Pasturing area intelligent water supply system based on wind-solar complementation

Abstract

The invention provides a pasturing area intelligent water supply system based on wind and light complementation, and relates to the field of water supply management, and the system comprises a wind and light complementation power generation module which comprises a wind power generation unit, a photovoltaic power generation unit and a wind and light complementation control unit, and the wind and light complementation control unit is used for controlling the wind power generation unit and the photovoltaic power generation unit to carry out wind and light complementation power generation; the drinking water monitoring module is used for a drinking water trough monitoring unit and a visual monitoring unit, the drinking water trough monitoring unit is used for collecting drinking water trough state information, and the visual monitoring unit is used for collecting drinking water trough images; and the water supply regulation and control module is used for determining livestock drinking information according to the drinking trough state information and the drinking trough image, determining optimal water supply parameters according to the livestock drinking information, and supplying water to the drinking trough according to the optimal water supply parameters.

Description

Technical Field

[0001] This invention relates to the field of water supply management, and in particular to a smart water supply system for pastoral areas based on wind-solar hybridization. Background Technology

[0002] The complex and harsh natural environment of pastoral areas, coupled with the dispersed settlements of herders, presents numerous challenges to water supply. Water supply in pastoral areas is characterized by high altitude, water scarcity, frost heave, long distances, and decentralized distribution. Water sources primarily rely on surface water, which is heavily polluted, and drinking water safety remains a persistent issue. Pasture water supply is mainly decentralized, with wells typically ranging from 110 to 300 mm in diameter and 25 to 225 meters in depth. Water is pumped using small-diameter pumps, and in most areas, internal combustion engines are used as the power source. However, the use of internal combustion engines in pastoral areas has many limitations: fuel transportation and storage are difficult, and the pollution from combustion adversely affects the grassland environment. Solar and wind energy, as clean and renewable energy sources, represent an effective exploration of new energy technology development and a viable pathway to promoting sustainable agricultural development.

[0003] In existing technologies, herders need to inspect water sources (such as wells and reservoirs) daily and manually turn water pumps on and off, which is time-consuming and labor-intensive (each inspection takes 2-3 hours). The lack of real-time monitoring often leads to pump damage or water waste due to idling or over-pumping (e.g., traditional systems have a water utilization rate of only 60%-70%).

[0004] Therefore, there is a need to provide a smart water supply system for pastoral areas based on wind and solar integration to improve the intelligence level of water supply in pastoral areas and reduce water waste. Summary of the Invention

[0005] This invention provides a smart water supply system for pastoral areas based on wind-solar hybrid power generation, comprising: a wind-solar hybrid power generation module, including a wind power generation unit, a photovoltaic power generation unit, and a wind-solar hybrid control unit, wherein the wind-solar hybrid control unit is used to control the wind power generation unit and the photovoltaic power generation unit to generate electricity in a wind-solar hybrid manner; a drinking water monitoring module, comprising a drinking trough monitoring unit and a visual monitoring unit, wherein the drinking trough monitoring unit is used to collect drinking trough status information, and the visual monitoring unit is used to collect drinking trough images; and a water supply control module, used to determine livestock drinking water information based on the drinking trough status information and drinking trough images, determine optimal water supply parameters based on the livestock drinking water information, and supply water to the drinking troughs according to the optimal water supply parameters.

[0006] Furthermore, the drinking trough monitoring unit includes a ring-shaped pressure sensor array, wherein the ring-shaped pressure sensor array includes multiple pressure sensors distributed in a ring, the spacing between any two adjacent pressure sensors is consistent, and the spacing between two adjacent pressure sensors is determined based on the sampling period.

[0007] Furthermore, the water supply control module determines livestock drinking information based on the drinking trough status information and the drinking trough image, including: preprocessing the drinking trough status information; identifying whether a livestock drinking event has occurred based on the preprocessed drinking trough status information; and when a livestock drinking event is identified, determining the livestock drinking behavior based on the drinking trough image, wherein the livestock drinking information includes at least the livestock drinking behavior.

[0008] Furthermore, the drinking trough status information includes pressure signals from multiple consecutive time points collected by the ring pressure sensor array; the water supply control module preprocesses the drinking trough status information, including: reconstructing the wave source coordinates based on the pressure signals from multiple consecutive time points collected by the ring pressure sensor array; and performing convolution enhancement processing on the pressure signals from multiple consecutive time points collected by the ring pressure sensor array based on the reconstructed wave source coordinates.

[0009] Furthermore, the water supply control module identifies whether a livestock drinking event has occurred based on the pre-processed drinking trough status information, including: determining a first pressure fluctuation feature based on pressure signals from multiple consecutive time points collected by the convolutionally enhanced ring pressure sensor array, wherein the first pressure fluctuation feature includes at least the fluctuation amplitude and frequency peak; determining whether to identify a livestock drinking event based on the first pressure fluctuation feature and a first condition set; if it is determined that a livestock drinking event should be identified, determining a second pressure fluctuation feature based on pressure signals from multiple consecutive time points collected by the convolutionally enhanced ring pressure sensor array, wherein the second pressure fluctuation feature includes at least the instantaneous pressure change, frequency energy, wave propagation speed, and normalized cross-correlation value of the spatial wavefront morphology; and identifying whether a livestock drinking event has occurred based on the second pressure fluctuation feature and the second condition set.

[0010] Furthermore, the water supply control module determines livestock drinking behavior based on the water trough image, including: determining multiple key points and multiple behavioral auxiliary points, wherein the multiple key points include at least multiple head key points and multiple neck key points; segmenting the water trough image into multiple target region images; for each target region image, using a head and neck posture estimation model, identifying the head and neck posture of the livestock in the target region image based on the multiple key points, and determining whether the livestock exhibits drinking behavior based on the head and neck posture of the livestock.

[0011] Furthermore, the loss function used to train the head and neck pose estimation model includes at least: , , in, This is the loss value. These are the weighting coefficients. For the predicted joint angle, For the actual joint angle, The L2 norm of the predicted angle and the range of the true angle. These are the weighting coefficients. To predict the KL divergence between the distribution and the true distribution, For the predicted motion posture probability distribution, This represents the true probability distribution of motion postures.

[0012] Furthermore, the water supply control module determines the optimal water supply parameters based on livestock drinking water information, including: determining the optimal water supply parameters based on the number of livestock exhibiting drinking behavior and the second pressure fluctuation characteristics.

[0013] Furthermore, the optimal water supply parameters include at least the optimal pumping frequency and the optimal water supply per unit time.

[0014] Furthermore, the water supply control module supplies water to the drinking trough according to the optimal water supply parameters, including: controlling the water pump to supply water to the drinking trough through the PLC according to the optimal water supply parameters.

[0015] Compared with existing technologies, the intelligent water supply system for pastoral areas based on wind-solar hybridization provided by this invention has at least the following beneficial effects: Wind and solar power generation units work collaboratively through a wind-solar hybrid control unit, utilizing solar energy during the day and wind energy at night or on cloudy days to achieve all-weather energy supply. This reduces reliance on traditional power grids or diesel generators, lowers electricity costs in pastoral areas, and reduces carbon emissions, aligning with green and sustainable development principles. Pastoral areas are typically far from the power grid, and energy supply is often unstable. The wind-solar hybrid system solves the power supply problem in pastoral areas through distributed generation. It also enhances the autonomy of pastoral infrastructure and reduces the risk of water supply system outages due to energy shortages.

[0016] The water trough monitoring unit collects status information such as water level and quality, while the visual monitoring unit uses image recognition technology to monitor livestock drinking behavior (such as drinking frequency and livestock numbers). This accurately grasps livestock's water needs, avoiding over- or under-watering. The water supply control module dynamically adjusts water supply parameters (such as water volume and duration) based on the water trough status information and image data. This reduces water waste, which is especially important in arid pastoral areas. It also reduces the frequency of manual inspections, saving labor costs. Attached Figure Description

[0017] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein: Figure 1This is a schematic diagram of a smart water supply system for pastoral areas based on wind-solar hybrid power, as shown in some embodiments of this specification. Figure 2 This is a schematic diagram of the structure of a ring pressure sensor array according to some embodiments of this specification; Figure 3 This is a schematic diagram of the water supply control process according to some embodiments of this specification. Detailed Implementation

[0018] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0019] Figure 1 This is a schematic diagram of a module of a wind-solar hybrid intelligent water supply system for pastoral areas, as shown in some embodiments of this specification. Figure 1 As shown, a smart water supply system for pastoral areas based on wind-solar hybrid power generation can include a wind-solar hybrid power generation module, a drinking water monitoring module, and a water supply control module.

[0020] The wind-solar hybrid power generation module includes a wind power generation unit, a photovoltaic power generation unit, and a wind-solar hybrid control unit, wherein the wind-solar hybrid control unit is used to control the wind power generation unit and the photovoltaic power generation unit to perform wind-solar hybrid power generation.

[0021] Specifically, the core function of a wind power generation unit is to convert wind energy into electrical energy. The blades of a wind turbine rotate under the influence of wind, driving the generator rotor to rotate and converting mechanical energy into electrical energy using the principle of electromagnetic induction. The efficiency of the wind turbine in capturing wind energy is closely related to wind speed, rotor diameter, and blade design. When the wind speed reaches the cut-in wind speed (generally 3-4 m / s), the turbine begins to generate electricity; as the wind speed increases, the power generation increases; however, when the wind speed exceeds the rated wind speed (generally 12-16 m / s), the turbine maintains a stable power output through pitch adjustment or other methods; when the wind speed reaches the cut-out wind speed (generally around 25 m / s), the turbine automatically stops operating to protect the equipment. Wind power generation units typically use direct-drive permanent magnet synchronous generators to improve reliability, outputting three-phase alternating current that is proportional to the wind speed.

[0022] Photovoltaic power generation units utilize the photovoltaic effect of solar panels to directly convert sunlight into electrical energy. Solar panels are primarily made of semiconductor materials. When sunlight shines on the panel, photons excite electrons in the semiconductor, generating electron-hole pairs. Under the influence of the internal electric field of the cell, the electrons and holes move towards the two electrodes, forming an electric current and achieving photoelectric conversion. The photovoltaic power generation unit outputs electrical energy in the form of direct current and uses an intelligent management core to realize functions such as charging, discharging, and inversion.

[0023] The wind-solar hybrid control unit is a key component of the entire power generation system, connecting the wind turbine unit, photovoltaic unit, battery, and load. It improves energy efficiency through maximum power point tracking (MPPT) technology based on different environmental conditions and provides comprehensive control over the input and output of the entire power generation system. Specific functions include: Power regulation: When the photovoltaic power generation system has high power generation efficiency, the control unit will automatically reduce the power output of the wind power generation system; conversely, when the wind power generation efficiency is high, the power output of the photovoltaic power generation system will be reduced. By adjusting the power output of the two systems in a timely manner, optimal power generation efficiency can be achieved.

[0024] Energy Management: The controller continuously switches and adjusts the battery bank's operating status based on changes in solar radiation intensity, wind speed, and load. On one hand, the adjusted electrical energy is directly sent to the DC or AC load; on the other hand, excess electrical energy is stored in the battery bank. When the generated power cannot meet the load demand, the controller sends energy from the battery to the load, ensuring the continuity and stability of the entire system.

[0025] Mode switching: The wind-solar hybrid power generation system can operate in three modes based on changes in wind and solar radiation: the wind turbine generator supplies power to the load alone; the photovoltaic power generation system supplies power to the load alone; and the wind turbine generator and the photovoltaic power generation system supply power to the load together. The wind-solar hybrid control unit is responsible for intelligently switching between these modes.

[0026] Protection functions: Provides overcharge and over-discharge protection for the battery, prevents excessive wind turbine speed, and ensures safe operation of the system.

[0027] As an example only, the photovoltaic power generation unit consists of 6 photovoltaic panels, and the wind power generation unit consists of a wind turbine. They are connected to a wind-solar hybrid control unit to ensure stable power output and meet the subsequent needs of supplying water to the drinking trough according to the optimal water supply parameters.

[0028] A smart water supply system for pastoral areas based on wind-solar hybridization may also include pumping and storage modules, including: DC water pump: Used to pump water from the water source to the reservoir. Pumping starts when the water level in the reservoir is below 15cm and stops when it is above 130cm.

[0029] AC water pump: Used to pump water from the reservoir to the drinking water tank. The pumping frequency is controlled by a frequency converter. The parameters of the frequency converter (such as current frequency value, frequency adjustment range, voltage value, current value, current speed, and current power value) are transmitted to the PLC controller in real time.

[0030] Water storage tank: An underground water storage tank with dimensions of 150cm high, 200cm long and 100cm wide is set up to store the pumped water.

[0031] A drinking water monitoring module is used for a drinking water trough monitoring unit and a visual monitoring unit, wherein the drinking water trough monitoring unit is used to collect drinking water trough status information, and the visual monitoring unit is used to collect drinking water trough images.

[0032] In some embodiments, the drinking trough monitoring unit includes a ring-shaped pressure sensor array, wherein the ring-shaped pressure sensor array includes a plurality of pressure sensors arranged in a ring, the spacing between any two adjacent pressure sensors is consistent, and the spacing between two adjacent pressure sensors is determined based on the sampling period.

[0033] Specifically, the spacing between two adjacent pressure sensors needs to meet the following constraints: , in, The distance between two adjacent pressure sensors. The sampling period is 50ms by default. The velocity of sound in water is c≈0.45m / s in clear water at 15℃, which is corrected to 0.3-0.4m / s for real-world scenarios. The equivalent theoretical maximum spacing is approximately 20mm.

[0034] Figure 2 This is a schematic diagram of the structure of a ring pressure sensor array according to some embodiments of this specification, such as... Figure 2 As shown, for illustrative purposes only, the annular pressure sensor array consists of eight pressure sensors evenly arranged in a ring array with a diameter of 80 cm and a spacing of 20 mm. The spacing of the pressure sensors meets the Nyquist-Shannon sampling criterion, ensuring accurate capture of the highest frequency of water wave propagation. The pressure sensor has a sensitivity of 0.1 Pa, corresponding to a water level change of 0.5 mm, and can detect flow velocity disturbances as low as 0.3 L / min caused by the tongue of cattle or sheep touching the water surface. The sensor employs a low-loss lithium niobate-on-insulator design, which greatly improves the Q value of the pressure sensor and ensures high-precision measurement.

[0035] Based on FPGA hardware triggering, all pressure sensors synchronously acquire data within 5μs, avoiding the time misalignment of traditional polling sampling and eliminating the calculation error introduced by the phase difference of wave propagation.

[0036] Compared to existing linear pressure sensor arrays, the ring pressure sensor array offers multi-dimensional performance improvements, as shown in Table 1.

[0037]

[0038] The annular pressure sensor array also features: Anti-clogging design: Sensor cavity self-cleaning mechanism (triggers a 0.2MPa gas pulse after each sampling); Thermodynamic compensation: Embedded Pt1000 temperature sensor corrects for water temperature and density changes (accuracy maintained at ±0.8mm across the entire temperature range from -20℃ to 50℃); Corrosion-resistant structure: SUS316L stainless steel shell + nano hydrophobic coating (contact angle > 150°). Health assessment: A self-diagnostic pulse test is performed every 5 minutes to assess sensor sensitivity drift through the impulse response function.

[0039] The water supply control module is used to determine livestock drinking water information based on the status information and image of the drinking trough, determine the optimal water supply parameters based on the livestock drinking water information, and supply water to the drinking trough according to the optimal water supply parameters.

[0040] In some embodiments, the water supply control module determines livestock drinking water information based on the drinking trough status information and the drinking trough image, including: The status information of the drinking trough is preprocessed, which includes pressure signals at multiple consecutive time points collected by the ring pressure sensor array. Based on the pre-processed status information of the water trough, identify whether a livestock drinking event has occurred; When a livestock drinking event is identified, the livestock drinking behavior is determined based on the image of the drinking trough, wherein the livestock drinking information includes at least the livestock drinking behavior.

[0041] In some embodiments, the water supply control module preprocesses the drinking trough status information, including: The wave source coordinates are reconstructed based on the pressure signals collected from multiple consecutive time points by the ring pressure sensor array. Based on the reconstructed wave source coordinates, the pressure signals from multiple consecutive time points acquired by the annular pressure sensor array are subjected to convolution enhancement processing.

[0042] Specifically, the wave source coordinates are reconstructed by the peak amplitude difference and phase difference between adjacent pressure sensors. As the wave propagates in the water tank, it passes through each pressure sensor sequentially. Due to the different locations of the wave source, the arrival times of the waves at each sensor will differ. For two adjacent sensors (e.g., sensor...), the wave source coordinates are reconstructed by the peak amplitude difference and phase difference between adjacent sensors. i and sensors jThere must be a time difference between the detection of the same wave peak by the two sensors, reflecting the time interval it takes for the wave to travel from the first sensor to the second. By analyzing and processing the signals collected by the pressure sensors, signal processing algorithms (such as peak detection algorithms) can be used to accurately identify the time point when the same wave peak appears in the signals of the two sensors, and then calculate the time difference. Multiplying this time difference by the wave velocity yields the distance difference. During wave propagation, there are not only time differences but also phase changes. The phase of the same wave detected by adjacent sensors will also differ, and this phase difference is also related to the location of the wave source and the wave propagation path. Through signal wave travel matching technology, the signals collected by the two sensors can be analyzed in detail to find their phase relationship, thereby obtaining phase difference information. The phase difference can further help verify the accuracy of the time difference measurement, and in some complex wave propagation situations, the phase difference may contain wave source location information that the time difference cannot reflect, which helps improve the positioning accuracy. Correlation analysis can be used for signal wave travel matching. By performing cross-correlation on the signals from two sensors, the peak position of the cross-correlation function corresponds to the optimal matching point between the two signals. The phase difference can be calculated based on the relationship between the peak position and the signal period. Using the adjacent sensor positions as foci, a hyperbola equation is established. According to the rigorous mathematical definition of a hyperbola, in a plane, the locus of points whose distance difference to two fixed points (foci) is constant is a hyperbola. In this problem, the two adjacent sensors... i and j The positions can be considered as the two foci of a hyperbola, and the distance difference calculated earlier is a constant in the definition of a hyperbola. Therefore, the wave source must be located between the two sensors. i and j The location of the wave source lies on a hyperbola with its focus. To accurately determine the location of the wave source, using only the hyperbolic equation established by a pair of adjacent sensors is insufficient, as a hyperbola has infinitely many points, making it impossible to uniquely determine the wave source coordinates. Therefore, at least two pairs of adjacent sensors (i.e., two hyperbolas) are needed. By solving for the intersection of these two hyperbolic equations, the coordinates of the wave source can be obtained. To improve the accuracy and reliability of the positioning, data from multiple pairs of adjacent sensors are used to establish multiple hyperbolic equations, and then optimization algorithms such as the least squares method are used to solve for the optimal estimate of the wave source coordinates.

[0043] For example, convolution enhancement can be performed according to the following formula: , in, This represents the enhanced waveform signal strength at position (x, y) and time t. Let be the raw pressure signal from the i-th sensor at time t. The dynamic weight of the i-th pressure sensor at the current moment can be calculated based on the following formula: , in, For signal-to-noise ratio, For direction matching items, Assess the health of the sensors. Let be the signal-to-noise ratio of the i-th sensor (unitless, range 0~10). To determine the propagation direction (in degrees) of the target fluctuation event, spatial spectrum estimation is performed starting from the pressure signals of 8 channels. The MUSIC (Multiple Signal Classification) algorithm is used to calculate the spatial spectrum, obtaining the angle spectrum P(θ), which represents the signal power distribution at different angles θ. Peak detection is performed on the angle spectrum to find the angle with the highest signal power. Kalman smoothing is then applied to the detected peak angles, and the Kalman smoothed angles represent the propagation direction of the target fluctuation event. Set the orientation matching sensitivity parameter (default 30°). The sensor's health status (0~1, 0 indicates failure) is... The value range is 0~1, and the weights are normalized to ensure that , A convolution kernel with directional parameterization is used to match the azimuth angle θi of the pressure sensor to enhance the wave signal in a specific direction. It can be modeled based on wave propagation theory. According to the water wave attenuation characteristics in each direction, eight convolution kernels in each direction are generated in advance, corresponding to eight pressure sensors respectively.

[0044] Effectively suppresses background noise from wind and waves (signal-to-noise ratio improved by 15dB).

[0045] In some embodiments, the water supply control module identifies whether a livestock drinking event has occurred based on the pre-processed drinking trough status information, including: Based on the pressure signals collected from multiple consecutive time points by the ring pressure sensor array after convolution enhancement processing, the first pressure fluctuation feature is determined, wherein the first pressure fluctuation feature includes at least the fluctuation amplitude and the main frequency peak. Based on the first pressure fluctuation characteristics and the first condition set, it is determined whether to identify livestock drinking events. The first condition set may include: 1. The fluctuation amplitude of at least 3 sensors is >2kPa, to prevent false judgments due to weak interference; 2. The main frequency peak is in the range of 1-5Hz, matching the livestock drinking rhythm. If it is determined that livestock drinking event identification is to be carried out, the second pressure fluctuation feature is determined based on the pressure signals collected by the ring pressure sensor array after convolution enhancement processing at multiple consecutive time points. The second pressure fluctuation feature includes at least the instantaneous pressure change, frequency energy, wave propagation speed and the normalized cross-correlation value of the spatial wavefront morphology. Based on the second pressure fluctuation characteristics and the second condition set, the system identifies whether a livestock drinking event has occurred. The second condition set may include: 1. Instantaneous pressure surge > 2 kPa, excluding minor disturbances such as falling leaves / light rain; 2. Frequency energy concentrated in the 1-5 Hz range, matching the fundamental frequency of livestock movement; 3. Wave propagation speed 0.3-0.6 m / s, eliminating rapid wind waves or mechanical vibrations; 4. Normalized cross-correlation values ​​of the spatial wavefront morphology conform to the hoof contact model, verifying biomechanical characteristics. Preferably, when the instantaneous pressure surge > 10 kPa, it indicates a large number of livestock drinking or an abnormal event (such as livestock falling), and the system activates a high-frequency sampling mode (100 Hz → 500 Hz) and initiates an abnormal impact absorption algorithm. If an abnormal event is determined, the water inlet solenoid valve is immediately closed.

[0046] The characteristics of the second pressure fluctuation can be determined in the following ways: 1. Fluctuation amplitude and frequency peak The filtered signal (1-10Hz bandpass) is rectified by full wave, and the root mean square value within the sliding window (2 seconds) is taken to obtain the fluctuation amplitude.

[0047] Perform a Fast Fourier Transform on the windowed signal with 1024 points, and use a Hanning window to reduce spectral leakage, extracting the frequency point with the maximum energy, and taking a resolution of 1Hz.

[0048] 2. Instantaneous pressure change Calculation: Real-time calculation of first-order differences , in, for For at any time t The collected pressure signal value, For a moment t Move forward Δ t The pressure signal value collected over a period of time. For time intervals.

[0049] Threshold trigger: If 3 consecutive It was determined to be a valid mutation.

[0050] 3. Wave propagation speed Calculate the cross-correlation function between two sensor signals. The cross-correlation function measures the similarity between two signals at different time delays. By finding the time delay corresponding to the peak of the cross-correlation function, and based on the position vectors of the two sensors (i.e., the actual distance between the two sensors) and the time delay, calculate the propagation speed of the wave.

[0051] 4. Normalized cross-correlation value of space wavefront morphology An 8×8 pressure signal covariance matrix R is constructed. The covariance matrix describes the statistical relationship between various sensor signals, and signal feature extraction and classification can be performed by analyzing the covariance matrix. Eigenvalue decomposition is performed on the covariance matrix R to obtain principal components. When the proportion of principal components (PC1) is greater than or equal to 90%, it is determined to be a single-point wave source. This indicates that the main energy of the signal is concentrated in a few characteristic directions, which conforms to the propagation characteristics of a single-point wave source. A wavenumber spectrum is generated using the MUSIC algorithm. The generated wavenumber spectrum is matched with the hoof-shaped contact model using a normalized cross-correlation method. When the normalized cross-correlation value is greater than or equal to 0.85, the wavefront morphology is considered to conform to the elliptical attenuation mode of the hoof-shaped contact model, which helps in further analysis and identification of the wave source type and propagation characteristics. In some embodiments, the water supply control module determines livestock drinking behavior based on the drinking trough image, including: Identify multiple key points and multiple behavioral auxiliary points, wherein the multiple key points include at least multiple head key points and multiple neck key points; The water trough image is segmented into multiple target region images, where each target region image can be an image containing a single animal. For each target region image, the head and neck posture estimation model is used to identify the head and neck posture of livestock in the target region image based on multiple key points, and to determine whether the livestock are drinking water based on their head and neck posture.

[0052] As an example only, several key points can be shown in Table 2.

[0053]

[0054] Existing animal posture detection algorithms (such as the AP-10K dataset) typically define fewer than 15 general animal keypoints, focusing on the limbs and trunk, rarely accurately annotating the head and neck structures. Commercial products (such as CattleTech Pro) use thermal imaging behavior analysis, relying on overall body posture changes rather than anatomical keypoints. In contrast, this system provides a biomechanical basis for accurate modeling of drinking behavior in cattle and sheep by identifying the above multiple keypoints. The keypoints cover the temporomandibular joint-eye-ear-lip linkage area, specifically designed to identify the drinking action sequence of "head down-mouth open-tongue protrusion-swallowing". The keypoints are non-coplanarly distributed in three-dimensional space (such as symmetrical ear roots + asymmetrical cervical spinous processes), so even if one side is occluded, the posture angle can still be calculated from the remaining points, improving the error tolerance rate by 61%.

[0055] Multiple behavioral aids may include the tongue contact point (a virtual point, calculated by inversion through the contraction of the neck extensor muscles), the atlas rotation fulcrum (a three-dimensional center of motion), and the anterior thoracic reference point (supporting the cervical kinematic chain).

[0056] The head and neck pose estimation model replaced PAFPN with VoVNet, reducing the computational cost of keypoint branches by 32%. Based on YOLOv7-Pose, we have made significant improvements and core extensions tailored to the characteristics of pastoral areas to establish a head and neck pose estimation model: 1. Backbone network: Retains the ELAN structure of YOLOv7, but adds an Anatomical Feature Enhancement Module (AFEM) in the shallow layers (layers 2-4), and guides the model to focus on the animal head and neck region through a 1×1 compressed channel followed by channel attention (ECA-Net).

[0057] 2. Neck structure: The original PAFPN was replaced with VoV-GSCSP (Ghost convolution + spatial pyramid pooling), which reduced the number of parameters by 38% while improving the detection accuracy of key points of small targets (mAP↑12.3%).

[0058] 3. Pose Branch: Expanded to a dual-path design — Coordinate prediction path: Outputs the (x,y) coordinates of 23 key points (using the DSNT differentiable space numerical method). Biomechanical constraint path: Introducing vector angle constraint loss (Cosine Similarity Loss) to ensure that the connection of the head bones conforms to anatomical rules.

[0059] The training of the head and neck pose estimation model is improved as follows: Masking data augmentation: Randomly erase 40% of the head and neck area (mimicking sand / leaf occlusion), forcing the model to infer the complete shape through residual local features; 2. Dynamic sparse training: Convolutional kernels that contribute >80% to keypoint prediction are selected through the Lottery Ticket Hypothesis, and finally 58% of the network parameters are retained; 3. By adopting the post-training quantization (PTQ) strategy of TensorRT, the accuracy loss is only 0.8% (compared to FP32), and the inference speed is improved by 2.3 times; 4. Optimize the utilization of Tensor Cores on the NVIDIA Jetson platform (>92%) to achieve real-time processing of 30FPS (input resolution 640×640).

[0060] In some embodiments, the loss function used to train the head and neck pose estimation model is: , , , , , , , in, is the total loss, The loss function is the keypoint spacing constraint. The loss function is the vector direction consistency constraint. Let the loss function be the smoothing loss function for the motion trajectory of key points. The joint physiological activity limitation penalty loss function. As weight, Greater than 0, for This is the loss value. These are the weighting coefficients. For the predicted joint angle, For the actual joint angle, To determine the L2 norm of the predicted joint angles relative to the true anatomical angles, the model is forced to predict joint angles close to the anatomical true range. Predictions that violate biomechanics will be penalized; for example, in cattle, cervical scoliosis exceeding 25° results in a surge in injury. These are the weighting coefficients. To predict the KL divergence between the predicted and actual distributions, the overall distribution of the predicted postures is constrained to conform to the actual anatomical movement patterns, such as the statistical regularity of the coordinated changes in joint angles when cattle and sheep lower their heads to drink water, thus avoiding out-of-group erroneous postures. For the predicted motion posture probability distribution, This represents the true probability distribution of motion postures. The weighting is dynamic, adjusted based on the importance of key points (e.g., significantly increasing the weight of core joints such as the temporomandibular joint and the root of the ear). To predict the Euclidean distance between keypoints i and j (e.g., the distance between the tip of the nose and the midpoint of the lower lip). Let i be the anatomical ground truth distance between key points i and j in the CT database. This is a tolerance threshold, determined based on biokinematics (e.g., temporomandibular joint distance tolerance ±3mm). For the predicted vector direction, The true vector direction in the CT database. Let be the predicted location of the k-th keypoint at time t. Let the position of the k-th key point be predicted at time t-1. For dynamic weighting, highly mobile joints (such as the temporomandibular joint) are assigned high weights (γ=1.5), while stable joints (such as the frontal eminence) are assigned low weights (γ=0.2). k is the index of the rotational joint. For the predicted angle of the k-th rotational joint, This represents the maximum physiological angular limit for the k-th rotational joint in the CT database. The loss function penalizes cases where the predicted angle exceeds the physiological angular limit.

[0061] The head and neck pose estimation model shows improvements over existing technologies, as shown in Table 3.

[0062]

[0063] In some embodiments, the water supply control module determines the optimal water supply parameters based on livestock drinking water information, including: The optimal water supply parameters are determined based on the number of livestock exhibiting drinking behavior and the characteristics of second-pressure fluctuations.

[0064] Among them, the optimal water supply parameters include at least the optimal pumping frequency and the optimal water supply per unit time.

[0065] For example, if drinking behavior of a single animal is detected, the pumping frequency is adjusted to 30Hz and the water supply is 100 liters per hour.

[0066] For example, if multiple livestock are detected drinking water simultaneously, the pumping frequency is dynamically adjusted according to the number of livestock. As an example, if three livestock are detected, the pumping frequency is adjusted to 60Hz, and the water supply is 300 liters per hour.

[0067] The design has a maximum pumping frequency of 100Hz, corresponding to a maximum water supply of 500 liters per hour, which is suitable for peak periods or scenarios where multiple livestock drink water at the same time.

[0068] Specifically, by using pressure sensors and visual acquisition technology, it is possible to monitor and determine in real time how many cattle are showing signs of wanting to drink. For example, pressure sensors can detect pressure changes near the water trough caused by livestock approaching or touching it, while visual acquisition devices can identify the shape and movement of the livestock, thus accurately determining the number of animals needing to drink. Based on livestock science theory, an adult cow drinks an average of 12 liters of water at a time, while a young cow drinks an average of 6 liters. Based on the number of adult and young cattle showing signs of wanting to drink, the total amount of water needed for a single pumping operation can be calculated. The optimal pumping frequency is then determined by comprehensively considering the number of animals showing signs of wanting to drink and the characteristics of pressure fluctuations. f If a large number of livestock intend to drink, the pumping frequency may need to be increased to ensure sufficient water supply. Simultaneously, based on the characteristics of the second pressure fluctuation, such as large instantaneous pressure changes or abnormal fluctuation propagation speeds, the pumping frequency may need to be adjusted to stabilize the water pressure and flow within the tank. For example, if frequent pressure changes are detected within the tank, the pumping frequency may need to be appropriately reduced to avoid excessive pressure fluctuations due to rapid pumping. The default operating frequency is 45Hz, but this will be adjusted based on actual conditions. The pumping frequency is determined by the calculated total pumping volume and estimated pumping time. t (Through precise timing via a timer in the PLC), the optimal water supply per unit time can be calculated. Simultaneously, by combining secondary pressure fluctuation characteristics, such as frequency energy and spatial wavefront morphology, it can be ensured that the water supply per unit time meets the livestock's drinking needs while maintaining a stable and uniform water flow distribution within the trough. For example, if the frequency energy indicates a high water wave vibration frequency, it may mean a fast water flow; in this case, the water supply per unit time can be adjusted appropriately to avoid excessively high water levels or rapid water flow. For example, the pumping volume can be calculated using the following formula: , , Where Q is the pumping capacity in cubic meters (m³), P is the pump shaft power in kilowatts (kW) (known from the nameplate), η is the pump efficiency in percentage (e.g., 0.73 for 73%) (known from the nameplate), f is the operating frequency or load rate in Hz (default 45Hz), t is the operating time in hours (h), H is the head in meters (m), and 2.73 is the unit conversion constant. Z is the height difference between the inlet and outlet, and hw is the water input loss, including friction head loss and local head loss. One bend ≈ 0.4 meters of pump head loss, and 10 meters of horizontal distance ≈ 1 meter of pump head loss.

[0069] In some embodiments, the water supply control module supplies water to the drinking trough according to optimal water supply parameters, including: The PLC controls the water pump to supply water to the drinking water tank according to the optimal water supply parameters.

[0070] Furthermore, to conserve water resources, prevent freezing in winter, and ensure livestock drink enough water due to weather conditions, the circular water trough was improved. The improved circular water trough consists of a circular body, an outward-sloping structure, a return pipe channel, and an electronic switch. The circular body, as the core container for livestock drinking, is a closed ring, providing a surrounding drinking area for multiple animals to drink from different locations simultaneously. The outer wall of the circular body is designed with an outward slope. This sloped structure has multiple functions: firstly, it reduces splashing during drinking, keeping the surrounding environment dry and clean; secondly, when the water level is high or there are fluctuations in the flow, it helps guide the water to flow smoothly to the periphery, creating favorable conditions for return. At the bottom of the outer perimeter of the circular water trough, there is a recessed pipe channel, similar in size to a flexible hose. This return pipe channel is continuously distributed along the bottom perimeter of the circular water trough to collect water returning from the trough. Its recessed design effectively gathers the water flow, ensuring the return water flows smoothly into the reservoir. The bottom of the return pipe channel is flat and smooth to reduce water flow resistance and allow for smooth water flow. An electronic switch is installed at the bottom of the return pipe channel. This electronic switch has precise control functions, allowing for accurate control of the opening and closing of the return pipe according to system settings and actual needs. For example, ... Figure 3 As shown, to prevent the reservoir below from overflowing due to rain, the electronic switch is set to only be on for 5 minutes and automatically shut off after the specified time, thereby achieving effective management of the backflow process and ensuring the stable operation of the entire water supply system.

[0071] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A smart water supply system for pastoral areas based on wind-solar hybridization, characterized in that, include: A wind-solar hybrid power generation module includes a wind power generation unit, a photovoltaic power generation unit, and a wind-solar hybrid control unit, wherein the wind-solar hybrid control unit is used to control the wind power generation unit and the photovoltaic power generation unit to perform wind-solar hybrid power generation. A drinking water monitoring module is used for a drinking water trough monitoring unit and a visual monitoring unit, wherein the drinking water trough monitoring unit is used to collect drinking water trough status information, and the visual monitoring unit is used to collect drinking water trough images; The water supply control module is used to determine livestock drinking water information based on the status information and image of the drinking trough, determine the optimal water supply parameters based on the livestock drinking water information, and supply water to the drinking trough according to the optimal water supply parameters.

2. The intelligent water supply system for pastoral areas based on wind-solar hybridization according to claim 1, characterized in that, The water tank monitoring unit includes a ring pressure sensor array, wherein the ring pressure sensor array includes multiple pressure sensors distributed in a ring, and the spacing between any two adjacent pressure sensors is consistent, and the spacing between two adjacent pressure sensors is determined based on the sampling period.

3. A smart water supply system for pastoral areas based on wind-solar hybridization according to claim 2, characterized in that, The water supply control module determines livestock drinking water information based on the status information and images of the water troughs, including: Preprocess the status information of the drinking trough; Based on the pre-processed status information of the water trough, identify whether a livestock drinking event has occurred; When a livestock drinking event is identified, the livestock drinking behavior is determined based on the image of the drinking trough, wherein the livestock drinking information includes at least the livestock drinking behavior.

4. A smart water supply system for pastoral areas based on wind-solar hybridization according to claim 3, characterized in that, The status information of the drinking trough includes pressure signals at multiple consecutive time points collected by a ring pressure sensor array; The water supply control module preprocesses the status information of the drinking fountains, including: The wave source coordinates are reconstructed based on the pressure signals collected from multiple consecutive time points by the ring pressure sensor array. Based on the reconstructed wave source coordinates, the pressure signals from multiple consecutive time points acquired by the annular pressure sensor array are subjected to convolution enhancement processing.

5. A smart water supply system for pastoral areas based on wind-solar hybridization according to claim 4, characterized in that, The water supply control module identifies whether a livestock drinking event has occurred based on the pre-treated water trough status information, including: Based on the pressure signals collected from multiple consecutive time points by the ring pressure sensor array after convolution enhancement processing, the first pressure fluctuation feature is determined, wherein the first pressure fluctuation feature includes at least the fluctuation amplitude and the main frequency peak. Based on the first pressure fluctuation characteristics and the first condition set, determine whether to identify livestock drinking events; If it is determined that livestock drinking event identification is to be carried out, the second pressure fluctuation feature is determined based on the pressure signals collected by the ring pressure sensor array after convolution enhancement processing at multiple consecutive time points. The second pressure fluctuation feature includes at least the instantaneous pressure change, frequency energy, wave propagation speed and the normalized cross-correlation value of the spatial wavefront morphology. Based on the second pressure fluctuation characteristics and the second condition set, it is possible to identify whether a livestock drinking event has occurred.

6. A smart water supply system for pastoral areas based on wind-solar hybridization according to claim 3, characterized in that, The water supply control module determines livestock drinking behavior based on images of the water troughs, including: Identify multiple key points and multiple behavioral auxiliary points, wherein the multiple key points include at least multiple head key points and multiple neck key points; The water trough image is segmented into multiple target region images; For each target region image, the head and neck posture estimation model is used to identify the head and neck posture of livestock in the target region image based on multiple key points, and to determine whether the livestock are drinking water based on their head and neck posture.

7. A smart water supply system for pastoral areas based on wind-solar hybridization according to claim 6, characterized in that, The loss function used to train the head and neck pose estimation model includes at least the following: ， ， in, This is the loss value. is the weight coefficient, For the predicted joint angle, For the actual joint angle, The L2 norm of the predicted angle and the range of the true angle. is the weight coefficient, To predict the KL divergence between the distribution and the true distribution, For the predicted motion posture probability distribution, This represents the true probability distribution of motion postures.

8. A smart water supply system for pastoral areas based on wind-solar hybridization according to claim 5, characterized in that, The water supply control module determines the optimal water supply parameters based on livestock drinking water information, including: The optimal water supply parameters are determined based on the number of livestock exhibiting drinking behavior and the characteristics of second-pressure fluctuations.

9. A smart water supply system for pastoral areas based on wind-solar hybridization according to any one of claims 1-8, characterized in that, The optimal water supply parameters include at least the optimal pumping frequency and the optimal water supply per unit time.

10. A smart water supply system for pastoral areas based on wind-solar hybridization according to any one of claims 1-8, characterized in that, The water supply control module supplies water to the drinking trough according to the optimal water supply parameters, including: The PLC controls the water pump to supply water to the drinking water tank according to the optimal water supply parameters.

Images