Sewage treatment remote monitoring system and method based on Internet of Things

Abstract

The invention discloses a sewage treatment remote monitoring system and method based on the Internet of Things, and the system comprises a data collection module which is used for collecting the solar illumination direction, the solar radiation intensity, and the operation state data of sewage treatment equipment; the data transmission module is used for sending the data to a remote monitoring platform in real time in a wireless communication mode; the illumination modeling module is used for extracting direct and scattered illumination characteristics based on a distribution model; the power generation capacity prediction module is used for predicting photovoltaic power generation power by combining the illumination model and the radiation data; the energy consumption prediction module is used for predicting the energy consumption level based on the power generation data and the equipment state; the frequency optimization module is used for comparing the power generation condition with the energy consumption condition and generating operation frequency adjusting parameters; and the remote control module is used for issuing a control instruction to realize dynamic adjustment of the operation frequency of the sewage treatment system. Intelligent matching and regulation of power generation and energy consumption are achieved, and the energy utilization efficiency and the automation level in the sewage treatment process are improved.

Description

Technical Field

[0001] This invention relates to the field of wastewater treatment technology, and in particular to a wastewater treatment remote monitoring system and method based on the Internet of Things. Background Technology

[0002] With the rapid development of the Internet of Things and new energy technologies, solar power has been widely used in distributed scenarios such as wastewater treatment. To improve equipment energy efficiency and achieve dynamic control, the integration of solar energy with frequency conversion technology—solar frequency conversion control—has become an important direction in current research and engineering practice. This technology dynamically adjusts the equipment's operating state by matching the solar power generation capacity with the motor's operating frequency, ensuring both wastewater treatment efficiency and improved energy utilization efficiency.

[0003] In existing technologies, the application of solar-powered frequency conversion control in remote monitoring of wastewater treatment still suffers from the following shortcomings: Insufficient modeling capability for sunlight variations: Existing systems struggle to accurately model the direct and diffuse characteristics of sunlight, making it impossible to dynamically predict power generation capacity based on actual sunlight conditions, leading to delayed or unreasonable frequency adjustment responses. Inaccurate matching of power generation and energy consumption: Most solutions lack linkage modeling between photovoltaic power generation and the energy consumption of wastewater treatment equipment, failing to achieve accurate prediction and adaptive adjustment of operating frequency, resulting in low equipment operating efficiency or energy waste. Rigid frequency conversion control mechanism: In traditional systems, frequency conversion control relies heavily on static rule settings, unable to dynamically adjust the frequency based on changes in sunlight and system load, resulting in a lack of flexibility and intelligence in the control method. Low level of intelligence in remote dispatching: In existing technologies, remote platforms mostly only play a monitoring role, lacking the ability to generate and issue control commands based on data prediction results, leading to insufficient closed-loop control of the system.

[0004] Therefore, how to provide IoT-based remote monitoring systems and methods for wastewater treatment is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a remote monitoring system and method for wastewater treatment based on the Internet of Things (IoT). This invention fully utilizes technologies such as illumination direction modeling, photovoltaic power generation prediction, improved spherical regression energy consumption analysis, and adaptive frequency regulation. It describes in detail the entire process of intelligent prediction and remote control of the operating status of wastewater treatment equipment by collecting environmental and equipment data through the IoT and combining it with modeling algorithms. It has the advantages of strong energy consumption adaptability, high control accuracy, and high system intelligence.

[0006] The wastewater treatment remote monitoring system and method based on the Internet of Things according to embodiments of the present invention includes the following steps:

[0007] Data acquisition module: used to collect data on the direction of sunlight incidence, solar radiation intensity, and the operating status of wastewater treatment equipment, and to attach a unified time label to all types of data;

[0008] Data transmission module: used to transmit collected data to a remote monitoring platform in real time via narrowband IoT, cellular networks or industrial wireless communication.

[0009] Illumination Modeling Module: Used to perform directional modeling of solar illumination data based on the Von Mises–Fisher distribution, generating the concentration and directional parameters of direct and diffuse components;

[0010] Power generation capacity prediction module: used to combine the results of illumination modeling with solar radiation data to calculate the time-series predicted value of photovoltaic power generation capacity;

[0011] Energy consumption prediction module: Used to input power generation prediction values ​​and equipment operating status data, and output energy consumption prediction results and confidence intervals through an improved spherical regression algorithm;

[0012] Frequency optimization module: used to compare power generation and energy consumption data, calculate operating frequency parameters, and perform adaptive adjustments based on the predicted confidence interval;

[0013] Remote control module: Used to generate time-stamped operating frequency adjustment commands, which are sent to the equipment terminal through the communication interface to control the dynamic operation of the sewage treatment system.

[0014] This invention proposes an IoT-based remote monitoring system for wastewater treatment, constructing a closed-loop process from data acquisition, wireless transmission, illumination modeling, power generation prediction, energy consumption estimation to remote frequency control. The system models the direction of solar illumination using the Von Mises–Fisher distribution, extracting direct and scattered characteristics, and predicts photovoltaic power generation capacity using solar radiation data. An improved spherical regression algorithm is then used to predict energy consumption and confidence intervals. The system adaptively generates operating frequency adjustment parameters based on the difference between power generation and energy consumption, and sends control commands to terminal devices via the IoT to achieve dynamic equipment adjustment. This method offers advantages such as high precision, strong real-time performance, and intelligent energy consumption response.

[0015] Optionally, modules can be integrated using the following methods:

[0016] S1. Collect data on the incident direction of sunlight, solar radiation intensity, and the operating status of wastewater treatment equipment, add time tags to each data point, and transmit them to a remote monitoring platform via IoT communication.

[0017] S2. Convert the solar illumination incident direction data into a spherical vector sequence, and use the VonMises–Fisher distribution for modeling to obtain the concentration parameter and mean direction parameter corresponding to each time label, which are used as the illumination modeling results.

[0018] S3. Based on the illumination modeling results and solar radiation intensity data, generate photovoltaic power generation capacity prediction data for each time tag;

[0019] S4. Input the photovoltaic power generation capacity prediction data and the sewage treatment equipment operation status data into the improved spherical regression algorithm, and output the energy consumption prediction data corresponding to each time tag;

[0020] S5. Compare the photovoltaic power generation capacity prediction data with the energy consumption prediction data, calculate the operating frequency of the sewage treatment equipment under each time tag, and generate the initial parameter sequence of the operating frequency.

[0021] S6. Optimize the initial parameter sequence of the operating frequency based on the illumination modeling results and energy consumption prediction data to generate a set of operating frequency adjustment parameters;

[0022] S7. Outputs remote control commands containing time stamps, operating frequency adjustment parameters, and energy consumption prediction data, serving as the control basis for the dynamic operation of wastewater treatment equipment.

[0023] This invention proposes an IoT-based remote monitoring method for wastewater treatment, establishing a closed-loop process from light and operational data acquisition, spherical vector modeling, photovoltaic power generation prediction, energy consumption direction estimation, adaptive frequency calculation, to control command generation. This effectively solves the problems of inaccurate energy scheduling, lagging frequency regulation, and insufficient remote control response in wastewater treatment. The method utilizes the Von Mises–Fisher distribution to model the incident light direction, extracting concentration and direction parameters. Combined with radiation intensity, it predicts photovoltaic power generation and constructs an energy consumption prediction model using an improved spherical regression algorithm, generating confidence intervals. The system dynamically generates and optimizes operating frequency parameters based on energy balance, ultimately outputting time-stamped control commands to achieve intelligent equipment operation. This method possesses high precision, real-time performance, and remote adaptability, making it suitable for green energy wastewater treatment systems under complex lighting environments.

[0024] Optionally, step S1 includes the following specific steps:

[0025] S11. Collect data on the direction of solar illumination and the intensity of solar radiation at the wastewater treatment site. The data on the direction of solar illumination consists of the azimuth angle and the elevation angle. The data on the intensity of solar radiation represents the light energy radiation density per unit time.

[0026] S12. Collect operating status data of sewage treatment equipment. The operating status data includes operating frequency, current value, input power and running time. Each set of data is simultaneously tagged with a time label during collection.

[0027] S13. Encapsulate the data on the incident direction of sunlight, the data on solar radiation intensity, and the data on the operating status of wastewater treatment equipment into data units according to a unified field structure. Each data unit contains a time tag, a data type identifier, and a corresponding numerical content.

[0028] S14. The data unit is sent to the remote monitoring platform in real time through Internet of Things (IoT) communication, including a two-way data link based on cellular communication, narrowband IoT, or industrial wireless communication.

[0029] S15. The remote monitoring platform includes a communication interface for receiving transmitted data units, a data processing structure for parsing, sorting, and verifying received data according to time tags, a data organization structure for constructing a time series data set after parsing, and a data storage structure for the constructed time series data set.

[0030] By collecting, structuring, and synchronizing data on sunlight direction, radiation intensity, and equipment operating status with time tags, this invention ensures the consistency and high timeliness of multi-source data in an IoT environment, solving the modeling bias problems caused by asynchronous data, inconsistent formats, and communication delays in wastewater treatment systems. A unified field structure is used to encapsulate data units, and real-time transmission is achieved through bidirectional links such as cellular and narrowband IoT. This enables remote monitoring platforms to parse, sort, and verify data according to time tags, constructing a stable and reliable time-series dataset. This structure provides a high-quality, low-latency data input foundation for subsequent illumination modeling and energy consumption prediction, significantly improving the accuracy and robustness of system operation.

[0031] Optionally, step S2 includes the following specific steps:

[0032] S21. Based on the sunlight incident direction data collected for each time tag, extract the azimuth and elevation angle information, convert them into unit spherical vectors, and construct a direction vector sequence with time tags: x(t) = [cosβ(t)cosα(t),cosβ(t)sinα(t),sinβ(t)]; where α(t) is the azimuth angle, β(t) is the elevation angle, and x(t) is the direction vector of the corresponding time tag.

[0033] S22. Divide the sequence of direction vectors within a continuous time period into two subsets according to their rate of change: the first subset is the set of direct components with low rate of change, and the second subset is the set of scattering components with high rate of change. Set a boundary point based on the rate of change threshold and assign each direction vector to the corresponding subset.

[0034] S23. Construct a first-layer VonMises–Fisher distribution model for the direct sunlight component set, and calculate the directional mean vector. The mean vector is obtained by summing the unit vectors within the direct sunlight component set, superimposing them with the photovoltaic array azimuth vector, and then normalizing the result. The calculation formula is as follows:

[0035]

[0036] Among them, x i Let b be the i-th unit vector in the direct set. q Let η be the orientation vector of the photovoltaic array, and η be the bias weighting coefficient.

[0037] S24. Construct a concentration parameter for the direct component based on the direction mean vector and the sample variance. The concentration parameter κ d (t) is calculated using the following formula:

[0038]

[0039] Where σ(t) is the angular variance of the unit vector relative to the mean direction, γ is the adjustment coefficient, and ∈ is a positive constant to avoid division by zero;

[0040] S25. Establish a second-layer Von Mises–Fisher distribution model for the scattering component set, setting the mean direction vector to a fixed direction or a zero vector, and setting the concentration parameter to a constant or determined by κ. d (t) Inverse proportional derivation;

[0041] S26. Based on the ratio of total solar radiation intensity to diffuse radiation intensity obtained by the sensor, calculate the direct radiation ratio under each time tag, define it as λ(t), and use it as the distribution mixing weight;

[0042] S27. Using λ(t) as a weighting factor, the direct and scattering distributions are linearly combined to form a hierarchical mixed modeling result. The modeling result includes: λ(t), κ d (t), κ s (t), μ d (t), μ s (t) constitutes the illumination modeling results for each time label.

[0043] This invention establishes a complete process for modeling illumination direction based on the Von Mises–Fisher distribution, encompassing raw illumination data conversion, component partitioning, model construction, and hybrid output. This effectively addresses the limitations of traditional illumination modeling methods, which cannot simultaneously represent direct and scattered characteristics and suffer from insufficient modeling accuracy. The method first extracts the azimuth and elevation angles for each time label, converting them into unit spherical vectors. These vectors are then divided into direct and scattered component sets based on their rate of change. Two Von Mises–Fisher models are constructed, and the direct direction is corrected by introducing the superposition of photovoltaic array direction vectors. The concentration is calculated based on sample variance, enhancing the stability of the direction estimation. Finally, by combining the total radiation and scattering ratios, weighting coefficients are dynamically generated to achieve a layered hybrid output. This improves the adaptability of illumination modeling to complex real-world illumination environments, providing a highly directional and discriminative data foundation for subsequent power generation prediction.

[0044] Optionally, step S3 includes the following specific steps:

[0045] S31. Extract the illumination modeling results corresponding to each time tag. The illumination modeling results include the concentration parameter and mean direction vector of the direct component, the concentration parameter and mean direction vector of the scattered component, and the weighting factor of the direct component.

[0046] S32. Extract the solar radiation intensity data corresponding to each time tag. The solar radiation intensity data includes the total radiation intensity value, the direct radiation intensity value, and the diffuse radiation intensity value.

[0047] S33. Construct an input feature vector from the illumination modeling parameters and solar radiation intensity data corresponding to each time tag. The feature vector includes: direct radiation concentration parameter, direct radiation mean direction vector component, scattering concentration parameter, scattering mean direction vector component, direct radiation weight, direct radiation intensity and scattered radiation intensity.

[0048] S34. Construct a training sample set based on time labels. The training sample set contains historical illumination modeling parameters, radiation intensity data and measured power generation data for the corresponding time.

[0049] S35. Construct a regression neural network model using the training sample set. The neural network is a multi-layer feedforward structure. Each layer uses a combination of linear transformation and nonlinear activation. The loss function is the mean square error between the predicted power generation and the measured value.

[0050] S36. Input the input feature vector corresponding to each time tag into the trained neural network model, and output the photovoltaic power generation prediction value corresponding to each time tag. All prediction values ​​constitute photovoltaic power generation capacity prediction data arranged in chronological order.

[0051] This invention establishes a systematic process for predicting power generation by constructing a neural network prediction method that integrates illumination modeling parameters and radiation intensity features. This process encompasses modeling result extraction, feature vector construction, sample set generation, and power generation prediction, effectively addressing the shortcomings of traditional methods such as insufficient understanding of illumination directionality and low prediction accuracy. The method extracts the direction and concentration parameters of direct and scattered components at each time label and constructs a multi-dimensional input feature vector by combining it with radiation intensity. A multi-layer feedforward neural network is trained using historical measured data, enabling the model to nonlinearly fit complex illumination-power generation relationships. The predicted output forms continuous time-series data of power generation capacity, providing high-precision, structured input support for subsequent energy consumption comparison and frequency optimization, significantly improving the system's responsiveness to illumination changes and the reliability of power generation prediction.

[0052] Optionally, step S4 includes the following specific steps:

[0053] S41. Based on each time tag, extract the photovoltaic power generation capacity prediction data and the sewage treatment equipment operation status data, and construct an input feature sequence. The input feature sequence includes the photovoltaic power generation power prediction value, equipment operating frequency, current, voltage, energy consumption direction change value of the previous time tag, and power difference.

[0054] S42. Construct an improved spherical regression algorithm, which includes a feature injection layer, an illumination weight gating layer, a spherical projection layer, and a confidence interval estimation module:

[0055] S421. In the feature injection layer, perform affine transformation and nonlinear mapping on the input features to obtain the hidden vector h(t);

[0056] S422. In the illumination weight gating layer, the direct component weight λ(t) from the illumination modeling result corresponding to the time label is introduced to construct a dual-channel gating mechanism for split calculation of the hidden vector:

[0057]

[0058] Among them, W d With W s These are the weight matrices for the direct and scattering channels, respectively;

[0059] S423. In the spherical projection layer, the gated output vector Perform an affine mapping and normalize to a unit sphere to obtain the energy dissipation direction prediction vector:

[0060]

[0061] Among them, W p With b p These are the parameters of the spherical mapping layer;

[0062] S424. In the confidence interval estimation module, based on the magnitude of the hidden vector... Construct the spherical cap angle θ c (t) = arcsinr(t), with Construct a spherical cap sampling set {v} containing M directional samples, centered at the cap. j (t)};

[0063] S425. Perform the inner product of each sampling direction vector and the predefined energy consumption reference vector e to calculate the set of predicted energy consumption values: And calculate the upper and lower boundaries E of the confidence interval for energy consumption prediction. low (t),E high (t), and the central prediction value

[0064] S43. Output the energy consumption direction prediction vector for each time tag. Energy consumption center predicted value E mid (t), along with the confidence interval boundary values, constitute the energy consumption prediction data.

[0065] This invention establishes an improved spherical regression algorithm that integrates illumination modeling weights and operational status features, creating an energy consumption modeling process from multi-dimensional feature injection and direction vector prediction to confidence interval output. This effectively addresses the problems of insufficient directional expression, uncontrollable confidence, and delayed response in traditional energy consumption prediction methods. The method uses the predicted power generation value and equipment operating status as the input feature sequence. Hidden features are extracted through affine transformation and nonlinear activation. A dual-channel gating mechanism is constructed by introducing the direct sunlight component weight from illumination modeling to achieve targeted guidance of energy consumption direction information. Subsequently, spherical affine projection is used to map the features to a unit sphere and construct a spherical cap sampling set. The energy consumption prediction value set is calculated by the inner product with the reference vector, ultimately outputting the direction prediction vector, the center prediction value, and the confidence boundary. This improves the directional expressiveness and confidence control capability of the prediction results, providing more reliable data support for frequency regulation.

[0066] Optionally, the improved spherical regression algorithm includes:

[0067] The feature injection layer includes: an initial input layer for receiving photovoltaic power generation capacity prediction data and sewage treatment equipment operation status data; an affine transformation layer for performing affine transformation on the input features to generate a preliminary feature representation; and a nonlinear activation layer for applying an activation function to the result of the affine transformation for nonlinear mapping.

[0068] The illumination weighting gate layer includes: a direct component weight input, used to receive the direct component weight in the illumination modeling result; and a dual-channel splitting mechanism, used to perform splitting calculation on the input features according to the direct component weight, generating direct channel and scattering channel outputs.

[0069] The spherical projection layer includes: an affine mapping layer for performing a linear transformation on the gated output; and a normalization processing layer for normalizing the mapped output to a unit sphere to generate the final energy consumption direction prediction vector.

[0070] The confidence interval estimation module includes: a modulus calculation unit for calculating the modulus of the hidden vector; a radius calculation unit for calculating the confidence radius; a spherical cap sampling unit for generating samples in multiple directions; and an energy consumption calculation unit for calculating the predicted energy consumption value based on the sampling direction and the energy consumption reference vector, and generating the upper and lower limits of the confidence interval.

[0071] This invention designs a structured, improved spherical regression algorithm to transform power generation prediction and operational status data into directional energy consumption expressions. It constructs a multi-layered computational model with spatial mapping and confidence estimation capabilities, effectively solving the problems of poor directional guidance, uncontrollable intervals, and weak structural interpretability in traditional energy consumption prediction methods. The algorithm consists of a feature injection layer, an illumination weight gating layer, a spherical projection layer, and a confidence interval estimation module. Layer by layer, it implements affine transformation and nonlinear mapping, then dual-channel gating based on direct illumination weights, and finally, directional vector prediction and confidence angle estimation on a unit sphere. By introducing a spherical cap sampling strategy and a reference vector inner product mechanism, it ultimately achieves closed-loop output of the upper and lower limits of the confidence interval, providing highly directional and reliable energy consumption prediction support for wastewater treatment frequency control.

[0072] Optionally, step S5 includes the following specific steps:

[0073] S51. Extract the photovoltaic power generation capacity prediction data and energy consumption prediction data corresponding to each time tag, which respectively represent the available power and estimated load power consumption of the sewage treatment plant.

[0074] S52. Based on each time tag, compare the photovoltaic power generation forecast with the energy consumption forecast to obtain the energy balance difference; when the photovoltaic power generation forecast is greater than the energy consumption forecast, it indicates that there is power redundancy, and when it is less than, it indicates that there is power shortage.

[0075] S53. Combining the energy balance difference with the operating frequency under the previous time tag, calculate the initial value of the operating frequency corresponding to the current time tag. The calculation rules include: when there is power redundancy, the operating frequency is set to be proportionally increased based on the previous frequency; when there is power shortage, the operating frequency is set to be proportionally decreased based on the previous frequency; when supply and demand are basically balanced, the operating frequency remains unchanged.

[0076] S54. Set the upper and lower limits of the operating frequency of the sewage treatment equipment, and make boundary corrections to the initial value of the above operating frequency so that it does not exceed the maximum operating frequency and is not lower than the minimum operating frequency.

[0077] S55. Arrange the calculated operating frequencies under all time tags into a vector in chronological order, and use it as the initial parameter sequence for the operating frequencies.

[0078] This invention establishes a closed-loop control process based on an energy balance-based operating frequency calculation mechanism. This process, from comparing power generation and energy consumption, deriving frequency adjustment rules, to generating an initial frequency sequence based on boundary constraints, effectively solves the problem of relying on manual experience and failing to match real-time supply and demand changes in wastewater treatment systems. The method calculates the energy difference based on the predicted photovoltaic power generation and energy consumption values ​​at each time point, and dynamically adjusts the current frequency in conjunction with the previous operating frequency, forming initial frequency parameters with directionality and trend. By introducing upper and lower limit control rules for frequency, it ensures that the adjustment results do not exceed limits or become unstable, ultimately generating a complete time-series initial frequency value, providing a reasonable starting point for subsequent frequency optimization modules. This method realizes an adaptive operation strategy based on energy matching, improving the flexibility, stability, and energy utilization efficiency of regulation.

[0079] Optionally, step S6 includes the following specific steps:

[0080] S61. Read the initial parameter sequence of the running frequency, the illumination modeling results and the energy consumption prediction data corresponding to each time tag, and match them one by one according to the time tag;

[0081] S62. Perform a difference operation on the initial parameter sequence of the operating frequency to obtain the frequency change of adjacent time tags; at the same time, calculate the rate of change of the weight of the direct component in the illumination modeling results and the rate of change of the energy consumption prediction data.

[0082] S63. Construct an adaptive adjustment coefficient based on the frequency change and the rate of change of illumination weight; when the direct illumination weight increases rapidly and the energy consumption prediction fluctuation decreases, the adaptive adjustment coefficient increases; when the direct illumination weight decreases or the energy consumption fluctuation increases, the adaptive adjustment coefficient decreases.

[0083] S64. Using an adaptive adjustment coefficient, the initial parameter sequence of the operating frequency is subjected to sliding window weighted smoothing to reduce high-frequency oscillations and retain the trend response; the smoothing result is used as a candidate frequency sequence.

[0084] S65. Compare the candidate frequency sequence with the energy consumption prediction data point by point. If the candidate frequency causes the energy consumption prediction value to exceed the upper limit of the confidence interval at any time label, the corresponding frequency is adjusted down proportionally. If it causes the energy consumption prediction to fall below the lower limit of the confidence interval, the corresponding frequency is adjusted up proportionally to form the adjusted frequency sequence.

[0085] S66. Apply threshold restrictions to the adjusted frequency sequence to keep it within the minimum and maximum frequency range allowed by the wastewater treatment equipment, thereby obtaining a set of operating frequency adjustment parameters that meet the requirements of the control interface.

[0086] S67. Save the set of operating frequency adjustment parameters in time-stamped order and output it to the remote monitoring platform for use in the next step of generating remote control commands.

[0087] This invention introduces an adaptive frequency optimization method based on dynamic construction of light intensity weights and predicted energy consumption changes. This forms an optimized closed loop from initial frequency value acquisition, dynamic smoothing, confidence interval verification to final frequency adjustment parameter generation, effectively solving the problems of large fluctuations, unstable response, and predicted deviations from control targets in wastewater treatment systems. The method first extracts the initial operating frequency value along with the corresponding light intensity modeling results and energy consumption prediction data, and calculates the adjustment coefficient based on their rate of change. The frequency sequence is adaptively smoothed to suppress ineffective oscillations. Furthermore, candidate frequencies are compared with confidence interval boundaries, dynamically adjusting or compensating for abnormal deviations to ensure that predicted energy consumption falls within a reasonable range. Finally, a set of frequency adjustment parameters that meets equipment limitations is formed, providing a stable, efficient, and intelligent control command basis for the remote control module.

[0088] Optionally, step S7 includes the following specific steps:

[0089] S71. Extract the operating frequency adjustment parameters and energy consumption prediction data corresponding to each time tag, and write the three types of information—time tag, operating frequency adjustment parameters, and energy consumption prediction data—into the instruction data structure in a unified field order.

[0090] S72. According to the byte order specified in the remote control protocol, allocate a message header, message body and digest check segment to the instruction data structure. The message header contains a timestamp, the message body records the operating frequency adjustment parameters and energy consumption prediction values ​​respectively, and the digest check segment uses a hash function to generate a check value.

[0091] S73. Perform quantization encoding on the operating frequency adjustment parameters in the message body to ensure that the frequency values ​​meet the requirements of integerization, fixed step size and minimum resolution; at the same time, perform floating-point compression encoding on the energy consumption prediction data to maintain the numerical accuracy not lower than the preset error threshold.

[0092] S74. Based on the bandwidth of the remote communication link, select a narrowband IoT or cellular network channel, and push the encoded instruction message to the wastewater treatment site control terminal within the specified time window. During the transmission process, add a sequence identifier to the message header.

[0093] S75. After receiving a command message, the field control terminal compares the time stamp in the message header with the clock. If the time stamp is earlier than the local clock, it executes immediately; if the time stamp is later than the local clock, it is buffered and executed only after clock synchronization.

[0094] S76. The field control terminal parses the message body, reads the operating frequency adjustment parameters and energy consumption prediction data, writes them into the drive register in the order of frequency first and energy consumption second, and records the execution receipt number.

[0095] S77. After receiving the acknowledgment number, the remote monitoring platform marks the instruction status as executed, archives the instruction content and execution time, and completes the remote control instruction output process that includes time stamps, operating frequency adjustment parameters and energy consumption prediction data.

[0096] This invention establishes a standardized, time-sequential remote control command generation and execution mechanism, creating a complete control loop from frequency and energy consumption prediction data encapsulation, message structure construction, encoding compression, network transmission to terminal parsing, execution, and status feedback. This effectively solves the problems of inconsistent control command formats, unstable communication, and unreliable execution responses in wastewater treatment systems. The method organizes operating frequency adjustment parameters and energy consumption prediction values ​​according to time tags, generating control messages with headers, bodies, and checksums. Quantization encoding and floating-point compression are used to improve transmission efficiency and accuracy. The messages are sent sequentially to the terminal via channels such as NB-IoT, and a time comparison mechanism ensures timely and accurate execution of control commands. The monitoring platform tracks the control loop based on feedback. This method enhances the system's remote control stability, communication security, and execution verifiability.

[0097] The beneficial effects of this invention are:

[0098] This invention introduces a solar illumination direction modeling method based on the Von Mises–Fisher distribution, and combines it with solar radiation intensity data to accurately extract the characteristics of direct and diffused light, significantly improving the accuracy of photovoltaic power generation capacity prediction and effectively solving the problem of coarse modeling in variable environments by traditional models.

[0099] This invention proposes an improved spherical regression algorithm that integrates equipment operating status data and illumination information to construct an energy consumption prediction model. It has the ability to express directionality and estimate confidence intervals, solving the problem that traditional energy consumption estimation cannot provide reliability assessment and improving the credibility and adaptability of prediction results.

[0100] This invention dynamically generates an initial value for the operating frequency by comparing the predicted power generation capacity with the energy consumption, and introduces an adaptive adjustment coefficient to smooth and optimize the frequency sequence, which significantly reduces the energy loss caused by drastic fluctuations in the operating frequency and achieves intelligent matching between energy consumption and energy supply. Attached Figure Description

[0101] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0102] Figure 1 The flowchart shows the wastewater treatment remote monitoring system and method based on the Internet of Things proposed in this invention.

[0103] Figure 2 This is a flowchart of the modeling process for extracting illumination features based on the Von Mises–Fisher distribution proposed in this invention.

[0104] Figure 3 This is a schematic diagram of the energy consumption prediction structure based on the improved spherical regression algorithm proposed in this invention. Detailed Implementation

[0105] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0106] refer to Figure 1-3 A wastewater treatment remote monitoring system and method based on the Internet of Things includes the following steps:

[0107] Data acquisition module: used to collect data on the direction of sunlight incidence, solar radiation intensity, and the operating status of wastewater treatment equipment, and to attach a unified time label to all types of data;

[0108] Data transmission module: used to transmit collected data to a remote monitoring platform in real time via narrowband IoT, cellular networks or industrial wireless communication.

[0109] Illumination Modeling Module: Used to perform directional modeling of solar illumination data based on the Von Mises–Fisher distribution, generating the concentration and directional parameters of direct and diffuse components;

[0110] Power generation capacity prediction module: used to combine the results of illumination modeling with solar radiation data to calculate the time-series predicted value of photovoltaic power generation capacity;

[0111] Energy consumption prediction module: Used to input power generation prediction values ​​and equipment operating status data, and output energy consumption prediction results and confidence intervals through an improved spherical regression algorithm;

[0112] Frequency optimization module: used to compare power generation and energy consumption data, calculate operating frequency parameters, and perform adaptive adjustments based on the predicted confidence interval;

[0113] Remote control module: Used to generate time-stamped operating frequency adjustment commands, which are sent to the equipment terminal through the communication interface to control the dynamic operation of the sewage treatment system.

[0114] In this embodiment, the modules are connected through the following method:

[0115] S1. Collect data on the incident direction of sunlight, solar radiation intensity, and the operating status of wastewater treatment equipment, add time tags to each data point, and transmit them to a remote monitoring platform via IoT communication.

[0116] S2. Convert the solar illumination incident direction data into a spherical vector sequence, and use the VonMises–Fisher distribution for modeling to obtain the concentration parameter and mean direction parameter corresponding to each time label, which are used as the illumination modeling results.

[0117] S3. Based on the illumination modeling results and solar radiation intensity data, generate photovoltaic power generation capacity prediction data for each time tag;

[0118] S4. Input the photovoltaic power generation capacity prediction data and the sewage treatment equipment operation status data into the improved spherical regression algorithm, and output the energy consumption prediction data corresponding to each time tag;

[0119] S5. Compare the photovoltaic power generation capacity prediction data with the energy consumption prediction data, calculate the operating frequency of the sewage treatment equipment under each time tag, and generate the initial parameter sequence of the operating frequency.

[0120] S6. Optimize the initial parameter sequence of the operating frequency based on the illumination modeling results and energy consumption prediction data to generate a set of operating frequency adjustment parameters;

[0121] S7. Outputs remote control commands containing time stamps, operating frequency adjustment parameters, and energy consumption prediction data, serving as the control basis for the dynamic operation of wastewater treatment equipment.

[0122] In this embodiment, step S1 includes the following specific steps:

[0123] S11. Collect data on the direction of solar illumination and the intensity of solar radiation at the wastewater treatment site. The data on the direction of solar illumination consists of the azimuth angle and the elevation angle. The data on the intensity of solar radiation represents the light energy radiation density per unit time.

[0124] S12. Collect operating status data of sewage treatment equipment. The operating status data includes operating frequency, current value, input power and running time. Each set of data is simultaneously tagged with a time label during collection.

[0125] S13. Encapsulate the data on the incident direction of sunlight, the data on solar radiation intensity, and the data on the operating status of wastewater treatment equipment into data units according to a unified field structure. Each data unit contains a time tag, a data type identifier, and a corresponding numerical content.

[0126] S14. The data unit is sent to the remote monitoring platform in real time through Internet of Things (IoT) communication, including a two-way data link based on cellular communication, narrowband IoT, or industrial wireless communication.

[0127] S15. The remote monitoring platform includes a communication interface for receiving transmitted data units, a data processing structure for parsing, sorting, and verifying received data according to time tags, a data organization structure for constructing a time series data set after parsing, and a data storage structure for the constructed time series data set.

[0128] In this embodiment, step S2 includes the following specific steps:

[0129] S21. Based on the sunlight incident direction data collected for each time tag, extract the azimuth and elevation angle information, convert them into unit spherical vectors, and construct a direction vector sequence with time tags: x(t) = [cosβ(t)cosα(t),cosβ(t)sinα(t),sinβ(t)]; where α(t) is the azimuth angle, β(t) is the elevation angle, and x(t) is the direction vector of the corresponding time tag.

[0130] S22. Divide the sequence of direction vectors within a continuous time period into two subsets according to their rate of change: the first subset is the set of direct components with low rate of change, and the second subset is the set of scattering components with high rate of change. Set a boundary point based on the rate of change threshold and assign each direction vector to the corresponding subset.

[0131] S23. Construct a first-layer VonMises–Fisher distribution model for the direct sunlight component set, and calculate the directional mean vector. The mean vector is obtained by summing the unit vectors within the direct sunlight component set, superimposing them with the photovoltaic array azimuth vector, and then normalizing the result. The calculation formula is as follows:

[0132]

[0133] Among them, x i Let b be the i-th unit vector in the direct set. q Let η be the orientation vector of the photovoltaic array, and η be the bias weighting coefficient.

[0134] S24. Construct a concentration parameter for the direct component based on the direction mean vector and the sample variance. The concentration parameter κ d (t) is calculated using the following formula:

[0135]

[0136] Where σ(t) is the angular variance of the unit vector relative to the mean direction, γ is the adjustment coefficient, and ∈ is a positive constant to avoid division by zero;

[0137] S25. Establish a second-layer Von Mises–Fisher distribution model for the scattering component set, setting the mean direction vector to a fixed direction or a zero vector, and setting the concentration parameter to a constant or determined by κ. d (t) Inverse proportional derivation;

[0138] S26. Based on the ratio of total solar radiation intensity to diffuse radiation intensity obtained by the sensor, calculate the direct radiation ratio under each time tag, define it as λ(t), and use it as the distribution mixing weight;

[0139] S27. Using λ(t) as a weighting factor, the direct and scattering distributions are linearly combined to form a hierarchical mixed modeling result. The modeling result includes: λ(t), κ d (t), κ s (t), μ d (t), μ s (t) constitutes the illumination modeling results for each time label.

[0140] In this embodiment, step S3 includes the following specific steps:

[0141] S31. Extract the illumination modeling results corresponding to each time tag. The illumination modeling results include the concentration parameter and mean direction vector of the direct component, the concentration parameter and mean direction vector of the scattered component, and the weighting factor of the direct component.

[0142] S32. Extract the solar radiation intensity data corresponding to each time tag. The solar radiation intensity data includes the total radiation intensity value, the direct radiation intensity value, and the diffuse radiation intensity value.

[0143] S33. Construct an input feature vector from the illumination modeling parameters and solar radiation intensity data corresponding to each time tag. The feature vector includes: direct radiation concentration parameter, direct radiation mean direction vector component, scattering concentration parameter, scattering mean direction vector component, direct radiation weight, direct radiation intensity and scattered radiation intensity.

[0144] S34. Construct a training sample set based on time labels. The training sample set contains historical illumination modeling parameters, radiation intensity data and measured power generation data for the corresponding time.

[0145] S35. Construct a regression neural network model using the training sample set. The neural network is a multi-layer feedforward structure. Each layer uses a combination of linear transformation and nonlinear activation. The loss function is the mean square error between the predicted power generation and the measured value.

[0146] S36. Input the input feature vector corresponding to each time tag into the trained neural network model, and output the photovoltaic power generation prediction value corresponding to each time tag. All prediction values ​​constitute photovoltaic power generation capacity prediction data arranged in chronological order.

[0147] In this embodiment, step S4 includes the following specific steps:

[0148] S41. Based on each time tag, extract the photovoltaic power generation capacity prediction data and the sewage treatment equipment operation status data, and construct an input feature sequence. The input feature sequence includes the photovoltaic power generation power prediction value, equipment operating frequency, current, voltage, energy consumption direction change value of the previous time tag, and power difference.

[0149] S42. Construct an improved spherical regression algorithm, which includes a feature injection layer, an illumination weight gating layer, a spherical projection layer, and a confidence interval estimation module:

[0150] S421. In the feature injection layer, perform affine transformation and nonlinear mapping on the input features to obtain the hidden vector h(t);

[0151] S422. In the illumination weight gating layer, the direct component weight λ(t) from the illumination modeling result corresponding to the time label is introduced to construct a dual-channel gating mechanism for split calculation of the hidden vector:

[0152]

[0153] Among them, W d With W s These are the weight matrices for the direct and scattering channels, respectively;

[0154] S423. In the spherical projection layer, the gated output vector Perform an affine mapping and normalize to a unit sphere to obtain the energy dissipation direction prediction vector:

[0155]

[0156] Among them, W p With b p These are the parameters of the spherical mapping layer;

[0157] S424. In the confidence interval estimation module, based on the magnitude of the hidden vector... Construct the spherical cap angle θ c (t) = arcsinr(t), with Construct a spherical cap sampling set {v} containing M directional samples, centered at the cap. j (t)};

[0158] S425. Perform the inner product of each sampling direction vector and the predefined energy consumption reference vector e to calculate the set of predicted energy consumption values: And calculate the upper and lower boundaries E of the confidence interval for energy consumption prediction. low (t),E high (t), and the central prediction value

[0159] S43. Output the energy consumption direction prediction vector for each time tag. Energy consumption center predicted value E mid (t), along with the confidence interval boundary values, constitute the energy consumption prediction data.

[0160] In this embodiment, the improved spherical regression algorithm includes:

[0161] The feature injection layer includes: an initial input layer for receiving photovoltaic power generation capacity prediction data and sewage treatment equipment operation status data; an affine transformation layer for performing affine transformation on the input features to generate a preliminary feature representation; and a nonlinear activation layer for applying an activation function to the result of the affine transformation for nonlinear mapping.

[0162] The illumination weighting gate layer includes: a direct component weight input, used to receive the direct component weight in the illumination modeling result; and a dual-channel splitting mechanism, used to perform splitting calculation on the input features according to the direct component weight, generating direct channel and scattering channel outputs.

[0163] The spherical projection layer includes: an affine mapping layer for performing a linear transformation on the gated output; and a normalization processing layer for normalizing the mapped output to a unit sphere to generate the final energy consumption direction prediction vector.

[0164] The confidence interval estimation module includes: a modulus calculation unit for calculating the modulus of the hidden vector; a radius calculation unit for calculating the confidence radius; a spherical cap sampling unit for generating samples in multiple directions; and an energy consumption calculation unit for calculating the predicted energy consumption value based on the sampling direction and the energy consumption reference vector, and generating the upper and lower limits of the confidence interval.

[0165] In this embodiment, step S5 includes the following specific steps:

[0166] S51. Extract the photovoltaic power generation capacity prediction data and energy consumption prediction data corresponding to each time tag, which respectively represent the available power and estimated load power consumption of the sewage treatment plant.

[0167] S52. Based on each time tag, compare the photovoltaic power generation forecast with the energy consumption forecast to obtain the energy balance difference; when the photovoltaic power generation forecast is greater than the energy consumption forecast, it indicates that there is power redundancy, and when it is less than, it indicates that there is power shortage.

[0168] S53. Combining the energy balance difference with the operating frequency under the previous time tag, calculate the initial value of the operating frequency corresponding to the current time tag. The calculation rules include: when there is power redundancy, the operating frequency is set to be proportionally increased based on the previous frequency; when there is power shortage, the operating frequency is set to be proportionally decreased based on the previous frequency; when supply and demand are basically balanced, the operating frequency remains unchanged.

[0169] S54. Set the upper and lower limits of the operating frequency of the sewage treatment equipment, and make boundary corrections to the initial value of the above operating frequency so that it does not exceed the maximum operating frequency and is not lower than the minimum operating frequency.

[0170] S55. Arrange the calculated operating frequencies under all time tags into a vector in chronological order, and use it as the initial parameter sequence for the operating frequencies.

[0171] In this embodiment, step S6 includes the following specific steps:

[0172] S61. Read the initial parameter sequence of the running frequency, the illumination modeling results and the energy consumption prediction data corresponding to each time tag, and match them one by one according to the time tag;

[0173] S62. Perform a difference operation on the initial parameter sequence of the operating frequency to obtain the frequency change of adjacent time tags; at the same time, calculate the rate of change of the weight of the direct component in the illumination modeling results and the rate of change of the energy consumption prediction data.

[0174] S63. Construct an adaptive adjustment coefficient based on the frequency change and the rate of change of illumination weight; when the direct illumination weight increases rapidly and the energy consumption prediction fluctuation decreases, the adaptive adjustment coefficient increases; when the direct illumination weight decreases or the energy consumption fluctuation increases, the adaptive adjustment coefficient decreases.

[0175] S64. Using an adaptive adjustment coefficient, the initial parameter sequence of the operating frequency is subjected to sliding window weighted smoothing to reduce high-frequency oscillations and retain the trend response; the smoothing result is used as a candidate frequency sequence.

[0176] S65. Compare the candidate frequency sequence with the energy consumption prediction data point by point. If the candidate frequency causes the energy consumption prediction value to exceed the upper limit of the confidence interval at any time label, the corresponding frequency is adjusted down proportionally. If it causes the energy consumption prediction to fall below the lower limit of the confidence interval, the corresponding frequency is adjusted up proportionally to form the adjusted frequency sequence.

[0177] S66. Apply threshold restrictions to the adjusted frequency sequence to keep it within the minimum and maximum frequency range allowed by the wastewater treatment equipment, thereby obtaining a set of operating frequency adjustment parameters that meet the requirements of the control interface.

[0178] S67. Save the set of operating frequency adjustment parameters in time-stamped order and output it to the remote monitoring platform for use in the next step of generating remote control commands.

[0179] In this embodiment, step S7 includes the following specific steps:

[0180] S71. Extract the operating frequency adjustment parameters and energy consumption prediction data corresponding to each time tag, and write the three types of information—time tag, operating frequency adjustment parameters, and energy consumption prediction data—into the instruction data structure in a unified field order.

[0181] S72. According to the byte order specified in the remote control protocol, allocate a message header, message body and digest check segment to the instruction data structure. The message header contains a timestamp, the message body records the operating frequency adjustment parameters and energy consumption prediction values ​​respectively, and the digest check segment uses a hash function to generate a check value.

[0182] S73. Perform quantization encoding on the operating frequency adjustment parameters in the message body to ensure that the frequency values ​​meet the requirements of integerization, fixed step size and minimum resolution; at the same time, perform floating-point compression encoding on the energy consumption prediction data to maintain the numerical accuracy not lower than the preset error threshold.

[0183] S74. Based on the bandwidth of the remote communication link, select a narrowband IoT or cellular network channel, and push the encoded instruction message to the wastewater treatment site control terminal within the specified time window. During the transmission process, add a sequence identifier to the message header.

[0184] S75. After receiving a command message, the field control terminal compares the time stamp in the message header with the clock. If the time stamp is earlier than the local clock, it executes immediately; if the time stamp is later than the local clock, it is buffered and executed only after clock synchronization.

[0185] S76. The field control terminal parses the message body, reads the operating frequency adjustment parameters and energy consumption prediction data, writes them into the drive register in the order of frequency first and energy consumption second, and records the execution receipt number.

[0186] S77. After receiving the acknowledgment number, the remote monitoring platform marks the instruction status as executed, archives the instruction content and execution time, and completes the remote control instruction output process that includes time stamps, operating frequency adjustment parameters and energy consumption prediction data.

[0187] Example 1:

[0188] To verify the effectiveness and feasibility of the IoT-based remote monitoring system and method for wastewater treatment proposed in this invention, a wastewater treatment plant was selected as the actual application scenario. This wastewater treatment plant has a daily treatment capacity of 6,000 tons and is equipped with a photovoltaic power supply system. However, in its current operation, it faces problems such as low energy utilization, reliance on manual experience for operation and control, and response delays. Especially during periods of significant fluctuation in sunlight, it is impossible to accurately adjust the equipment operating frequency based on real-time power generation capacity, leading to increased mains power consumption and high equipment energy consumption.

[0189] Prior to the application of this invention's system, the treatment plant operated at a fixed frequency. Each day, on-duty personnel set the operating frequency of equipment such as aeration blowers and water pumps based on experience, which in most cases failed to reflect changes in photovoltaic power generation capacity in real time. Furthermore, the system lacked an energy consumption prediction model, leading to frequent overload or inefficiency issues.

[0190] After introducing the system of this invention, light sensors, radiation intensity meters, and operational status acquisition modules were deployed at the wastewater treatment site. The system collects data such as solar illumination direction (azimuth and elevation angle), total solar radiation intensity, diffuse radiation intensity, and real-time operating frequency, current, and voltage of key equipment. All data is tagged with a unified time stamp and transmitted to a remote monitoring platform via narrowband Internet of Things (IoT). The platform uses the Von Mises–Fisher distribution for illumination modeling, extracting the direction vectors and concentration parameters of direct and diffuse components; combined with radiation data, it predicts photovoltaic power generation. Subsequently, using the predicted power generation and equipment operating status as inputs, an improved spherical regression algorithm is applied to generate predicted energy consumption values ​​and confidence intervals.

[0191] Based on the discrepancy between energy consumption forecasts and power generation forecasts, the platform calculates the initial operating frequency of the equipment and performs smoothing optimization by combining frequency variation trends, adaptive adjustment factors, and confidence boundaries. Finally, it generates frequency control commands that can be recognized by the terminal. These control commands are sent to the field control system via a cellular communication interface to achieve automatic frequency adjustment of the operating equipment.

[0192] To verify the effectiveness of the technical solution of this invention, the entire month of October 2024 was selected as the trial operation period. Two independent operating units were configured at the wastewater treatment plant. One unit served as the experimental group, deploying the system of this invention, while the other unit served as the control group, continuing to use the original manually set operating strategy. Key indicators such as energy consumption level, photovoltaic power generation utilization rate, operating frequency adjustment response, treatment efficiency, and dependence on mains power were recorded for both groups during this period. Some data are summarized below:

[0193] Table 1. Comparison of system operation data under different control methods (October 2024)

[0194] Indicator Item Control group (traditional manual setting) Experimental group (system of this invention) Average daily equipment operating frequency (Hz) 43.6 42.3 Daily operating frequency adjustment times 1.7 4.5 Average daily utilization rate of photovoltaic power generation (%) 78.9 91.2 Percentage of electricity consumption from mains power (%) 36.5 24.7 Energy consumption prediction confidence interval hit rate (%) not applicable 94.6 Average daily water processing capacity (tons) 5770 5930 Overload operation period (per minute / day) 41 9 Number of alarms (related to abnormal energy consumption) 5 times 1 time

[0195] The data shows that the experimental group had a slightly lower daily operating frequency than the control group, but the operating frequency was adjusted more frequently. This indicates that the system of this invention can respond more flexibly to dynamic changes in sunlight and load, reducing energy waste and equipment fatigue. Meanwhile, in terms of photovoltaic resource utilization, the experimental group achieved a daily utilization rate of 91.2%, which is 12.3 percentage points higher than the control group, effectively reducing dependence on mains power and lowering operating costs.

[0196] Regarding mains power consumption, the proportion of mains power in the experimental group decreased to 24.7%, a decrease of 11.8 percentage points compared to the control group. Combined with the energy consumption prediction confidence interval hit rate as high as 94.6%, it indicates that the system prediction model has high accuracy, can detect possible load fluctuations in advance and make reasonable adjustments, thereby significantly reducing the overload operation time to only 9 minutes per day, far lower than the 41 minutes in the control group.

[0197] The system also showed improved processing capacity during operation. The experimental group processed an average of 160 tons more water per day than the control group, indicating that more reasonable frequency regulation helps optimize the operating load and improve overall processing efficiency. At the same time, the number of energy consumption anomaly alarms decreased significantly, and the equipment operated more stably.

[0198] As can be seen from this embodiment, the system provided by the present invention, through the organic coordination of four major modules—illuminance modeling, power generation prediction, energy consumption analysis, and frequency optimization control—establishes a closed-loop link between data acquisition, modeling analysis, and control execution. In actual operation, it effectively alleviates problems such as response lag, inefficient regulation, and energy waste existing in traditional systems, thereby improving the overall operating efficiency and intelligence level of the system.

[0199] It is worth noting that the experimental group did not significantly outperform the control group in all time periods. During periods of continuous overcast and rainy weather, although the solar radiation modeling and prediction still functioned normally, the gap between the two groups narrowed as photovoltaic output approached its limit, indicating that the optimization effect of this system still depends on certain renewable resource conditions. However, overall, the system of this invention demonstrates high practical value in the context of fluctuating solar energy resources.

[0200] In summary, this embodiment demonstrates the application value of the present invention in actual wastewater treatment scenarios, clearly shows its technical features and beneficial effects, and has good scalability and engineering adaptability.

[0201] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A wastewater treatment remote monitoring system based on the Internet of Things, characterized in that, include: Data acquisition module: used to collect data on the direction of sunlight incidence, solar radiation intensity, and the operating status of wastewater treatment equipment, and to attach a unified time label to all types of data; Data transmission module: used to transmit collected data to a remote monitoring platform in real time via narrowband IoT, cellular networks or industrial wireless communication. Illumination Modeling Module: Used to perform directional modeling of solar illumination data based on the Von Mises–Fisher distribution, generating the concentration and directional parameters of direct and diffuse components; Power generation capacity prediction module: used to combine the results of illumination modeling with solar radiation data to calculate the time-series predicted value of photovoltaic power generation capacity; Energy consumption prediction module: Used to input power generation prediction values ​​and equipment operating status data, and output energy consumption prediction results and confidence intervals through an improved spherical regression algorithm; Frequency optimization module: used to compare power generation and energy consumption data, calculate operating frequency parameters, and perform adaptive adjustments based on the predicted confidence interval; Remote control module: Used to generate time-stamped operating frequency adjustment commands, which are sent to the equipment terminal through the communication interface to control the dynamic operation of the sewage treatment system.

2. A remote monitoring method for wastewater treatment based on the Internet of Things, characterized in that, Includes the following steps: S1. Collect data on the incident direction of sunlight, solar radiation intensity, and the operating status of wastewater treatment equipment, add time tags to each data point, and transmit them to a remote monitoring platform via IoT communication. S2. Convert the solar illumination incident direction data into a spherical vector sequence, and use the VonMises–Fisher distribution for modeling to obtain the concentration parameter and mean direction parameter corresponding to each time label, which are used as the illumination modeling results. S3. Based on the illumination modeling results and solar radiation intensity data, generate photovoltaic power generation capacity prediction data for each time tag; S4. Input the photovoltaic power generation capacity prediction data and the sewage treatment equipment operation status data into the improved spherical regression algorithm, and output the energy consumption prediction data corresponding to each time tag; S5. Compare the photovoltaic power generation capacity prediction data with the energy consumption prediction data, calculate the operating frequency of the sewage treatment equipment under each time tag, and generate the initial parameter sequence of the operating frequency. S6. Optimize the initial parameter sequence of the operating frequency based on the illumination modeling results and energy consumption prediction data to generate a set of operating frequency adjustment parameters; S7. Outputs remote control commands containing time stamps, operating frequency adjustment parameters, and energy consumption prediction data, serving as the control basis for the dynamic operation of wastewater treatment equipment.

3. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, S1 specifically includes: S11. Collect data on the direction of solar illumination and the intensity of solar radiation at the wastewater treatment site. The data on the direction of solar illumination consists of the azimuth angle and the elevation angle. The data on the intensity of solar radiation represents the light energy radiation density per unit time. S12. Collect operating status data of sewage treatment equipment. The operating status data includes operating frequency, current value, input power and running time. Each set of data is simultaneously tagged with a time label during collection. S13. Encapsulate the data on the incident direction of sunlight, the data on solar radiation intensity, and the data on the operating status of wastewater treatment equipment into data units according to a unified field structure. Each data unit contains a time tag, a data type identifier, and a corresponding numerical content. S14. The data unit is sent to the remote monitoring platform in real time through Internet of Things (IoT) communication, including a two-way data link based on cellular communication, narrowband IoT, or industrial wireless communication. S15. The remote monitoring platform includes a communication interface for receiving transmitted data units, a data processing structure for parsing, sorting, and verifying received data according to time tags, a data organization structure for constructing a time series data set after parsing, and a data storage structure for the constructed time series data set.

4. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, S2 specifically includes: S21. Based on the sunlight incident direction data collected for each time tag, extract the azimuth and elevation angle information, convert them into unit spherical vectors, and construct a direction vector sequence with time tags: x(t) = [cosβ(t)cosα(t),cosβ(t)sinα(t),sinβ(t)]; where α(t) is the azimuth angle, β(t) is the elevation angle, and x(t) is the direction vector of the corresponding time tag. S22. Divide the sequence of direction vectors within a continuous time period into two subsets according to their rate of change: the first subset is the set of direct components with low rate of change, and the second subset is the set of scattering components with high rate of change. Set a boundary point based on the rate of change threshold and assign each direction vector to the corresponding subset. S23. Construct a first-layer VonMises–Fisher distribution model for the direct sunlight component set, and calculate the directional mean vector. The mean vector is obtained by summing the unit vectors within the direct sunlight component set, superimposing them with the photovoltaic array azimuth vector, and then normalizing the result. The calculation formula is as follows: Where, x i Let b be the i-th unit vector in the direct set. q Let η be the orientation vector of the photovoltaic array, and η be the bias weighting coefficient. S24. Construct a concentration parameter for the direct component based on the direction mean vector and the sample variance. The concentration parameter κ d (t) is calculated using the following formula: Where σ(t) is the angular variance of the unit vector relative to the mean direction, γ is the adjustment coefficient, and ∈ is a positive constant to avoid division by zero; S25. Establish a second-layer Von Mises–Fisher distribution model for the scattering component set, setting the mean direction vector to a fixed direction or a zero vector, and setting the concentration parameter to a constant or determined by κ. d (t) Inverse proportional derivation; S26. Based on the ratio of total solar radiation intensity to diffuse radiation intensity obtained by the sensor, calculate the direct radiation ratio under each time tag, define it as λ(t), and use it as the distribution mixing weight; S27. Using λ(t) as a weighting factor, the direct and scattering distributions are linearly combined to form a hierarchical mixed modeling result. The modeling result includes: λ(t), κ d (t), κ s (t), μ d (t), μ s (t) constitutes the illumination modeling results for each time label.

5. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, S3 specifically includes: S31. Extract the illumination modeling results corresponding to each time tag. The illumination modeling results include the concentration parameter and mean direction vector of the direct component, the concentration parameter and mean direction vector of the scattered component, and the weighting factor of the direct component. S32. Extract the solar radiation intensity data corresponding to each time tag. The solar radiation intensity data includes the total radiation intensity value, the direct radiation intensity value, and the diffuse radiation intensity value. S33. Construct an input feature vector from the illumination modeling parameters and solar radiation intensity data corresponding to each time tag. The feature vector includes: direct radiation concentration parameter, direct radiation mean direction vector component, scattering concentration parameter, scattering mean direction vector component, direct radiation weight, direct radiation intensity and scattered radiation intensity. S34. Construct a training sample set based on time labels. The training sample set contains historical illumination modeling parameters, radiation intensity data and measured power generation data for the corresponding time. S35. Construct a regression neural network model using the training sample set. The neural network is a multi-layer feedforward structure. Each layer uses a combination of linear transformation and nonlinear activation. The loss function is the mean square error between the predicted power generation and the measured value. S36. Input the input feature vector corresponding to each time tag into the trained neural network model, and output the photovoltaic power generation prediction value corresponding to each time tag. All prediction values ​​constitute photovoltaic power generation capacity prediction data arranged in chronological order.

6. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, S4 specifically includes: S41. Based on each time tag, extract the photovoltaic power generation capacity prediction data and the sewage treatment equipment operation status data, and construct an input feature sequence. The input feature sequence includes the photovoltaic power generation power prediction value, equipment operating frequency, current, voltage, energy consumption direction change value of the previous time tag, and power difference. S42. Construct an improved spherical regression algorithm, which includes a feature injection layer, an illumination weight gating layer, a spherical projection layer, and a confidence interval estimation module: S421. In the feature injection layer, perform affine transformation and nonlinear mapping on the input features to obtain the hidden vector h(t); S422. In the illumination weight gating layer, the direct component weight λ(t) from the illumination modeling result corresponding to the time label is introduced to construct a dual-channel gating mechanism for split calculation of the hidden vector: Among them, W d With W s These are the weight matrices for the direct and scattering channels, respectively; S423. In the spherical projection layer, the gated output vector Perform an affine mapping and normalize to a unit sphere to obtain the energy dissipation direction prediction vector: Among them, W p with b p These are the parameters of the spherical mapping layer; S424. In the confidence interval estimation module, based on the magnitude of the hidden vector... Construct the spherical cap angle θ c (t) = arcsinr(t), with Construct a spherical cap sampling set {v} containing M directional samples, centered at the cap. j (t)}; S425. Perform the inner product of each sampling direction vector and the predefined energy consumption reference vector e to calculate the set of predicted energy consumption values: And calculate the upper and lower boundaries E of the confidence interval for energy consumption prediction. low (t),E high (t), and the central prediction value S43. Output the energy consumption direction prediction vector for each time tag. Energy consumption center forecast value E mid (t), along with the confidence interval boundary values, constitute the energy consumption prediction data.

7. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, The improved spherical regression algorithm specifically includes: The feature injection layer includes: an initial input layer, an affine transformation layer, and a nonlinear activation layer; The illumination weighting gate layer includes: direct component weighting input; and a dual-channel splitting mechanism to generate direct channel and scattering channel outputs. The spherical projection layer includes: an affine mapping layer, a normalization processing layer, and a final energy consumption direction prediction vector. The confidence interval estimation module includes: a modulus calculation unit, a radius calculation unit, a spherical cap sampling unit, and an energy consumption calculation unit.

8. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, S5 specifically includes: S51. Extract the photovoltaic power generation capacity prediction data and energy consumption prediction data corresponding to each time tag, which respectively represent the available power and estimated load power consumption of the sewage treatment plant. S52. Based on each time tag, compare the photovoltaic power generation forecast with the energy consumption forecast to obtain the energy balance difference; when the photovoltaic power generation forecast is greater than the energy consumption forecast, it indicates that there is power redundancy, and when it is less than, it indicates that there is power shortage. S53. Combining the energy balance difference with the operating frequency under the previous time tag, calculate the initial value of the operating frequency corresponding to the current time tag. The calculation rules include: when there is power redundancy, the operating frequency is set to be proportionally increased based on the previous frequency; when there is power shortage, the operating frequency is set to be proportionally decreased based on the previous frequency; when supply and demand are basically balanced, the operating frequency remains unchanged. S54. Set the upper and lower limits of the operating frequency of the sewage treatment equipment and perform boundary correction on the initial value of the operating frequency. S55. Arrange the calculated operating frequencies under all time tags into a vector in chronological order, and use it as the initial parameter sequence for the operating frequencies.

9. The wastewater treatment remote monitoring method based on the Internet of Things according to claim 2, characterized in that, S6 specifically includes: S61. Read the initial parameter sequence of the running frequency, the illumination modeling results and the energy consumption prediction data corresponding to each time tag, and match them one by one according to the time tag; S62. Perform a difference operation on the initial parameter sequence of the operating frequency to obtain the frequency change of adjacent time tags; at the same time, calculate the rate of change of the weight of the direct component in the illumination modeling results and the rate of change of the energy consumption prediction data. S63. Construct an adaptive adjustment coefficient based on the frequency change and the rate of change of illumination weight; when the direct illumination weight increases rapidly and the energy consumption prediction fluctuation decreases, the adaptive adjustment coefficient increases; when the direct illumination weight decreases or the energy consumption fluctuation increases, the adaptive adjustment coefficient decreases. S64. Using an adaptive adjustment coefficient, the initial parameter sequence of the operating frequency is subjected to sliding window weighted smoothing to reduce high-frequency oscillations and retain the trend response; the smoothing result is used as a candidate frequency sequence. S65. Compare the candidate frequency sequence with the energy consumption prediction data point by point. If the candidate frequency causes the energy consumption prediction value to exceed the upper limit of the confidence interval at any time label, the corresponding frequency is adjusted down proportionally. If it causes the energy consumption prediction to fall below the lower limit of the confidence interval, the corresponding frequency is adjusted up proportionally to form the adjusted frequency sequence. S66. Apply threshold restrictions to the adjusted frequency sequence to keep it within the minimum and maximum frequency range allowed by the wastewater treatment equipment, thereby obtaining a set of operating frequency adjustment parameters that meet the requirements of the control interface. S67. Save the set of operating frequency adjustment parameters in time-stamped order and output it to the remote monitoring platform for use in the next step of generating remote control commands.

Images