A method and system for in-situ measurement of electrical conductivity in saline-alkali paddy fields

Abstract

This invention relates to the field of farmland water and salinity regulation technology, and particularly to a method and system for in-situ measurement of electrical conductivity in saline-alkali paddy fields. The method includes: acquiring geometric and baseline data; obtaining a complex transfer response matrix through complex frequency excitation and synchronous detection; applying thermal pulses and salinity steps within the same measurement range; calculating the electro-thermal kernel function and the electro-salt kernel function; compensating for and calibrating the observations; assembling a time-varying top boundary forward model and performing multi-frequency joint inversion to obtain the root zone equivalent conductivity and pore water conductivity; forming an intervention level map based on the early salinization index, connectivity index, and hotspot mask; outputting parameters for intermittent rinsing, water source mixing, and drainage linkage; and updating the next round of channels and frequencies based on the criterion of maximizing information content to achieve closed-loop control. This invention improves dimensional consistency, disturbance resistance, and spatial resolution.

Description

Technical Field

[0001] This invention relates to the field of farmland water and salt regulation technology, and in particular to a method and system for in-situ measurement of electrical conductivity in saline-alkali paddy fields. Background Technology

[0002] The spatial and temporal patterns of salinity directly affect seedling survival, tillering, and grain-filling quality, and are key variables restricting yield and yield stability. On the production side, there is a need for continuously operating electrical proxy indicators at the field scale to support field zoning mapping, irrigation and drainage start-up and shutdown criteria, salinity return early warning, variety salt tolerance comparison, and engineering modification evaluation. On the water conservancy side, there is a need to parameterize water allocation, drainage, and canal system scheduling to reduce trial-and-error costs and improve water resource utilization efficiency. On the industry side, there is a need to integrate ground monitoring with remote sensing inversion and yield prediction models to form a usable closed loop from monitoring to decision-making. Summary of the Invention

[0003] To address the numerous problems existing in the prior art, this invention provides a method and system for in-situ measurement of electrical conductivity in saline-alkali paddy fields. This invention deploys an electrode array and a water layer sensor in the field, applies a complex frequency current, and introduces a thermal pulse and a salinity step within the same measurement range. It calculates the electro-thermal kernel function and the electro-salt kernel function, compensates for the complex transfer response, and calibrates its dimensions. A time-varying top boundary is assembled, and multi-frequency joint inversion is performed to obtain the root zone equivalent conductivity and pore water conductivity, generating an early salinization index and a connectivity index, thus realizing a closed loop of monitoring-decision-control.

[0004] A method for in-situ determination of electrical conductivity in saline-alkali paddy fields includes the following steps: Acquire topographic and irrigation / drainage information, set up measurement profiles, establish electrode arrays and water layer conductivity measurement devices, generate injection and receiving channel sets, acquire water layer conductivity and water depth baselines, and form geometric models and boundary baseline data; A current excitation and synchronous detection are applied to the injection channel at a preset frequency set to obtain the complex transfer response matrix; thermal pulses and salinity steps are implemented in the same measurement range, and the electro-salt kernel function and electro-thermal kernel function are calculated based on the response and the change in water layer conductivity. The complex transfer response is compensated and dimensionally calibrated based on the electro-salt kernel function and the electro-thermal kernel function. A forward model with a time-varying top boundary is assembled and multi-frequency joint inversion is performed to obtain the root region equivalent conductivity and pore water conductivity distribution. The time derivative is calculated and the early salt return index, connectivity index and hotspot mask are generated. A control plan is generated, and the next round of channel and frequency plans are updated after retesting and inversion.

[0005] Preferably, acquiring topographic and irrigation / drainage information includes determining the direction and location of the measurement profile based on topographic data; establishing an electrode array and a water layer conductivity measurement device includes connecting electrodes through a multiplexed switch matrix to generate an injection channel and a receiving channel set; acquiring water layer conductivity and water depth baselines and forming a geometric model and boundary baseline data includes recording the electrode spatial coordinates and the zero-point offset of the water layer conductivity measurement device.

[0006] Preferably, applying current excitation to the injection channel at a preset frequency set includes using a parallel multi-sine coding sequence; synchronous detection includes using phase-locked signal processing to extract amplitude and phase; and performing a reciprocity check before generating the complex transfer response matrix to eliminate unqualified channels.

[0007] Preferably, implementing thermal pulses and salinity steps within the same measurement range includes sequentially applying thermal pulses and salinity steps; calculating the electro-salt kernel function and the electro-thermal kernel function includes dual-input deconvolution with the water layer conductivity change time series and temperature rise time series as inputs and the complex transfer response change time series as output, and applying causal constraints and non-negativity constraints to the dual-input deconvolution.

[0008] Preferably, the complex transfer response is compensated for the effects of temperature and water content based on the electro-thermal kernel function, and the complex transfer response is dimensionally calibrated based on the electro-salt kernel function. The compensated and dimensionally calibrated complex transfer response is then converted into an observation consistent with the conductivity dimension for multi-frequency joint inversion.

[0009] Preferably, the assembly includes a forward model with a time-varying top boundary, which employs a quasi-static conductivity equation and Robin boundary conditions; the multi-frequency joint inversion includes constructing a multi-frequency data consistency term and a regularization term, the regularization term including anisotropic total variation regularization, structural sparsity regularization derived from temperature change time series, and salt conservation constraints.

[0010] Preferably, the multi-frequency joint inversion is solved using an iterative optimization method, the data consistency term uses a robust loss function to suppress abnormal channels, and a non-negative constraint is applied to the equivalent conductivity distribution in the root region.

[0011] Preferably, the early salt return index is calculated based on the non-negative part of the time derivative of the equivalent conductivity distribution in the root region, the connectivity index is calculated based on the ratio of the conductivity gradient to the conductivity gradient norm along the measurement profile direction, and the hotspot mask is generated based on the confidence threshold and connected component discrimination.

[0012] Preferably, the generation of the control plan includes forming an intervention level map based on the early salt return index, connectivity index, and hotspot mask, and generating intermittent rinsing parameters, water source mixing parameters, and drainage linkage parameters; updating the next round of channel and frequency plan after retesting and inversion includes selecting the injection channel and frequency set based on the criterion of maximizing the information content of the sensitivity matrix.

[0013] A system for in-situ measurement of electrical conductivity in saline-alkali paddy fields, used to implement the aforementioned method for in-situ measurement of electrical conductivity in saline-alkali paddy fields, the system comprising: The baseline module is used to acquire topographic and irrigation / drainage information, set up measurement profiles, establish electrode arrays and water layer conductivity measurement devices, generate injection and receiving channel sets, acquire water layer conductivity and water depth baselines, and form geometric models and boundary baseline data. The geometric models, boundary baseline data, and injection and receiving channel sets are then provided to the controlled measurement module. The controlled measurement module is used to apply current excitation to the injection channel at a preset frequency set and simultaneously detect to obtain the complex transfer response matrix. It implements thermal pulse and salinity step in the same measurement range, calculates the electro-salt kernel function and electro-thermal kernel function based on the complex transfer response and the change in water layer conductivity, and provides the complex transfer response matrix, electro-salt kernel function, electro-thermal kernel function and the change in water layer conductivity to the inversion diagnostic module. The inversion diagnostic module is used to compensate and dimensionally calibrate the complex transfer response based on the electro-salt kernel function and the electro-thermal kernel function. It assembles a forward model containing a time-varying top boundary and performs multi-frequency joint inversion to obtain the root zone equivalent conductivity distribution and pore water conductivity distribution. It calculates the time derivative and generates the early salt return index, connectivity index and hot spot mask, and provides the above results to the control closed-loop module. The control closed-loop module is used to generate a control plan based on the root region equivalent conductivity distribution, early salt return index, connectivity index and hotspot mask, organize retesting and inversion to form updated results, and update the next round of channel and frequency plans accordingly and feed them back to the controlled measurement module and inversion diagnosis module.

[0014] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows: By using parallel multi-sinusoidal complex frequency excitation and synchronous detection, the effect of robustly acquiring amplitude and phase information and improving the signal-to-noise ratio within a single measurement range is achieved. By using thermal pulse and salinity step calibration with the same measurement range, compensation for the effects of temperature and water content and dimensional calibration of electrical observations were achieved. By assembling a time-varying top boundary and performing multi-frequency joint inversion, an adaptive fitting effect on the changes in field surface water depth and water layer conductivity was achieved, reducing model mismatch. By using anisotropic total variation and structural sparsity and salt conservation constraints derived from temperature change time series, clear boundaries, artifact suppression and physical consistency were achieved. By generating early salt return index, connectivity index and hotspot mask, the risk classification and spatial priority determination effect for operation is achieved; By updating channels and frequencies based on the criterion of maximizing information content of the sensitivity matrix, the self-optimization of the measurement range design and the improvement of the input-output ratio were achieved. By using intermittent rinsing, water source mixing, and parameterized output of drainage linkage, the engineering implementation effect of monitoring-decision-execution was achieved. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of the method of the present invention; Figure 2 This is a structural block diagram of the system of the present invention. Detailed Implementation

[0016] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0017] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0018] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0019] like Figure 1 As shown, a method for in-situ determination of electrical conductivity in saline-alkali paddy fields includes the following steps: Acquire topographic and irrigation / drainage information, set up measurement profiles, establish electrode arrays and water layer conductivity measurement devices, generate injection and receiving channel sets, acquire water layer conductivity and water depth baselines, and form geometric models and boundary baseline data; Preferably, acquiring topographic and irrigation / drainage information includes determining the direction and location of the measurement profile based on topographic data; establishing an electrode array and a water layer conductivity measurement device includes connecting electrodes through a multiplexed switch matrix to generate an injection channel and a receiving channel set; acquiring water layer conductivity and water depth baselines and forming a geometric model and boundary baseline data includes recording the electrode spatial coordinates and the zero-point offset of the water layer conductivity measurement device.

[0020] This embodiment focuses on the field setup and baseline establishment in saline-alkali paddy fields. The goal is to generate reusable geometric models and boundary baseline data without altering field production conditions, providing stable input for subsequent measurements and inversions. The method follows the sequence of "information acquisition - profile determination - device deployment - channel generation - baseline calibration - data archiving".

[0021] First, topographic and irrigation / drainage information is acquired. Based on elevation data generated by GPS or aerial surveying, and vector maps of irrigation ditches and field ridges, slope aspect and water runoff paths are extracted to determine the direction and location of the measurement profile. The profile should span low-lying areas and drainage boundaries, and its length should cover the planned monitoring zone. The start and end coordinates and orientation angles of the profile are output as the deployment reference.

[0022] An electrode array and a water layer conductivity measuring device were linearly deployed along the profile. Corrosion-resistant thin metal rods were used as electrodes, equidistantly buried in the lower part of the cultivated layer to ensure stable contact resistance. The water layer conductivity measuring device was installed at the shallow water ends of the profile, with the sensor head submerged in the field water layer. Its installation height, cable routing, and environmental obstructions were all recorded. Electrode leads were uniformly connected to a multiplexed switch matrix, forming programmable injection and receiving ports. After wiring was completed, no-load noise and contact resistance were read channel by channel, and abnormal ports were eliminated and reset.

[0023] Generate sets of injection and reception channels. Based on three geometric relationships—adjacent, spanning, and long base distance—pre-select a set of channel pairs, and then use a simple scoring function to filter for combinations with better coverage and discriminability. The scoring function is as follows: ,in, Scoring of channel candidates This is the normalized value for the electrode spacing. This is the normalized value of the geometric coverage angle. This is the normalized value of the contact resistance at both ports. , , The scores are used as weights. Injection and receiving channels are selected from highest to lowest and saved as a configuration file for easy reuse later.

[0024] The water layer conductivity and water depth baseline were obtained. Conductivity was continuously acquired and water temperature was recorded using a four-electrode method, and the readings were converted to temperature using a two-point water temperature calibration coefficient. Water depth was measured using a pressure or acoustic sensor, and zero-point offset correction was performed. The calculation relationship is as follows: ,in, This is the actual water depth. For instrument readings, The zero-point offset was recorded during installation. The stable segment of conductivity versus water depth was used as the baseline value, and the original time series was retained for comparison.

[0025] A geometric model and boundary baseline data are generated. The geometric model includes profile lines, three-dimensional coordinates of each electrode, sensor locations, and cable routing. The boundary baseline data includes the water layer conductivity baseline, water depth baseline, temperature conversion factor, and zero-point offset of the water layer conductivity measuring device. To ensure consistency, the same coordinate system and unit system are used uniformly, and all fields are accompanied by the acquisition time, equipment number, and personnel signature for easy traceability.

[0026] The principle of this embodiment is as follows: by quantitatively identifying the terrain and irrigation / drainage patterns, the profile is aligned with the main hydraulic direction, improving the identifiability of subsequent current paths; by using a multiplexed switch matrix, discrete electrodes are abstracted into programmable channel resources, facilitating rapid switching of excitation and acquisition combinations in subsequent measurements; and by zero-point and temperature correction of water layer conductivity and depth, a reliable top boundary baseline is provided to the forward model. The effects are reflected in improved channel reciprocity pass rate, enhanced consistency of repeated measurements, and improved comparability between data from different dates.

[0027] To facilitate implementation, it is recommended to encapsulate profile design, channel generation, and baseline acquisition into a one-click workflow: input topographic and irrigation / drainage data, output a complete engineering package containing profile geometry, channel list, and baseline parameters. This engineering package serves as the standardized input for subsequent measurements and inversions, and can be directly used to generate field work orders and data entry templates.

[0028] A current excitation and synchronous detection are applied to the injection channel at a preset frequency set to obtain the complex transfer response matrix; thermal pulses and salinity steps are implemented in the same measurement range, and the electro-salt kernel function and electro-thermal kernel function are calculated based on the response and the change in water layer conductivity. This embodiment provides a practical process and algorithm for controlled current excitation, synchronous detection, and dual-pulse calibration within the same measurement range.

[0029] First, parallel multi-sinusoidal current sequences are generated for each injection channel from a preset frequency set, with amplitudes limited to ensure safety for personnel and crops. The acquisition end records the voltage signals of each receiving channel at a uniform sampling rate, and performs DC-DC removal and power frequency interference suppression on the raw data. Phase-locked detection is performed for each frequency to calculate the quadrature components at the same frequency, obtaining the amplitude and phase, which are then organized into a complex transfer response matrix according to the combination of injection and receiving channels. To ensure quality, reciprocity and repeatability checks are performed, rejecting unqualified channels and recording the mask.

[0030] Thermal pulses and salinity steps are performed sequentially within the same measurement window. Thermal pulses involve short-term heating via an electrically heated strip laid along the profile, with temperature sensors on the surface recording the temperature rise time series. Salinity steps involve adding a pre-weighed sodium chloride solution to the water layer, with the water layer conductivity measuring device continuously recording the water layer conductivity change time series, and short-term stirring after addition to achieve homogenization. Neither of these operations alters the electrode and channel configuration; only perturbation calibration is performed on the same complex transfer response within the same window.

[0031] Align the changes in the complex transfer response with the two input sequences along a unified time axis to form a two-input deconvolution problem. Select a representative output quantity as the calibration object; this can be the magnitude or phase change of the complex transfer response. Employ a regularized least-squares estimation kernel function and apply causal and non-negative projection constraints in the time domain, iterating until the residuals converge. To improve robustness, an initial solution can be provided in the frequency domain first, followed by applying constraints and slight smoothing in the time domain.

[0032] Core computing uses the following expressions:

[0033]

[0034]

[0035] in, To relay the response, For frequency, For integration time, For receiving channel voltage time series, For the selected response change time series, This is a time series of changes in the electrical conductivity of the water layer. This is a time series of temperature rises. For the electro-salt kernel function, For electrothermal kernel functions, For noise terms, and is the regularization coefficient, and * is the convolution operator. For Imposing nonnegativity and causality constraints on Applying causal constraints, in this invention A unified representation of time variables.

[0036] In terms of system implementation, the signal source and current source are connected to the electrodes via a multiplexed switch matrix. The data acquisition equipment synchronously acquires the voltage and reference signal from the receiving channel, as well as data from water layer conductivity and temperature sensors. The control unit provides the measurement script: the timing and safety thresholds for the baseline segment, frequency segment, thermal pulse segment, and salinity step segment. The data processing module completes time synchronization, quality control, phase-locked detection, matrix assembly, and deconvolution solving, and outputs the kernel function, confidence index, and quality label.

[0037] The principle of this embodiment is to establish a traceable relationship between the electrical response and aquifer conditions using controlled thermal and salinity perturbations within the same geometry and time window. The effects are threefold: first, consistent extraction of the complex transfer response across multiple frequencies; second, anchoring the electrical response dimensions to changes in aquifer conductivity and temperature through kernel functions; and third, providing a stable basis for subsequent inversion and compensation. The above process relies only on common hardware and implementable software algorithms, facilitating repeated execution and large-scale deployment in the field.

[0038] Preferably, applying current excitation to the injection channel at a preset frequency set includes using a parallel multi-sine coding sequence; synchronous detection includes using phase-locked signal processing to extract amplitude and phase; and performing a reciprocity check before generating the complex transfer response matrix to eliminate unqualified channels.

[0039] This embodiment provides a practical process and algorithm for controlled current excitation, synchronous detection, and reciprocal control within the same measurement range, generating a complex transfer response matrix as the basis for subsequent calculations.

[0040] The system consists of a signal source and a constant current source, a switch matrix, acquisition equipment, a reference signal generator, and control software. The control unit generates a parallel multi-sine injection sequence according to a preset frequency set. For each injection channel, excitation and acquisition of all frequencies are completed within a single measurement range, reducing deviations caused by environmental changes. To reduce peak values ​​and inter-channel crosstalk, a fixed amplitude and adjustable phase are used. The phase is updated before each startup through simple iteration, resulting in a low peak-to-average waveform and low correlation between different injection channels at the same frequency. The injection sequence is denoted as:

[0041] in, For the injected current waveform, The number of frequencies For the first One frequency, The amplitude at that frequency. This refers to the phase of that frequency. The upper limit of the amplitude is determined by the safety of personnel and crops, and the specific value is set in the control software parameters.

[0042] The acquisition unit records the voltage time series of each receiving channel at a uniform sampling rate, and simultaneously records the reference sine and cosine. To suppress power frequency and wireless interference, band-limited filtering and DC component removal are used in the pre-processing stage, with the filtering bandwidth covering all preset frequencies. Phase-locked detection calculates the in-phase and quadrature components at each frequency, forming information including amplitude and phase, which is then assembled into a complex response value according to the combination of injection and reception. The calculation of the complex transfer response uses:

[0043] in, To relay the response, To receive channel voltage, For integration time, The target frequency is used. The integration time and frequency set are uniformly set in the task configuration.

[0044] To ensure data reliability, reciprocity control is performed before forming the complex transfer response matrix. Specifically, two measurements are taken for each electrode pair: "injected at A, received at B" and "injected at B, received at A," to obtain... and The relative difference in amplitude and the phase difference are used as the criteria for judgment.

[0045]

[0046] in, For the relative difference in amplitude, For phase difference, For amplitude, For phase. If any index at any frequency exceeds the preset threshold, the channel is marked as unqualified and removed from the matrix construction, and the reason is recorded for on-site inspection of contact resistance and wiring.

[0047] The data processing software workflow includes: task parameter loading, phase iteration solution, waveform distribution and execution, raw data acquisition, time synchronization and preprocessing, phase-locked loop detection, reciprocity control and mask generation, and complex transfer response matrix writing. For traceability, the matrix file includes timestamps, injection and receiving channel indices, a frequency list, integration time, threshold settings, and reciprocity flags. The principle of this workflow is to project the voltage signal onto a basis function of the same frequency under a sinusoidal reference of known frequency, thereby robustly obtaining the amplitude and phase directly related to the electrical parameters. The reciprocity check utilizes the premise of dielectric linearity and boundary stability to quickly expose poor contact and wiring errors.

[0048] The effectiveness of this embodiment is reflected in three aspects: First, parallel multi-sine waves and unified sampling shorten the measurement time and reduce the impact of environmental fluctuations; second, phase-locked detection can still stably extract amplitude and phase information under low signal-to-noise ratio, improving the usable bandwidth of the matrix; and third, reciprocal quality control can detect abnormal channels on-site, preventing bad data from entering subsequent calculations. The entire process relies on common electrical survey hardware and a general data acquisition platform. Parameters and thresholds can be adjusted on-site according to the electrical properties and noise conditions of farmland, facilitating promotion and replication.

[0049] Preferably, implementing thermal pulses and salinity steps within the same measurement range includes sequentially applying thermal pulses and salinity steps; calculating the electro-salt kernel function and the electro-thermal kernel function includes dual-input deconvolution with the water layer conductivity change time series and temperature rise time series as inputs and the complex transfer response change time series as output, and applying causal constraints and non-negativity constraints to the dual-input deconvolution.

[0050] This embodiment provides an implementable process and algorithm for completing thermal pulses and salinity steps within the same measurement range, thereby obtaining the electro-salt kernel function and the electro-thermal kernel function.

[0051] The process is executed sequentially: First, an electric heating belt is activated on the cross-section to generate a thermal pulse. The power and duration are set by the control unit, and a surface temperature sensor records the temperature rise time series. After the thermal pulse ends, a natural cooling phase begins, maintaining the channel and frequency configuration while continuously acquiring the complex transfer response. Subsequently, a pre-weighed sodium chloride solution is added to the surface water layer and briefly stirred. The water layer conductivity measuring device records the conductivity change time series, and the complex transfer response continues to be continuously acquired. To avoid excessive disturbance, the dosage is set to the target conductivity increment using the following formula: ,in, For the target conductivity increment, For salt quality, Because of the water depth, For the area of ​​action, This is a temperature conversion factor. The control unit provides dosage prompts and records the dosing time.

[0052] Data alignment uses a unified time axis. The temperature rise time series and the water layer conductivity change time series are aligned to the sampling period using linear interpolation, and the complex transfer response is taken as the change at the same time baseline as both. To compensate for the hysteresis caused by mixing and diffusion, the time delays of the two inputs are first estimated by cross-correlation with the changes in the complex transfer response, and integer sample shift is performed to avoid misinterpreting the time delay as the shape of the kernel function during deconvolution.

[0053] Kernel function estimation uses a two-input convolution model: ,in, For the change in the complex transitive response, This is a time series of changes in the electrical conductivity of the water layer. This is a time series of temperature rises. For the electro-salt kernel function, For electrothermal kernel functions, For noise terms, For convolution. Discrete implementation uses least squares and L2 regularization:

[0054] The constraints are causality and non-negativity: the kernel function is set to zero before time zero, and the negative components of the electro-salt kernel function are truncated to zero after each iteration. Regularization coefficients are automatically initialized based on baseline noise power and input energy, and then fine-tuned using the residual-curvature piecewise linear method. The solution process involves an initial solution in the frequency domain followed by time-domain projection: a preliminary kernel function is obtained using fast convolution, then causal and non-negativity projection and mild smoothing are performed in the time domain, iterating repeatedly until the residual change is less than a set threshold. To prevent instability caused by excessively high input correlation, energy normalization is performed on the time series of water layer conductivity changes and temperature rise before processing, and the kernel function time window is shortened when the cross-correlation exceeds a threshold.

[0055] Quality control comprises three aspects: first, verifying the consistency between residual variance and baseline noise; second, checking the range of kernel function energy and peak time to exclude results significantly exceeding physical expectations; and third, rapidly retesting a channel using repeated small-dose salinity steps to compare the consistency of the two electro-salt kernel functions. Qualified results are stored in an array, including kernel function values, sampling intervals, and start times, along with confidence indicators and processing logs, for direct retrieval in subsequent compensation and dimensional calibration.

[0056] The principle of this process is to establish a mapping between complex transfer response changes and water layer conditions using small-amplitude, measurable thermal and salinity disturbances under fixed geometry and frequency configurations. The effects are as follows: thermal pulses provide a basis for compensating for the influence of temperature and water content, and salinity steps provide a calibration scale for the electrical response to conductivity dimensions. Together, these ensure that the data terms and boundary terms in subsequent inversion have a stable and consistent source. The hardware uses common heating belts, temperature sensors, and conductivity sensors. Control and solution are implemented by a general-purpose data acquisition and calculation module, facilitating repeated execution and portability in the field.

[0057] The complex transfer response is compensated and dimensionally calibrated based on the electro-salt kernel function and the electro-thermal kernel function. A forward model with a time-varying top boundary is assembled and multi-frequency joint inversion is performed to obtain the root region equivalent conductivity and pore water conductivity distribution. The time derivative is calculated and the early salt return index, connectivity index and hotspot mask are generated. A control plan is generated, and the next round of channel and frequency plans are updated after retesting and inversion.

[0058] In this embodiment, compensation and dimensional calibration, forward assembly and multi-frequency joint inversion are completed on the same dataset. Based on the results, zoning indicators and control plans are generated, and finally, the channel and frequency plans are updated with short-term retest closed loop.

[0059] First, temperature and moisture content effects are compensated for in the complex transfer response, and dimensional calibration is performed. Using the electrothermal kernel function and the electrosalt kernel function as calibration bases, the correction is calculated after alignment with a unified time axis. ,in, For the change in the complex transitive response, For electrothermal kernel function, This is a temperature rise time series, and * denotes convolution. The steady-state gain of the electric salt kernel function is then used as the dimensional anchor point:

[0060]

[0061] in, For calibrated frequency domain observations, For Fourier transform, For the electro-salt kernel function, For its steady-state gain, This is the integration window. Compensation and calibration are performed automatically by the software module, which outputs a weighted observation vector and quality label.

[0062] The assembly includes a forward model with a time-varying top boundary. Based on the water layer conductivity and water depth time series obtained in the previous stage, the change of the top boundary transducer coefficient over time is calculated and integrated with the electrode geometry and mesh to form a forward operator. Numerical implementation employs finite element discretization and single-frequency decoupled solution to reduce computational load and maintain consistency with multi-frequency data.

[0063] The joint inversion uses multi-frequency data consistency as the primary term and sparsity and anisotropic constraints as secondary terms to solve for the equivalent conductivity distribution in the root region. The objective function is denoted as:

[0064] in, For the equivalent conductivity distribution, For a set of frequencies, Forward operator, For frequency band weighting, For total variation, The unit vector is connected in the main direction. The coefficients are regularization factors. The solution employs iterative optimization with non-negativity constraints applied to the equivalent conductivity distribution. The termination criterion is the convergence of the data residuals and the objective function. To ensure feasibility, a starting parameter template and an automatic adjustment strategy are provided, prioritizing convergence to a data-consistent solution and then gradually enhancing sparsity and anisotropy to stabilize the boundary.

[0065] The pore water conductivity distribution is obtained from the equivalent conductivity distribution. A monotonic mapping function is established based on the indoor calibration of soil samples to convert the equivalent conductivity into pore water conductivity. ,in, The electrical conductivity of pore water, Equivalent conductivity For the monotonic function calibrated on site, This represents a spatial location vector, i.e., a point location within the soil inversion domain of the root zone. The mapping adopts a piecewise linear or log-linear form, with parameters obtained from soil column tests or historical data fitting. The time derivative and discriminant index are generated. The time derivative field is obtained by differencing the equivalent conductivity between two adjacent measurements:

[0066] in, For time derivative, The time interval is the duration of two measurements. The early salt return index is calculated using the non-negative part of the time derivative: The connectivity index is taken as the gradient percentage along the main direction:

[0067] in, To prevent small values ​​with zero denominators, a hotspot mask is generated based on the uncertainty and the area of ​​the connected region. The output vector boundary is used to control the partitioning.

[0068] A control plan is formulated. The three indicators are synthesized using a scoring function: ,in, For rating, Non-negative truncation, The conductivity threshold, Assign weights. For connected regions within the hotspot mask, assign scores and grades to generate control parameters for intermittent rinsing, water source mixing, and drainage linkage, and create an execution list and schedule.

[0069] After execution, a short-term retest is triggered, and the channel and frequency plan is updated in a closed loop. The retest uses a small number of channels and representative frequencies to quickly compare the difference in equivalent conductivity and the change in time derivative before and after the intervention. To improve the information content of the next round of measurements, a new set of channels and frequencies is selected based on the sensitivity matrix. ,in, For the candidate set, This is the sensitivity matrix. The control unit updates the injection and reception combinations and frequency list based on this matrix and writes it back to the field script.

[0070] The principle of this embodiment is to complete the physical compensation and dimensional unification of data under the same caliber, and then solve the electrical distribution through multi-frequency consistency and structural constraint stability. Over time, indicators reflecting the salt return trend and hydraulic channels are generated, and finally the diagnosis is transformed into actionable control parameters. The entire process relies on common measurement and control equipment and a general numerical solution framework. The parameters can be adjusted according to field conditions, which is convenient for engineering deployment and long-term operation.

[0071] Preferably, the complex transfer response is compensated for the effects of temperature and water content based on the electro-thermal kernel function, and the complex transfer response is dimensionally calibrated based on the electro-salt kernel function. The compensated and dimensionally calibrated complex transfer response is then converted into an observation consistent with the conductivity dimension for multi-frequency joint inversion.

[0072] This embodiment provides a directly implementable process and core calculations for the processing step of "compensation and dimensional calibration - construction of invertible observations".

[0073] The inputs are the time series of complex transfer response changes, temperature rise time series, water layer conductivity change time series, electrothermal kernel function, and electrosalt kernel function. First, the three types of time series are resampled at a uniform sampling step size, and the time delays of the input and output are determined by cross-correlation. Then, integer sample alignment is performed, followed by compensation and calibration calculations.

[0074] The effects of temperature and moisture content are compensated by deconvolution subtraction of the complex transfer response using an electrothermal kernel function: ,in, For the change in the complex transitive response, For electrothermal kernel function, This is a temperature rise time series, and * represents the convolution operator. This represents the change in response after compensation. To suppress noise, the kernel length is set to no more than the sum of the thermal pulse duration and the cooling time, and zero-extension is used at the boundary. A slight moving average is applied to the results. Dimensional calibration is performed based on the steady-state gain of the electro-salt kernel function, mapping the compensated response to the conductivity dimension:

[0075]

[0076] in, For the electro-salt kernel function, For steady-state gain, The length of the integration window. For frequency domain observations, This is a Fourier transform. Its physical meaning is: the proportion of the electrical response caused by a unit change in conductivity of the water layer under steady state. ,by Normalizing the compensated frequency domain response yields an observation consistent with the dimensions of conductivity. If implemented in the time domain, it is possible to... As the equivalent conductivity variation curve, the discrete Fourier transform is then taken for each frequency window to obtain... .

[0077] To ensure stability, two quality control measures are implemented: one is to... The confidence interval determination is as follows: if the interval is too wide due to noise or input correlation, the weight of that channel is reduced; secondly, for... The amplitude-phase continuity test detects jumps and then reverts to a shorter integration window and recalculates. Weights and a quality mask are included when constructing the observation vector for each frequency; the weights are derived from the prior reciprocity test and the uncertainty assessment of the kernel function.

[0078] In terms of system architecture, the data processing module includes a compensation unit, a calibration unit, and an observation construction unit. The compensation unit reads the electrothermal kernel function and the temperature rise time series and performs convolution subtraction; the calibration unit calculates the steady-state gain based on the electrosalt kernel function and performs dimension normalization; the observation construction unit organizes the data into an equivalent conductivity observation matrix of the complex transfer response according to channel and frequency, and outputs a frequency list, weights, and a quality mask. For easy traceability, metadata such as kernel function version, integration window, sampling rate, and alignment delay are recorded.

[0079] The principle of this process is to use an electrothermal kernel function to eliminate the influence of temperature and water content on the electrical response within the same measurement window, and then use an electrosalt kernel function to provide dimensional anchors, transforming relative electrical quantities into conductivity-dimensional observations with engineering significance. The benefits are: consistent units between the observed measurements and the multi-frequency forward model; enhanced comparability across measurement ranges; reduced influence of slowly varying thermal fields in the low-frequency band; and more robust initial inversion values ​​and regularization weights. In actual deployment, only conventional signal processing and numerical integration are required, and parameter values ​​are adaptively adjusted according to the intensity of field noise and disturbances, facilitating long-term operation in the field.

[0080] Preferably, the assembly includes a forward model with a time-varying top boundary, which employs a quasi-static conductivity equation and Robin boundary conditions; the multi-frequency joint inversion includes constructing a multi-frequency data consistency term and a regularization term, the regularization term including anisotropic total variation regularization, structural sparsity regularization derived from temperature change time series, and salt conservation constraints.

[0081] This embodiment provides an feasible process for forward model assembly and multi-frequency joint inversion, highlighting the synergistic effect of time-varying top boundaries and three types of regularization to ensure consistency with field data and on-site conditions.

[0082] The geometric model and electrode coordinates are discretized into a finite element mesh with triangular or quadrilateral elements. The injected and received dipoles are mapped according to the mesh nodes. The top boundary adopts the Robin condition updated over time, with water conductivity and water depth as driving variables. The control end pushes the boundary coefficient sequence according to the time axis, while the side and bottom boundaries are taken as flux-free. The potential field is solved independently for each frequency, and the electrode response is cached to facilitate parallel computing and anomaly backtracking.

[0083] The core computation retains only the following expressions:

[0084]

[0085]

[0086] For the equivalent conductivity distribution, For frequency The potential field, For the normal derivative, The time-varying top boundary transducer. For the top boundary source term; For a set of frequencies, For assembling a single-frequency forward map with a time-varying top boundary. For observations consistent with the dimensions of electrical conductivity; For total anisotropy, the penalty along the main hydraulic direction is weaker. The structural field is extracted from the time series of temperature changes. For scale weighting; The salt conservation residual reflects the balance error of salt in transport and dispersion obtained by the equivalent conductivity mapping. This is the regularization weight.

[0087] The implementation steps are as follows: First, forward assembly: Load the mesh file, electrode table, injection receiver list, and top boundary coefficient sequence into the control software to complete the assembly of the single-frequency stiffness matrix and the construction of source terms; solve for the potential at each frequency and generate the predicted response. Second, data term construction: Read the compensated and dimensionally calibrated observations, align them by channel and frequency, set weights, and reduce the weight of low-quality channels. Third, canonical field generation: Form the structural field from the peak value or arrival time of the temperature rise, and perform scale normalization; estimate the main hydraulic direction from the canal direction or historical water level changes. Fourth, solution control: Initialize the equivalent conductivity distribution with a smooth baseline, adopt iterative optimization with non-negative projection, first converge the data terms with low weights, and then gradually increase the weights of anisotropic and structural terms; the salinity conservation term is added with soft constraints, only restricting non-physical fluctuations in the time series. Termination is based on the synchronous convergence of the data residuals and the target value, and uncertainty assessment is output.

[0088] At the system level, the model assembly unit is responsible for updating and verifying the mesh and boundaries; the inversion unit is responsible for the iteration strategy and weight adaptation; and the quality management unit evaluates the residual spectrum, boundary misalignment, and structural consistency after each iteration, triggering weight fine-tuning or channel masking. The output includes the equivalent conductivity distribution, pore water conductivity mapping, uncertainty grid, and complete configuration log, ensuring reproducibility and traceability.

[0089] The principle of this process is to constrain the electrical distribution with multi-frequency information under the drive of real boundaries, and to robustly modify the spatial morphology and temporal evolution of the solution by using structural information derived from temperature and salinity balance. The results are reflected in a clearer interface, better cross-frequency consistency, and a time-varying variation that better matches the irrigation, drainage, and salt return processes, facilitating integration with subsequent zonal regulation and long-term operation.

[0090] Preferably, the multi-frequency joint inversion is solved using an iterative optimization method, the data consistency term uses a robust loss function to suppress abnormal channels, and a non-negative constraint is applied to the equivalent conductivity distribution in the root region.

[0091] This embodiment presents the solution process for multi-frequency joint inversion, highlighting the engineering implementation of iterative optimization, robust loss suppression of abnormal channels, and non-negative constraints, and outputs the equivalent conductivity distribution in the root region.

[0092] The input consists of a set of frequencies, a forward operator, observations consistent with the conductivity dimension, channel weights, and a regularization term from the previous step. The initial equivalent conductivity distribution is a baseline smooth field, and the frequency is determined using a phased strategy from mid-frequency to full-frequency to avoid interference from low signal-to-noise ratio bands in early iterations.

[0093] Data consistency is achieved using Huber robust loss, the core expression of which is as follows: , For the first The next iteration in frequency The weighted residual vector on; For frequency The channel weight matrix; In order to be in The predicted observation vector obtained by forward mapping at the location; For the first The equivalent conductivity distribution of the next iteration; For frequency The observation vector after compensation and dimensional calibration; For a single frequency element in a preset frequency set.

[0094]

[0095]

[0096] Huber loss function; The effect function of Huber loss; To be Limit to range Operators; For residual scalars or components; This is the Huber threshold parameter.

[0097]

[0098] For the first The total gradient direction of the next iteration For a preset frequency set; For frequency The sensitivity matrix in transpose of the location; The vector of weighted residuals after influence function transformation; In order to be in The gradient of the composition regularization term; It is a weighted sum of anisotropic total variation regularization, structural sparsity regularization, and salt conservation soft constraint.

[0099]

[0100] The updated equivalent conductivity distribution; For unit-by-unit nonnegative projection operators; For the first The step size of the next iteration; For the first The overall gradient direction of the next iteration; This represents the equivalent conductivity distribution before this update. The step size is determined through a backtracking search; the update stops when both the relative increment and the residual decrease are below the threshold, or when the maximum number of iterations is reached.

[0101]

[0102] The relative parameter increment threshold is used to determine whether the equivalent conductivity distribution has basically stabilized between two adjacent iterations; The normalized robust residual threshold is used to determine whether the residual based on robust loss has been reduced to the expected level. For the first The residual vector of the influence function of each iteration; The residual vector of the influence function at initialization; This is the equivalent conductivity distribution after this round of updates; This represents the equivalent conductivity distribution before this update.

[0103] Anomaly channel suppression follows a two-stage strategy: First, large residuals are truncated using the Huber function within the residual domain. Second, the number of times "low-impact weights" are accumulated along the channel dimension; if a threshold is exceeded, the channel is either assigned a low weight or temporarily blocked, and then retested in the next batch. To avoid local ill-posedness, block coordinate updates are performed by frequency band grouping: first, the regularization term weights are fixed and solved for several rounds in the mid-frequency domain, then low-frequency and high-frequency domains are gradually incorporated, with the step size and threshold re-estimated each time they are incorporated. .

[0104] The key implementation points are as follows: First, the sensitivity matrix returns a vector Jacobian product via a "forward first, adjoint first" interface, eliminating the need for explicit matrix assembly and ensuring memory efficiency. Second, non-negativity constraints employ element-wise projection to ensure the physical feasibility of the equivalent conductivity distribution. Third, regularization weights are adaptive with iteration: when boundary misalignment and excessively smoothed residual spectra or stripe artifacts occur, the weights of anisotropic and structural terms are fine-tuned. Fourth, the computation is organized in batches: each batch contains a fixed subset of frequencies and channels, and batches inherit the final values ​​from the previous batch, improving convergence stability.

[0105] The output includes the equivalent conductivity distribution, residual statistics, and uncertainty approximation (diagonal Fisher approximation), which are archived together with the channel masking log and parameter files for easy reproduction and comparison. This process uses a small number of core expressions as its framework, achieving robust and constrained multi-frequency inversion through feasible line search, projection, and adjoint differentiation. It can converge stably in field environments with abnormal channels and noise disturbances.

[0106] Preferably, the early salt return index is calculated based on the non-negative part of the time derivative of the equivalent conductivity distribution in the root region, the connectivity index is calculated based on the ratio of the conductivity gradient to the conductivity gradient norm along the measurement profile direction, and the hotspot mask is generated based on the confidence threshold and connected component discrimination.

[0107] This embodiment provides a process for generating the early salt return index, connectivity index, and hotspot mask, oriented towards long-term field operation and batch processing. The inputs are the root zone equivalent conductivity distribution of two adjacent measurement ranges, the corresponding confidence grid, and the measurement profile direction.

[0108] First, geometric registration and pixel-level alignment are performed on adjacent measurement maps in the same coordinate system. Weighted cropping is applied to boundary regions to ensure spatial consistency. To suppress occasional noise, a slight temporal smoothing of the equivalent conductivity distribution is performed only within a short time window of a single pixel to avoid weakening the true changes.

[0109] The early salt return index is calculated based on the non-negative part of the time derivative of the equivalent conductivity. The core expression is:

[0110]

[0111] in, For position With time Equivalent conductivity at the point; The time for two measurements; The time derivative; This is an early salt return index. It is primarily implemented using bilateral differential or adjacent measurement range differentials. If the measurement range intervals are unequal, it is normalized according to the actual time difference.

[0112] The connectivity index is used to reflect the trend of salinity channels along the measurement profile. The core expression is:

[0113] in, The spatial gradient of equivalent conductivity; This is a unit vector along the direction of the measurement profile; To prevent small positive numbers with a denominator of zero; The connectivity index is used. The gradient employs central difference, with one-sided difference at the boundaries, and the direction vector... It is given directly from the cross-sectional design.

[0114] Hotspot masks are generated under confidence constraints. First, a mask is constructed using a confidence raster, retaining pixels larger than a threshold to obtain the confidence mask. Then, within this mask, a joint threshold segmentation of the early salt return index and connectivity index is performed to form the initial hotspot raster. Subsequently, morphological opening and closing operations are used to remove scattered noise, patches with areas smaller than the threshold are discarded, and connected components are labeled. The boundary vector and statistics (area, mean, quantiles) of each connected component are output. The threshold and structuring element size are used as extrinsic parameters, calibrated once using historical field data and crop salt tolerance levels, and then used long-term.

[0115] To ensure stable results, this embodiment sets up quality control in three stages: first, time difference checks and abnormal interval alarms are performed before the time derivative is calculated; second, the gradient norm histogram is checked after the gradient is calculated, and if an overall offset occurs, the process is backtracked to the registration stage for verification; and third, the consistency between the slenderness ratio of the patches and the profile direction is verified after the connected components are marked to eliminate false hotspots caused by boundary mismatch.

[0116] The output includes three types of products: pixel-by-pixel early salinity index raster, connectivity index raster, and hotspot mask; patch-by-patch connected domain vectors and attribute tables; and grading suggestion fields for subsequent regulation. The attribute tables record the source measurement time, threshold parameters, minimum and average confidence levels, ensuring traceability. The above process relies only on common raster operations, differencing, and morphological operations. The software is implemented modularly: the temporal processing unit, spatial gradient unit, and connected domain analysis unit are executed sequentially, and parameters are managed through configuration files, facilitating replication and batch execution across different fields. This method, without adding additional sensing hardware, transforms electrical changes into actionable risk and channel indicators, providing direct spatial objects and priorities for subsequent intermittent leaching and drainage linkage.

[0117] Preferably, the generation of the control plan includes forming an intervention level map based on the early salt return index, connectivity index, and hotspot mask, and generating intermittent rinsing parameters, water source mixing parameters, and drainage linkage parameters; updating the next round of channel and frequency plan after retesting and inversion includes selecting the injection channel and frequency set based on the criterion of maximizing the information content of the sensitivity matrix.

[0118] This embodiment revolves around "control plan generation - retesting closed loop - channel and frequency update". The inputs are the root region equivalent conductivity distribution, early salt return index, connectivity index, hotspot mask, and a list of available water sources and drainage facilities. The outputs are the intervention level map, intermittent rinsing parameters, water source mixing parameters, and drainage linkage parameters. After retesting and rapid inversion, the injection channels and frequency set for the next round of measurement are updated.

[0119] First, an intervention level map is generated. An intervention score is calculated for each pixel and aggregated into connected components within a hotspot mask to obtain the classification results and the work block. The core expression is:

[0120] in, For intervention scoring, Equivalent conductivity For the threshold, Non-load disconnection, This is an early indicator of salt return. It is the connectivity index. Weights. Based on the connected components. The quantile setting level is used to output vector work blocks. Intermittent rinsing parameters are generated for each work block. A target conductivity threshold and allowable decrease are set, and the single-pulse water depth is estimated based on the intensity and growth trend within the block. ,in, For the depth of a single pulse, The mean value within the block. The mean value within the block. This is a planning coefficient. The number of pulses and the interval are given based on the field infiltration rate and the operation period, forming a "water volume-interval-sequence" list. The water mixing parameters are calculated based on the set target conductivity of the influent to determine the proportion of clean water.

[0121] in, This represents the volume fraction of pure water. The target conductivity of the influent is... and The conductivity values ​​are for clean water and slightly brackish water, respectively. Write the matching instructions for the pump station and valves.

[0122] The drainage linkage parameters combine connectivity and empirical time delays to set the activation timing and duration. The salt discharge delay is estimated using the response time delay of the previous salinity step, defined as follows: ,in, To delay the start of drainage, This represents the peak time lag of the electrical response to the step change in water layer conductivity during the previous measurement cycle. Longer opening durations and lower closing thresholds are set for high-connectivity blocks to create a gate and tributary scheduling table.

[0123] After execution, retesting and rapid inversion are performed within a short time window to verify the decrease in equivalent conductivity and early salt return index changes of the current work block, forming block-level performance indicators and uncertainties. The results are written back to the data warehouse as the basis for updating the measurement design.

[0124] The next round of channel and frequency planning will adopt the criterion of maximizing information content. A sensitivity matrix will be constructed based on the current solution, and a greedy selection process will be performed on candidate injection-reception-frequency combinations. ,in, For the sensitivity matrix, For the candidate channel set, For the candidate frequency set, Let the matrix be a determinant. The algorithm starts with an empty set and successively adds elements to make the matrix determinant... The combination with the highest gain is selected until the range and energy limits are reached; combinations marked as low quality by reciprocity or quality masking are not included in the candidates. The updated injection and reception lists, frequency list, and execution timing are output.

[0125] The principle behind the above process is to transform diagnostic data into direct parameters for water allocation, timing, and drainage, and to make the next round of measurements more sensitive to key areas through information criteria. The effects are reflected in prioritizing hotspot treatment, providing evidence-based water resource utilization, coordinating drainage and infiltration, and continuously self-optimizing the measurement process, thereby obtaining a more reliable spatial-temporal field of conductivity and a more controllable risk of salinization with the same investment.

[0126] like Figure 2As shown, an in-situ conductivity measurement system for saline-alkali paddy fields is used to implement the aforementioned in-situ conductivity measurement method for saline-alkali paddy fields. The system includes: The baseline deployment module is used to acquire topographic and irrigation / drainage information, deploy measurement profiles, establish electrode arrays and water layer conductivity measurement devices, generate injection and receiving channel sets, acquire water layer conductivity and water depth baselines, and form geometric models and boundary baseline data. It then provides the geometric model, boundary baseline data, and injection and receiving channel sets to the controlled measurement module. The baseline deployment module consists of a differential satellite positioning receiver or RTK handheld device, a portable topographic surveyor, a ruler, and positioning stakes, forming a field positioning unit. The electrode array uses corrosion-resistant metal electrodes, shielded cables, and terminals. A multiplexed switch matrix is ​​housed in a waterproof enclosure. The water layer conductivity measurement device includes conductivity and temperature probes, and the water depth sensor is either pressure-type or ultrasonic. The data logger collects GNSS, sensor, and electrode coordinates, outputs gigabit Ethernet or RS-485, is powered by a lithium battery and solar panel, and generates and distributes channel lists, geometric models, and boundary baseline data.

[0127] The controlled measurement module applies current excitation to the injection channel at a preset frequency set and synchronously detects to obtain the complex transfer response matrix. It implements thermal pulses and salinity steps within the same measurement range. Based on the complex transfer response and changes in water layer conductivity, it calculates the electro-salt kernel function and electro-thermal kernel function, and provides the complex transfer response matrix, electro-salt kernel function, electro-thermal kernel function, and water layer conductivity changes to the inversion diagnostic module. The controlled measurement module generates parallel multi-sinusoidal excitations using a programmable current source and function generator, and selects the injection channel via a multiplexed switch matrix. The receiving end employs a synchronous sampling multi-channel analog-to-digital converter, pre-band-limited filtering and isolation amplification, and phase-locked detection implemented in a DSP / FPGA or embedded industrial computer, outputting the complex transfer response matrix. Thermal pulses are provided by an insulating heating belt and a power controller, while salinity steps are achieved by a metering pump and a mixer. The entire link uses a unified GPS / BeiDou PPS clock, is equipped with leakage protection and overcurrent limiting, has an IP65 enclosure, and features gigabit Ethernet backhaul.

[0128] The inversion diagnostic module is used to compensate and dimensionally calibrate the complex transfer response based on the electro-salt kernel function and the electro-thermal kernel function. It assembles a forward model including a time-varying top boundary and performs multi-frequency joint inversion to obtain the root region equivalent conductivity distribution and pore water conductivity distribution. It calculates the time derivative and generates the early salt return index, connectivity index, and hotspot mask, and provides the results to the control closed-loop module. The inversion diagnostic module uses an industrial computer or weather-resistant edge server, configured with a multi-core CPU, sufficient memory, and SSD array, and optionally equipped with GPU to accelerate finite element solution. The local network includes a NAS for archiving raw and intermediate results. The rack has a built-in UPS and temperature-controlled fan, supporting 4G / 5G and Ethernet redundancy. The interface receives data streams and baseline libraries from the control and measurement module. A field monitor and keyboard and mouse are used for debugging. It runs compensation and dimensional calibration, forward assembly, and multi-frequency joint inversion, and generates equivalent and pore water conductivity distributions and diagnostic indices in real time.

[0129] The control closed-loop module is used to generate a control plan based on the root region equivalent conductivity distribution, early salt return index, connectivity index, and hotspot mask. It organizes retesting and inversion to form updated results, and updates the next round of channel and frequency plans accordingly, feeding them back to the controlled measurement module and the inversion diagnosis module. The control closed-loop module uses a PLC or RTU as the core control unit to drive the variable frequency pump, electromechanical valves, and electric gates, and collects online conductivity, flow, and water level sensors. The control cabinet includes relays, circuit breakers, and EMC filters, and supports Modbus-TCP / RTU and Ethernet switching. The upper-level SCADA / edge gateway issues intermittent rinsing parameters, water source mixing ratio, and drainage linkage timing, and receives retest results. It reserves manual bypass and emergency stop, buzzer and SMS / WeChat alarms, and writes closed-loop parameters back to the controlled measurement and inversion diagnosis modules.

[0130] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0131] The above are merely embodiments of this application and are not intended to limit the scope 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 the claims of this application.

Claims

1. A method for in-situ determination of electrical conductivity in saline-alkali paddy fields, characterized in that, Includes the following steps: Acquire topographic and irrigation / drainage information, set up measurement profiles, establish electrode arrays and water layer conductivity measurement devices, generate injection and receiving channel sets, acquire water layer conductivity and water depth baselines, and form geometric models and boundary baseline data; Acquiring topographic and irrigation / drainage information includes determining the direction and location of the measurement profile based on topographic data; establishing an electrode array and aquatic conductivity measurement device includes connecting electrodes through a multiplexed switch matrix to generate a set of injection and receiving channels. Acquiring water layer conductivity and water depth baselines and forming geometric models and boundary baseline data includes recording electrode spatial coordinates and zero-point offset of the water layer conductivity measuring device; A complex transfer response matrix is ​​obtained by applying current excitation to the injection channel at a preset frequency set and simultaneously detecting it. Thermal pulses and salinity steps were applied in the same measurement range, and the electro-salt kernel function and electro-thermal kernel function were calculated based on the response and changes in water layer conductivity. Applying current excitation to the injection channel at a preset frequency set includes using a parallel multi-sine coding sequence; synchronous detection includes using phase-locked signal processing to extract amplitude and phase; performing a reciprocity check before generating the complex transfer response matrix to eliminate unqualified channels; Implementing thermal pulses and salinity steps within the same measurement range includes sequentially applying thermal pulses and salinity steps; calculating the electro-salt kernel function and the electro-thermal kernel function includes dual-input deconvolution with the water layer conductivity change time series and temperature rise time series as inputs and the complex transfer response change time series as output, and applying causal constraints and non-negativity constraints to the dual-input deconvolution; Compensation and dimensional calibration of the complex transfer response are performed based on the electro-salt kernel function and the electro-thermal kernel function. A forward model with a time-varying top boundary is assembled and multi-frequency joint inversion is performed to obtain the root region equivalent conductivity and pore water conductivity distribution. The time derivative is calculated and the early salt return index, connectivity index and hotspot mask are generated. A control plan is generated, and the next round of channel and frequency plans are updated after retesting and inversion. The assembly includes a forward model with a time-varying top boundary, which employs a quasi-static conductivity equation and Robin boundary conditions. The multi-frequency joint inversion includes the construction of a multi-frequency data consistency term and a regularization term. The regularization term includes anisotropic total variation regularization, structural sparsity regularization derived from temperature change time series, and salt conservation constraints. The early salt return index is calculated based on the non-negative part of the time derivative of the equivalent conductivity distribution in the root region, the connectivity index is calculated based on the ratio of the conductivity gradient to the conductivity gradient norm along the measurement profile direction, and the hotspot mask is generated based on the confidence threshold and connected component discrimination. The generation of the control plan includes forming an intervention level map based on the early salt return index, connectivity index, and hotspot mask, and generating intermittent rinsing parameters, water source mixing parameters, and drainage linkage parameters; after retesting and inversion, the next round of channel and frequency plan is updated, including selecting the injection channel and frequency set based on the criterion of maximizing the information content of the sensitivity matrix.

2. The method according to claim 1, characterized in that, The complex transfer response is compensated for by temperature and water content based on the electro-thermal kernel function, and dimensionally calibrated based on the electro-salt kernel function. The compensated and dimensionally calibrated complex transfer response is then converted into an observation consistent with the conductivity dimension for multi-frequency joint inversion.

3. The method according to claim 1, characterized in that, The multi-frequency joint inversion is solved using an iterative optimization method. The data consistency term uses a robust loss function to suppress abnormal channels, and a non-negative constraint is applied to the equivalent conductivity distribution in the root region.

4. A system for in-situ measurement of electrical conductivity in saline-alkali paddy fields, used to implement the method for in-situ measurement of electrical conductivity in saline-alkali paddy fields according to any one of claims 1 to 3, characterized in that, The system includes: The baseline module is used to acquire topographic and irrigation / drainage information, set up measurement profiles, establish electrode arrays and water layer conductivity measurement devices, generate injection and receiving channel sets, acquire water layer conductivity and water depth baselines, and form geometric models and boundary baseline data. The geometric models, boundary baseline data, and injection and receiving channel sets are then provided to the controlled measurement module. The controlled measurement module is used to apply current excitation to the injection channel at a preset frequency set and simultaneously detect to obtain the complex transfer response matrix. It implements thermal pulse and salinity step in the same measurement range, calculates the electro-salt kernel function and electro-thermal kernel function based on the complex transfer response and the change in water layer conductivity, and provides the complex transfer response matrix, electro-salt kernel function, electro-thermal kernel function and the change in water layer conductivity to the inversion diagnostic module. The inversion diagnostic module is used to compensate and dimensionally calibrate the complex transfer response based on the electro-salt kernel function and the electro-thermal kernel function. It assembles a forward model containing a time-varying top boundary and performs multi-frequency joint inversion to obtain the root zone equivalent conductivity distribution and pore water conductivity distribution. It calculates the time derivative and generates the early salt return index, connectivity index and hot spot mask, and provides the above results to the control closed-loop module. The control closed-loop module is used to generate a control plan based on the root region equivalent conductivity distribution, early salt return index, connectivity index and hotspot mask, organize retesting and inversion to form updated results, and update the next round of channel and frequency plans accordingly and feed them back to the controlled measurement module and inversion diagnosis module.

Images