupling law of layered soil-water-thermal-salt systems, the prediction of power supply margin, and the assessment of equipment status are not considered in a joint decision-making process, the sampling strategy adjustment has only one dimension, making it difficult to achieve the optimal balance between observation quality and power consumption under complex field conditions.

[0021] Meanwhile, the operation and maintenance functions of peripheral devices such as photovoltaic panel cleaning, equipment compartment temperature control, and sound wave bird deterrence operate independently, and each makes start-stop judgments based on a single threshold of its own sensor, without sharing status or cooperating with each other.

[0022] For example, in an independent peripheral operation and maintenance scheme, functions such as photovoltaic panel snow removal and de-icing, sound wave bird deterrence, and equipment compartment heating are treated as independent peripheral control modules, each executing preset actions based on local sensor inputs. Because there is a lack of unified scheduling among these operation and maintenance functions prioritizing core observation tasks, under low-power constraints, non-core peripherals may compete with core observation tasks for energy, reducing the ability to ensure observation continuity.

[0023] It should be noted that the defects in the solutions in the prior art are the result of practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by the embodiments of this application in the following text should be regarded as contributions to this application in the process of invention and creation, and should not be understood as technical content known to those skilled in the art.

[0024] Based on the discovery of the above-mentioned technical problems, this embodiment provides an adaptive observation method. For example... Figure 1 As shown, the method includes: S1 acquires real-time operating condition characteristic information from multiple sensors.

[0025] S2 uses real-time operating condition feature information to obtain search results from the permafrost observation knowledge base.

[0026] S3 calls the locally configured task inference model to process the inference information, including real-time operating condition feature information and retrieval results, to obtain an observation strategy that matches the current operating condition.

[0027] This can be understood as follows: the observation equipment can acquire real-time operating condition feature information from multiple sensors, and use this information to obtain retrieval results from the permafrost observation knowledge base, so that the strategy generation has domain knowledge constraints rather than experience settings; then it calls the locally configured task reasoning model to process the reasoning information containing the feature information and retrieval results, and outputs an observation strategy that matches the current operating conditions.

[0028] Thus, the generation process of the observation strategy is always based on the real-time perceived state of the permafrost itself, and therefore can respond to phase change dynamics, supporting the production of high-fidelity, high-resolution, and high-reliability long-term observation data.

[0029] It should be noted that the observation device, as an electronic device, can be, but is not limited to, a mobile terminal, tablet computer, laptop computer, desktop computer, or embedded device.

[0030] Taking an embedded device as an example, it includes a low-power edge controller, which can be one or more of the following: ARM Cortex-M series microcontrollers, RISC-V architecture microcontrollers, or low-power ARM Cortex-A series application processors. It is also equipped with non-volatile storage media to store the parameters of the task inference model and the permafrost observation knowledge base.

[0031] Optionally, the observation device can also communicate with a server, which can be a single server or a group of servers. The server group can be centralized or distributed (e.g., the servers can be a distributed system). In some embodiments, the server can be local or remote relative to the user terminal. In some embodiments, the server can be implemented on a cloud platform; by way of example only, the cloud platform can include private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, inter-cloud, multi-cloud, etc., or any combination thereof. In some embodiments, the server can be implemented on an electronic device having one or more components.

[0032] To make the solution provided in this embodiment clearer, the following is combined with... Figure 1 Each step of the method is described in detail. However, it should be understood that the operations in the flowchart may not be implemented in sequence, and steps without logical contextual relationships may be reversed in order or implemented simultaneously. Furthermore, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0033] See also Figure 1The following is a detailed explanation of step S1 in the diagram: S1 acquires real-time operating condition characteristic information from multiple sensors.

[0034] The observation equipment can acquire real-time operational characteristic information from multiple sensors. This information originates from four types of sensor units configured in the sensing and acquisition layer: near-surface meteorological sensing unit, underground permafrost stratification sensing unit, equipment status monitoring unit, and signal conditioning and preprocessing unit. Details of each unit are as follows: The near-surface meteorological sensing unit can collect one or more parameters from the following: air temperature, air humidity, atmospheric pressure, wind speed, wind direction, precipitation, snow depth, and light intensity or solar radiation intensity. The underground permafrost stratification sensing unit deploys multiple measuring points along the depth direction in the active layer and deep permafrost. Each measuring point is equipped with one or more of the following: soil temperature sensor, soil moisture sensor, soil conductivity or salinity sensor, and soil heat flux sensor, to reflect the dynamics of water, heat, and salt in the permafrost. The equipment status monitoring unit collects one or more operating parameters from the following: state of charge (SOC) of the energy storage battery, battery voltage, charging and discharging current, output power of the photovoltaic array, output power of the wind turbine, internal temperature of the equipment compartment, online status of each functional module, and communication link quality indication. The raw signals output by the above three types of units are uniformly processed by the signal conditioning and preprocessing unit to complete filtering and noise reduction, jitter reduction, timestamp synchronization, physical range rationality check, and initial marking of outliers, ultimately forming structured, marked, and time-consistent real-time operating condition characteristic information.

[0035] Optionally, for data points that clearly exceed the physically reasonable range, the signal conditioning and preprocessing unit only generates anomaly markers without discarding or replacing the original values. The markers are transmitted along with the data to ensure information integrity and traceability.

[0036] Thus, the real-time operating condition information obtained by the observation equipment is not a single-dimensional reading, but a multi-source fusion result that covers meteorological conditions, permafrost physical conditions and the equipment's own operating conditions, and can reflect the true evolution of permafrost.

[0037] Based on the above explanation of real-time operating condition characteristic information, the following will continue to discuss... Figure 1 Step S2 will be explained in detail: S2 uses real-time operating condition feature information to obtain search results from the permafrost observation knowledge base.

[0038] Regarding step S2, it should be noted that related technologies typically employ fixed-period sampling or rely on single-parameter threshold adjustments for observation behavior, lacking structured knowledge support for permafrost physical processes. Their knowledge organization is relatively loose, failing to partition and correlate freeze-thaw mechanisms, historical operational experience, sampling strategies, and equipment operation and maintenance logic. Therefore, when facing complex and ever-changing field conditions, the equipment cannot quickly match phase transition stage determination rules from existing knowledge, struggles to find historical conditions most closely matching the current weather, permafrost, and equipment status, and cannot retrieve appropriate sampling or operation and maintenance strategy templates as needed.

[0039] To address the above issues, the permafrost observation knowledge base includes partitions for permafrost freeze-thaw mechanism rules, historical operating conditions, sampling strategy rules, and equipment operation and maintenance strategies. The search results include phase transition stage determination rules, approximate operating condition entries, target sampling strategy templates, and target operation and maintenance strategy templates. The following optional implementation methods are provided for step S2: S2-1 uses real-time operating condition feature information to match phase change stage determination rules from the frozen soil freeze-thaw mechanism rule partition.

[0040] In practical applications, the observation equipment can build and maintain a structured permafrost observation knowledge base, which is clearly divided into four functional areas: permafrost freeze-thaw mechanism rule area, historical working condition area, sampling strategy rule area, and equipment operation and maintenance strategy area.

[0041] The observation equipment can use real-time operating condition characteristic information to match the phase change stage determination rules from the frozen soil freeze-thaw mechanism rule partition. This partition stores the freeze-thaw phase change temperature window, water content change rate threshold, and heat flux change rate threshold corresponding to different frozen soil types and different depth layers.

[0042] For example, for permafrost with high salinity, the phase transition temperature window can be corrected to -0.5℃ to +0.5℃. The criteria for judgment include whether the temperature of multiple soil layers is within this window, whether the residence time near zero degrees reaches a cumulative threshold, whether the rate of change in moisture content exceeds the corresponding limit, and whether the rate of change in heat flux changes from negative to positive or increases significantly. Based on this, the observation equipment classifies the current state into four categories: freezing period, thawing period, critical period of freeze-thaw phase transition, or stable freezing period. The determination of the critical period of freeze-thaw phase transition requires that the soil temperature of at least two adjacent depth layers is within the phase transition temperature window, and that the rate of change in moisture content or the rate of change in heat flux exceeds a preset threshold. The stable freezing period requires that the temperature of multiple soil layers remain below the lower limit of the window for a long period and that the rate of change approaches zero.

[0043] S2-2, based on the real-time operating condition feature information, obtain the approximate operating condition entry from the historical operating condition partition that has the closest feature distance to the real-time operating condition feature information.

[0044] The observation equipment can retrieve the approximate operating condition entry with the closest feature distance to the real-time operating condition feature information from the historical operating condition partition. This partition stores the typical operating condition feature vectors recorded in historical observations, along with their corresponding effective sampling and control strategies.

[0045] For example, before a severe snowstorm, there was a combination of changes: air pressure continuously dropping above the first threshold, humidity continuously rising above the second threshold, wind speed increasing and exceeding the third threshold, and temperature decreasing. Another example is that before an extreme cold snap, there were records of a sudden drop in temperature within a preset time window, extremely low rate of change in light intensity, and wind direction consistent with the typical path of a local cold wave. Yet another example is that before a period of persistent icing, there were observations of ambient temperature near freezing, humidity consistently above the threshold, and a mismatch between the decrease in photovoltaic array output power and the decrease in light intensity. These historical entries are all stored in a structured manner, containing complete multi-dimensional feature vectors and the effective responses that were validated at the time.

[0046] The observation equipment calculates the vector of the current real-time operating condition feature information and the feature distance of each historical entry, selects the one with the smallest distance as the approximate operating condition entry, and then retrieves the associated historical observation strategies and performance evaluations.

[0047] S2-3 uses real-time operating condition feature information to match the target sampling strategy template from the sampling strategy rule partition.

[0048] The observation equipment can use real-time operating condition characteristic information to match the target sampling strategy template from the sampling strategy rule partition. This partition stores sampling frequency configuration parameters under different freeze-thaw stages, different power supply margin levels, and different extreme weather risk levels.

[0049] For example, when the system is identified as being in a critical period of freeze-thaw phase change and the battery SOC is higher than the first protection threshold, a first-level encrypted sampling mode is used, increasing the sampling frequency of core parameters (soil temperature, water content, salinity, and heat flux) in the active layer and adjacent deep layers to once every 5 minutes. When the system is identified as being in the same stage but the SOC is between the first and second protection thresholds, a second-level encrypted sampling mode is used, increasing only the frequency of core parameters in the core layers while reducing the sampling frequency of non-core layers and non-core parameters. When the system is identified as being in a stable freezing period, a sparse sampling mode is used, uniformly reducing all parameters in all layers to once every 2 hours. When the extreme weather risk level is determined to be high, a conservative sampling mode is used, maintaining only the minimum continuous sampling of core parameters in the core layers (e.g., once every 30 minutes) while suspending the collection of non-core parameters. These templates are all pre-stored in the form of structured entries, each entry clearly associated with a specific combination of operating conditions and executable sampling actions, without relying on manual intervention or remote commands.

[0050] The observation equipment compares the current real-time operating condition characteristics with the triggering conditions of each template, directly hits and calls the most suitable target sampling strategy template, and provides it for reference in the subsequent reasoning process.

[0051] S2-4 uses real-time operating condition feature information to match the target operation and maintenance strategy template from the equipment operation and maintenance strategy partition.

[0052] The monitoring equipment can use real-time operating condition characteristic information to match the target operation and maintenance strategy template from the equipment operation and maintenance strategy partition. This partition stores photovoltaic panel cleaning trigger conditions, sound wave bird deterrence trigger conditions, equipment compartment heating strategies, battery low temperature protection strategies, and fault self-healing strategies.

[0053] For example, the triggering conditions for the photovoltaic snow removal unit include snow depth exceeding a threshold, photovoltaic array output power drop exceeding a threshold and duration meeting a standard, and ambient temperature change rate below a threshold while humidity change rate exceeds a threshold; the triggering conditions for the acoustic bird deterrent unit include an abnormal downward trend in photovoltaic array output power under the premise of no snowfall and stable light intensity, and this trend matches the bird activity time rules recorded in the knowledge base, while excluding factors such as snow accumulation, ice accumulation, and device aging; the equipment compartment heating strategy dynamically adjusts the heating power and air duct circulation status based on the current internal temperature of the equipment compartment, the operating temperature range of the energy storage battery, and the power supply margin; the battery low temperature protection strategy automatically limits high current discharge and starts preheating when the air temperature is below a threshold and the battery SOC is at a medium to low level; the graded fault self-healing strategy executes parameter rollback, module reset, or enters low power protection mode according to the anomaly type, duration, and impact range.

[0054] The monitoring equipment compares real-time operating condition characteristics with the triggering logic of each template item by item, accurately hits and calls the most suitable target operation and maintenance strategy template, and uses it as reference information for subsequent reasoning.

[0055] Based on the above examples illustrating the search results, we will continue with the following... Figure 1 Step S3 will be explained in detail: S3 calls the locally configured task inference model to process the inference information, including real-time operating condition feature information and retrieval results, to obtain an observation strategy that matches the current operating condition.

[0056] As an optional implementation, the inference information also includes the hardware constraint parameters and power supply status parameters of the observation equipment. The observation equipment calls the locally configured task inference model to process the real-time operating condition feature information, retrieval results, hardware constraint parameters and power supply status parameters to obtain the observation strategy.

[0057] The hardware constraints include the processing power, storage capacity, and peripheral interface limitations of the ARM Cortex-M series microcontroller, RISC-V architecture microcontroller, or low-power ARM Cortex-A series application processor used in the low-power edge controller; the power supply status parameters include one or more of the following: state of charge (SOC) of the energy storage battery, battery voltage, charging and discharging current, output power of the photovoltaic array, output power of the wind turbine, internal temperature of the equipment compartment, online status of each functional module, and communication link quality indication.

[0058] When the observation equipment calls the locally configured task inference model, it takes the above parameters, real-time operating condition feature information, and retrieval results as input to obtain an observation strategy that matches the current operating condition.

[0059] The task inference model is a constrained task inference model that has undergone quantization compression, structural pruning, and knowledge distillation, enabling it to be deployed on low-power edge controllers. Its inputs are explicitly limited to a vector of structured condition feature information, the retrieval results returned by the edge knowledge retrieval engine, hardware constraint parameters, and power supply status parameters. The outputs are recommended instructions selected from a predefined set of control instructions and confidence levels.

[0060] It should be noted that, unlike traditional large language models, this task inference model does not generate natural language control text. All outputs are mapped to existing instructions in a predefined control instruction set. The instruction set includes six major categories of instructions: sampling control, power consumption scheduling, temperature control and protection, peripheral device operation and maintenance, data quality control, and fault self-healing. Each category has several specific instruction numbers and corresponding parameter ranges.

[0061] Among these, sampling control includes instructions to adjust the sampling frequency of near-surface meteorological parameters, underground permafrost stratification parameters, and equipment status parameters. For example, increasing the sampling frequency of core parameters in core strata during the critical period of freeze-thaw phase change, and decreasing the sampling frequency of non-core parameters during the stable freezing period.

[0062] Power scheduling includes instructions for managing the flow of power between solar photovoltaic arrays, wind turbines, and cryogenic energy storage battery packs. For example, maximum power point tracking (MPPT) technology is used to optimize the power extraction efficiency of photovoltaic and wind power, and the power supply priority and duration of each functional module are adjusted according to power scheduling instructions issued by the edge intelligence core layer.

[0063] Temperature control and protection measures include temperature control and protection instructions issued by the edge intelligence core layer, which are used to adjust the heating power of the equipment compartment and the air circulation status inside the equipment compartment, so that the energy storage battery, edge controller, data acquisition module and communication module operate within the allowable temperature range.

[0064] Peripheral maintenance commands include photovoltaic panel cleaning commands and bird deterrence commands. The photovoltaic panel cleaning command activates the cleaning mechanism on the photovoltaic panel surface to remove snow or ice. For example, triggering conditions for the cleaning mechanism include snow depth exceeding a threshold, a decrease in photovoltaic array output power exceeding a threshold for both magnitude and duration, and an ambient temperature change rate below a threshold while a humidity change rate exceeds a threshold. The bird deterrence command uses low-frequency sound waves to drive away birds perched on the photovoltaic panel surface. For example, triggering conditions for the bird deterrence command include an abnormal decrease in photovoltaic array output power under conditions of no snowfall and stable light intensity, a decrease matching the bird activity time patterns recorded in the knowledge base, and excluding factors such as snow accumulation, ice buildup, and device aging.

[0065] Data quality control includes instructions for independently classifying and storing raw observation data, post-quality control observation data, and data processing logs. In this embodiment, the raw observation data storage partition is set to read-only or append-only mode, disallowing subsequent quality control processes from overwriting or deleting it; the post-quality control observation data storage partition stores data after anomaly marking, missing value completion, or noise filtering; the data processing log storage partition stores the raw data value, processing result value, processing rule number, processing timestamp, and operator identifier for each quality control processing operation.

[0066] like Figure 2 As shown, the observation equipment receives the raw observation data and performs the following data quality control operations: (1) Handling values ​​outside the physical range: For data values ​​that exceed the physical measurement range of the sensor or the reasonable physical range of the permafrost environment, an abnormal label is generated and, in principle, it is not automatically replaced with other values. Only after communication conditions allow and remote manual confirmation can a manually corrected value be used to replace the data and recorded in the log. If necessary, a separate quality control copy of the data can be generated for subsequent analysis, but the original data is still stored independently. (2) Short-term missing value completion: For missing data caused by short-term sensor failure or communication packet loss, the missing data is completed by linear interpolation, adjacent value averaging or knowledge base matching based on adjacent time series data, adjacent depth data of the same layer or historical similar working conditions data, and the completion label and confidence score are written into the data after quality control. (3) Sensor drift handling: For sensor output values ​​that gradually deviate from the historical baseline or the logical relationship between adjacent layers, drift judgment is made based on historical baseline data, adjacent layer data and redundant sensor data, drift tags are generated, and sensor recalibration process is initiated or maintenance alarm is generated when necessary. (4) Noise filtering of sudden changes: For single-point sudden change data caused by electromagnetic interference or instantaneous anomalies, under the premise of meeting the physical constraints of permafrost evolution (e.g., the temperature change rate does not exceed the physical upper limit), sliding window smoothing or median filtering is used for processing, and noise labels and numerical comparison records before and after processing are generated. (5) Long-term anomaly handling: For abnormal data whose duration exceeds the threshold, generate a fault alarm and do not automatically complete it into high confidence data to avoid introducing systematic errors.

[0067] Following the quality control process described above, the observation equipment stores the data processing results in three independent storage partitions: the original observation data storage partition, the post-quality control observation data storage partition, and the data processing log storage partition. Details of each partition are as follows: The raw observation data storage partition stores the unmodified observation data output by the sensing and acquisition layer. This partition is set to read-only or append-only mode, and upper-layer applications are not allowed to overwrite or delete the data that has been written. Each raw data record includes sensor ID, raw value, timestamp, signal quality indicator and raw data check code. The post-quality control observation data storage partition stores the observation data after processing such as anomaly marking, missing value completion, noise filtering, or drift correction. Each record in this partition includes the original data reference (linked to the corresponding record in the original data storage partition), the processed value, the processing type label, the processing rule number, the confidence score, and the quality control completion timestamp. The data processing log storage partition stores detailed logs for each quality control processing operation. Each log record in this partition includes the original data value, the processed data value, the exception type, the triggering condition, the processing rule number, the correction range, the confidence score, the estimation label (if applicable), and the processing timestamp.

[0068] Fault self-healing includes fault level classification based on fault type, duration, and impact range, and corresponding self-healing instructions. For example, a level 1 fault corresponds to parameter anomalies or short-term communication anomalies, and the instruction executes parameter rollback to the previous valid value or resampling; a level 2 fault corresponds to single-function module failure, and the instruction executes module software reset or power restart; a level 3 fault corresponds to power supply anomalies, temperature control anomalies, or multi-module collaborative failure, and the instruction executes entry into low-power protection mode or safe sleep mode, and waits for external interrupt to wake it up.

[0069] Because the inference observation strategy takes into account real-time operating conditions, retrieval results, hardware constraint parameters, and power supply status parameters, when the battery SOC is low, the model automatically tends to choose a low-power sampling frequency and suspend the operation of non-core peripherals; when the communication link quality is poor, the model avoids generating instructions that require remote confirmation; when the internal temperature of the equipment compartment is close to the lower limit of the controller's operation, the model triggers temperature control protection instructions to ensure system stability.

[0070] For example, when the input includes the judgment result of "critical period of freeze-thaw phase change", "battery SOC is higher than the first protection threshold" and "the controller has sufficient memory", the recommended instruction number output by the model corresponds to the instruction of level 1 encrypted sampling, with parameters of {sampling frequency = 5 minutes, core parameters = temperature / moisture content / heat flux} and confidence level of 0.92.

[0071] When the input includes "PV power drops beyond the threshold and lasts for 30 minutes", "snow depth > 15cm", and "controller computing power load > 85%", the model will output the instruction corresponding to the instruction number as the start of PV snow sweeping and the corresponding execution duration parameters.

[0072] It should also be noted that in related technologies, when the confidence level of the control command output by the task inference model is lower than a preset threshold, the command is usually discarded directly without any alternative response. Due to the lack of a tiered response mechanism for the consequences of command failure, the equipment cannot determine whether the command is associated with critical observation windows such as the freeze-thaw phase transition.

[0073] To address this issue, the observation strategy includes recommended instructions selected from a pre-defined set of control instructions, along with confidence levels, such as... Figure 3 As shown, the method provided in this embodiment further includes: S4. If the confidence level is less than the reliability threshold, determine whether there is a fallback instruction in the permafrost observation knowledge base that matches the real-time operating condition characteristics and has the highest priority. If yes, proceed to step S5; otherwise, proceed to step S6.

[0074] S5 executes the fallback instruction.

[0075] S6 uses historical control instructions that have been successfully executed in the previous cycle.

[0076] This can be understood as follows: the observation equipment can generate an observation strategy that includes recommended instructions and confidence levels. When the confidence level is less than the reliability threshold, it first checks whether there is a fallback instruction in the permafrost observation knowledge base that matches the real-time operating condition characteristics and has the highest priority. If it exists, the fallback instruction is executed; if it does not exist, the historical control instruction that was successfully executed in the previous cycle is used. The fallback instruction comes from the rule entries in the knowledge base.

[0077] For example, when the confidence level of the recommended instruction for the condition of "snowstorm precursor" is lower than the reliability threshold, the highest priority rule associated with the combination of "continuous decrease in air pressure + increase in humidity + increase in wind speed" pre-stored in the knowledge base is triggered to activate the equipment compartment preheating and photovoltaic snow removal mechanism standby; similarly, when the confidence level of the recommended instruction for the condition of "extreme cold weather precursor" is lower than the reliability threshold, the battery low temperature protection strategy associated with "sudden drop in temperature + low sunshine + typical cold wave wind direction" in the knowledge base is invoked.

[0078] The so-called historical control instructions that have been successfully executed specifically refer to instructions that did not trigger alarms or cause failures after execution in the previous cycle, and whose data quality control confidence level met the standards. For example, the sparse sampling instruction that was successfully executed once every 2 hours during the stable freeze period in the previous cycle.

[0079] In this way, even when the model output is unreliable, the observation equipment can still maintain basic observation functions by relying on knowledge base rules or verified historical actions.

[0080] It should also be noted that if the historical operating condition partition lacks dynamic update capability, the knowledge base will remain stagnant in the initial configuration state for a long time. When encountering new operating conditions that are similar to those in the past but have not been included, the device cannot match an effective strategy and can only rely on general rules or model inference, resulting in strategy adaptation deviation.

[0081] To address this issue, the permafrost observation knowledge base includes historical working condition zoning and real-time working condition characteristic information obtained from the quality control processing of raw observation data. For example... Figure 4 As shown, this adaptive observation method also includes: S7. If the observation strategy is executed successfully and no alarm is triggered, obtain the maximum similarity of real-time operating condition feature information in the historical operating condition partition.

[0082] S8. If the maximum similarity is less than the deduplication threshold, then generate knowledge entries including real-time working condition feature information, observation strategy, and execution results of the observation strategy and update them to the permafrost observation knowledge base.

[0083] In this embodiment, the observation equipment can dynamically update the historical working condition partitions in the permafrost observation knowledge base, thereby enabling the autonomous accumulation of knowledge related to the observation strategy.

[0084] The real-time operating condition feature information is obtained from the raw observation data through quality control processing, including fused features such as multi-layer soil temperature, moisture content, salinity, heat flux, meteorological parameters, and equipment status. If the observation strategy is executed successfully and no alarm is triggered, the maximum similarity of the real-time operating condition feature information in the historical operating condition partition is obtained; if the similarity is less than the deduplication threshold, it indicates that the current operating condition is significantly different from the existing entries and has new value, so a structured knowledge entry is generated and updated to the historical operating condition partition.

[0085] This entry explicitly includes real-time operating condition characteristics, the observation strategy adopted, the evaluation of the strategy's execution effect (success / failure / partial success), and the entry entry timestamp.

[0086] For example, if a photovoltaic snow removal command is executed under the precursor of continuous icing and power generation is successfully restored, and the feature distance between this condition and all historical entries in the knowledge base is greater than a set threshold, then this operation and maintenance experience is added as a new knowledge entry. Similarly, if the equipment compartment heating and battery discharge strategies are adjusted in conjunction during extreme cold weather to ensure that the core sensors operate continuously for more than 72 hours without triggering alarms, this combined strategy is considered a new knowledge entry.

[0087] In addition, when the communication link is available, the observation equipment can receive remotely sent knowledge base update data packets (including new rules, update thresholds, and new historical working condition cases) and update the local permafrost observation knowledge base incrementally; when the communication link is unavailable, high-confidence new observation data that has passed quality control can be temporarily stored locally and uploaded to the remote server in batches after the communication is restored, and a joint update of the knowledge base can be requested.

[0088] The monitoring equipment can also generate alarm information based on the equipment status monitoring results and fault self-healing records. The alarm information is divided into three levels according to the fault level: emergency alarm, important alarm and general alarm, and is reported using different communication resources. It also periodically (e.g., monthly or quarterly) summarizes the equipment status statistics, sampling strategy change records, power consumption event records, fault self-healing records, battery health trends and data quality control statistics to generate structured operation and maintenance reports for storage, archiving and remote reporting.

[0089] In this way, the knowledge base accumulated by the historical operating condition partitions with each reliable operation can support more accurate and robust strategy matching and reasoning decisions in the future.

[0090] Based on the same inventive concept as the adaptive observation method provided in this embodiment, this embodiment also provides an adaptive observation device, which includes at least one software functional module that can be stored in a memory or embedded in an electronic device. The software functional module and computer program included in the electronic device, etc., are described below. Please refer to... Figure 5 Functionally, the device may include: The operating condition monitoring module 11 is used to acquire real-time operating condition characteristic information from multiple sensors; Knowledge enhancement module 12 is used to obtain retrieval results from the permafrost observation knowledge base using real-time working condition feature information; The strategy execution module 13 is used to call the locally configured task inference model to process the inference information, including real-time operating condition feature information and retrieval results, to obtain an observation strategy that matches the current operating condition.

[0091] Optionally, the permafrost observation knowledge base includes permafrost freeze-thaw mechanism rule partitioning, historical working condition partitioning, sampling strategy rule partitioning, and equipment operation and maintenance strategy partitioning; The search results include phase transition stage determination rules, approximate operating condition entries, target sampling strategy templates, and target operation and maintenance strategy templates; The knowledge enhancement module 12 utilizes real-time operating condition feature information to obtain retrieval results from the permafrost observation knowledge base in the following ways: Real-time working condition feature information is used to match phase change stage determination rules from the frozen soil freeze-thaw mechanism rule partitioning. Based on the real-time operating condition feature information, obtain the approximate operating condition entry that is closest to the feature distance of the real-time operating condition feature information from the historical operating condition partition. The target sampling strategy template is matched from the sampling strategy rule partition using real-time operating condition feature information; The target operation and maintenance strategy template is matched from the equipment operation and maintenance strategy partition using real-time operating condition feature information.

[0092] Optionally, the inference information also includes the hardware constraint parameters and power supply status parameters of the observation equipment.

[0093] Optionally, the strategy execution module 13 calls the locally configured task inference model to process the inference information, including real-time operating condition feature information and retrieval results, to obtain an observation strategy that matches the current operating condition, including: The locally configured task inference model is invoked to process real-time operating condition characteristics, retrieval results, hardware constraint parameters, and power supply status parameters to obtain the observation strategy.

[0094] Optionally, the task reasoning model is obtained by knowledge distillation, structural pruning, and quantization compression of a large model deployed in the cloud.

[0095] Optionally, the observation strategy includes recommended instructions selected from a preset control instruction set and a confidence level. The knowledge enhancement module 12 is also used to determine whether there is a fallback instruction in the permafrost observation knowledge base that matches the real-time working condition feature information and has the highest priority if the confidence level is less than the reliability threshold. If so, execute the fallback instruction; If not, then the historical control instructions that were successfully executed in the previous cycle will be used.

[0096] Optionally, the permafrost observation knowledge base includes historical working condition partitions, and the real-time working condition feature information is obtained by quality control processing of the original observation data. The strategy execution module 13 is also used to obtain the maximum similarity of the real-time working condition feature information in the historical working condition partitions if the observation strategy is executed successfully and no alarm is triggered. If the maximum similarity is less than the deduplication threshold, a knowledge entry including real-time working condition feature information, observation strategy, and the execution result of the observation strategy is generated and updated in the permafrost observation knowledge base.

[0097] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0098] It should also be understood that if the above embodiments are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0099] Therefore, this embodiment also provides a storage medium, which is a computer-readable storage medium. This storage medium stores a computer program, which, when executed by a processor, implements the adaptive observation method provided in this embodiment. The storage medium can be any medium capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0100] Please refer to Figure 6 The electronic device provided in this embodiment may include a processor 22 and a memory 21. The memory 21 stores a computer program, and the processor implements the adaptive observation method provided in this embodiment by reading and executing the computer program corresponding to the above-described embodiments in the memory 21.

[0101] See also Figure 6 The electronic device also includes a communication unit 23. The memory 21, processor 22 and communication unit 23 are electrically connected to each other directly or indirectly through system bus 24 to realize data transmission or interaction.

[0102] The memory 21 can be an information recording device based on any electronic, magnetic, optical, or other physical principles, used to record execution instructions, data, etc. In some embodiments, the memory 21 can be, but is not limited to, volatile memory, non-volatile memory, memory drive, etc.

[0103] In some embodiments, the volatile memory may be random access memory (RAM); in some embodiments, the non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, etc.; in some embodiments, the storage drive may be a disk drive, solid-state drive, any type of storage disk (such as optical disc, DVD, etc.), or similar storage media, or a combination thereof.

[0104] The communication unit 23 is used for sending and receiving data, and can be, but is not limited to, network, Bluetooth, infrared, serial port, or near-field wireless communication methods. For example, LoRa spread spectrum communication, BeiDou short message communication, low-power cellular communication, or other on-demand wake-up communication methods can be used. When communication conditions permit, it mainly uploads summary data, alarm information, and key quality control results to reduce communication energy consumption and communication costs. Under conditions of no network or weak network, it enters a sleep mode while maintaining the task of generating observation strategies.

[0105] In some implementations, the network may include a wired network, a wireless network, a fiber optic network, a telecommunications network, an intranet, the Internet, a local area network (LAN), a wide area network (WAN), wireless local area networks (WLANs), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network, or a near field communication (NFC) network, or any combination thereof. In some embodiments, the network may include one or more network access points. For example, the network may include wired or wireless network access points, such as base stations and / or network switching nodes, through which one or more components of the service request processing system can connect to the network to exchange data and / or information.

[0106] The processor 22 may be an integrated circuit chip with signal processing capabilities, and may include one or more processing cores (e.g., a single-core processor or a multi-core processor). By way of example only, the processor described above may include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), an Application Specific Instruction-set Processor (ASIP), a Graphics Processing Unit (GPU), a Physics Processing Unit (PPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), a Programmable Logic Device (PLD), a controller, a microcontroller unit, a Reduced Instruction Set Computing (RISC) computer, or a microprocessor, or any combination thereof.

[0107] Understandable. Figure 6 The structure shown is for illustrative purposes only. Electronic devices may also have more advanced features. Figure 6 Showing more or fewer components, or having with Figure 6 The different configurations shown. Figure 6 The components shown can be implemented using hardware, software, or a combination thereof.

[0108] It should be understood that the apparatus and methods disclosed in the above embodiments can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0109] The above descriptions are merely various embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive observation method, characterized in that, The method includes: Acquire real-time operating condition characteristic information from multiple sensors; The search results are obtained from the permafrost observation knowledge base using the real-time operating condition feature information. The locally configured task inference model is invoked to process the inference information, including the real-time operating condition feature information and the retrieval results, to obtain an observation strategy that matches the current operating condition.

2. The adaptive observation method according to claim 1, characterized in that, The permafrost observation knowledge base includes permafrost freeze-thaw mechanism rule partitioning, historical working condition partitioning, sampling strategy rule partitioning, and equipment operation and maintenance strategy partitioning. The search results include phase transition stage determination rules, approximate operating condition entries, target sampling strategy templates, and target operation and maintenance strategy templates. Retrieval results from the permafrost observation knowledge base using the aforementioned real-time operating condition feature information include: The phase transition stage determination rule is matched from the frozen soil freeze-thaw mechanism rule partition using the real-time working condition feature information. Based on the real-time operating condition feature information, obtain the approximate operating condition entry that has the closest feature distance to the real-time operating condition feature information from the historical operating condition partition; The target sampling strategy template is matched from the sampling strategy rule partition using the real-time operating condition feature information; The target operation and maintenance strategy template is matched from the equipment operation and maintenance strategy partition using the real-time operating condition feature information.

3. The adaptive observation method according to claim 1, characterized in that, The inference information also includes the hardware constraint parameters and power supply status parameters of the observation equipment.

4. The adaptive observation method according to claim 3, characterized in that, The locally configured task inference model is invoked to process the inference information, including the real-time operating condition feature information and the retrieval results, to obtain an observation strategy that matches the current operating condition, including: The observation strategy is obtained by calling the locally configured task inference model to process the real-time operating condition feature information, the retrieval results, the hardware constraint parameters, and the power supply status parameters.

5. The adaptive observation method according to claim 1, characterized in that, The task reasoning model is obtained by knowledge distillation, structural pruning, and quantization compression of a large model deployed in the cloud.

6. The adaptive observation method according to claim 1, characterized in that, The observation strategy includes recommended instructions selected from a preset set of control instructions and confidence levels, and the method further includes: If the confidence level is less than the reliability threshold, then it is determined whether there is a fallback instruction in the frozen soil observation knowledge base that matches the real-time working condition feature information and has the highest priority. If so, then execute the fallback instruction; If not, then the historical control instructions that were successfully executed in the previous cycle will be used.

7. The adaptive observation method according to claim 1, characterized in that, The permafrost observation knowledge base includes historical working condition partitions, and the real-time working condition feature information is obtained by quality control processing of the raw observation data. The method further includes: If the observation strategy is executed successfully and no alarm is triggered, the maximum similarity of the real-time operating condition feature information in the historical operating condition partition is obtained. If the maximum similarity is less than the deduplication threshold, a knowledge entry including the real-time working condition feature information, the observation strategy, and the execution result of the observation strategy is generated and updated in the permafrost observation knowledge base.

8. An adaptive observation device, characterized in that, The device includes: The operating condition monitoring module is used to acquire real-time operating condition characteristic information from multiple sensors; The knowledge enhancement module is used to obtain retrieval results from the permafrost observation knowledge base using the real-time working condition feature information; The strategy execution module is used to call the locally configured task inference model to process the inference information, including the real-time operating condition feature information and the retrieval results, to obtain an observation strategy that matches the current operating condition.

9. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the adaptive observation method according to any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the adaptive observation method according to any one of claims 1-7.

Images