Digital agricultural management system based on Internet of Things

Abstract

The invention relates to the technical field of agricultural digital management, and discloses a digital agricultural management system based on the Internet of Things, and the system comprises an environment sensing module which is configured to collect real-time environment data inside and outside a target agricultural facility and real-time energy consumption data of regulation and control equipment in the facility, collecting the real-time environment data and the real-time energy consumption data to form historical environment data and historical energy consumption data; and the thermal inertia modeling module is configured to receive the historical environment data and the historical energy consumption data formed by the environment sensing module. A self-adaptive regulation and control engine is adopted, an equipment control instruction is dynamically generated according to real-time environment data and a preset crop growth demand curve, dynamic optimization and accurate execution of a control strategy are achieved, and compared with an existing control scheme adopting a fixed threshold value or a rigid time program, the control scheme has the advantages that the efficiency is high; the problems that a control strategy is rigid and cannot actively adapt to external environment changes and crop growth stage requirements are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital agricultural management, in particular to a digital agricultural management system based on the Internet of Things. BACKGROUND

[0002] The Internet of Things is a new network communication technology emerging in recent years. The Internet of Things uses local network or Internet communication technology to connect sensors, controllers, machines, personnel and objects together in a new way, forming a network that connects people and things, and realizes informatization, remote management and control and intelligentization.

[0003] The Internet of Things in agriculture aims to promote the automation, informatization and intelligentization of agriculture, and has attracted great attention at home and abroad. Some western countries have begun to use special resource satellites to monitor farmland resources in real time, use sensor information fusion technology to monitor agricultural ecology in real time, and use precision fertilization and irrigation systems to improve fertilizer utilization. China uses GPS positioning technology and other wireless sensor network technologies to collect, transmit and manage farmland parameters, and uses ground monitoring stations to build agricultural environment monitoring systems.

[0004] For example, the Chinese patent application with publication number CN117610847A discloses a digital agricultural management system based on the Internet of Things, which includes a processing module, an observation module, an environment module, a crop module, a management platform, a warning module, and an automatic control subsystem. The processing module is used to analyze and process the obtained agricultural planting related information, the observation module is used to obtain and record the agricultural planting related information, the environment module is used to collect and adjust the environmental information, and the crop module is used to manage the information and operation of plants from planting to harvesting.

[0005] For example, the Chinese patent application with publication number CN116306057A discloses an agricultural digital management system and method based on the Internet of Things, which includes a data acquisition module for real-time acquisition of water body parameter information; a data processing module for receiving water body parameter information collected by the data acquisition module, and constructing a breeding pond early warning management model based on the water body parameter information; and an Internet of Things server for real-time monitoring of early warning information of each breeding pond according to the constructed breeding pond early warning management model, and real-time transmission of the early warning information to a mobile monitoring terminal.

[0006] The above patent has the following shortcomings:

[0007] In the agricultural management system, a preset fixed threshold or a rigid time program is usually used to drive the environmental control equipment. This control method cannot adjust itself according to the real-time changes of the external climate and the physiological needs of crops in different growth periods, resulting in rigid control strategy and inability to provide a continuous optimal environment for crop growth.

[0008] The prior art limits the control target to making the environment monitoring parameter reach the preset value when performing the environment adjustment task, and the regulation process does not take the energy consumption as a feedback or optimization index, so that the control system cannot evaluate and constrain the energy consumption generated by the adjustment behavior while achieving the environment standard, resulting in low energy use efficiency and high operation cost.

[0009] The existing greenhouse control system fails to build an internal correlation model between the environment variable and the energy consumption parameter, and therefore does not have the prediction ability of the environment change trend, so that the regulation behavior is always lagging behind the actual change of the environment, cannot intervene in advance, and responds passively when the environment condition deviates from the ideal interval. The lag of the regulation behavior may damage the crops, and the overshoot and fluctuation of the environment parameter may be caused due to the inability to predict the regulation end point, and then trigger the reverse compensation regulation, resulting in repeated start and stop of the equipment and additional energy loss.

[0010] The fixed parameter control mode adopted by the prior art cannot actively adapt to and meet the variability demand of the crops in the whole life cycle, resulting in the inability to provide the optimal growth conditions for the crops. If adjustment is to be made, the system parameters must be frequently reset by manual operation, increasing the management burden.

[0011] Therefore, the present application provides a digital agricultural management system based on the Internet of Things to solve the above problems. SUMMARY

[0012] In view of the deficiencies of the prior art, the present application provides a digital agricultural management system based on the Internet of Things to solve the problems in the background art.

[0013] To achieve the above object, the present application is implemented by the following technical scheme: a digital agricultural management system based on the Internet of Things, comprising:

[0014] An environment perception module configured to collect real-time environment data inside and outside a target agricultural facility and real-time energy consumption data of the regulation equipment inside the facility, and to collect the real-time environment data and the real-time energy consumption data to form historical environment data and historical energy consumption data;

[0015] A thermal inertia modeling module configured to receive the historical environment data and the historical energy consumption data formed by the environment perception module, and to build a thermal response model containing the facility thermodynamic parameters in combination with the preset structure parameters of the target agricultural facility;

[0016] An adaptive regulation engine configured to receive the real-time environment data collected by the environment perception module, to call the thermal response model built by the thermal inertia modeling module, and to generate a device control instruction containing the prediction result of the operation of the regulation equipment based on a preset crop growth demand curve;

[0017] An energy-saving feedback module is configured to, after the execution of the device control instruction, obtain the actual environment data and the actual energy consumption data collected by the environment perception module, compare the data collected by the environment perception module with the predicted results contained in the device control instruction, generate a model calibration signal fed back to the thermal inertia modeling module and a strategy weight adjustment signal fed back to the adaptive control engine according to the deviation generated by the comparison.

[0018] Preferably, the environment perception module comprises an edge computing gateway deployed on site, which is responsible for the preliminary cleaning and formatting of the raw data collected by the sensor, and then uploading the structured data.

[0019] Preferably, the thermal response model is a mathematical model taking the equivalent heat capacity and the equivalent total heat exchange coefficient of the target agricultural facility as core variables, for quantifying the delay effect and the change rate of the internal environment temperature of the facility caused by the predictive control behavior.

[0020] Preferably, the crop growth demand curve is a data set storing the optimal environment parameter range corresponding to different growth stages of the crop, and the adaptive control engine calls the corresponding optimal environment parameter range as the control target according to the current growth stage of the crop.

[0021] Preferably, the adaptive control engine generates the device control instruction by simulating and evaluating multiple alternative control strategies through a comprehensive cost function, and selects the alternative control strategy that minimizes the value of the comprehensive cost function as the optimal device control instruction.

[0022] The comprehensive cost function is a function that quantifies the deviation of the environment parameter from the optimal target interval and the predicted energy consumption caused by the execution of the alternative control strategy by means of weighted summation.

[0023] Preferably, the model calibration signal generated by the energy-saving feedback module is used to adjust the values of the equivalent heat capacity and the equivalent total heat exchange coefficient in the thermal response model.

[0024] The strategy weight adjustment signal generated by the energy-saving feedback module is used to adjust the weight coefficients of the deviation of the environment parameter from the optimal target interval and the predicted energy consumption in the comprehensive cost function.

[0025] Preferably, the environment perception module further comprises:

[0026] A raw data collection unit is configured to be deployed at the sensors and energy consumption monitoring devices inside and outside the target agricultural facility to periodically obtain raw data points containing measurement values.

[0027] The data validity discrimination unit is configured to receive the raw data points acquired by the raw data acquisition unit, filter the raw data points through a validity determination condition, discard the raw data points that do not meet the validity determination condition, and confirm the raw data points that meet the condition as valid data points, wherein the validity determination condition is:

[0028] The measurement value of the raw data point x i must fall within a dynamic threshold interval calculated from the data in the latest time window, and the calculation formula of the dynamic threshold interval is:

[0029] [μ-kσ, μ+kσ], wherein x i is the measurement value of the current raw data point, μ is the arithmetic mean of the measurement values of all raw data points in the latest time window, σ is the standard deviation of the measurement values of all raw data points in the latest time window, and k is a preset confidence coefficient greater than zero.

[0030] The data smoothing filter unit is configured to receive the valid data points confirmed by the data validity discrimination unit, calculate a string of continuous valid data points by using a moving average algorithm to generate a smoothed filter value, and the calculation formula of the moving average algorithm is:

[0031] wherein, is the smoothed filter value generated at time t, N is the number of valid data points contained in the moving average window, x t-i is the measurement value of the valid data point at time t-i; and

[0032] The structured packaging unit is configured to receive the smoothed filter values generated by the data smoothing filter unit, combine each smoothed filter value with a unique sensor identifier and an acquisition time stamp to generate a structured data record.

[0033] The data collection and uploading unit is configured to combine multiple structured data records generated by the structured packaging unit into a data packet, and upload the data packet.

[0034] Preferably, the thermal inertia modeling module further comprises:

[0035] The data correlation aggregation unit is configured to receive the data packet uploaded by the data collection and uploading unit in the environment perception module, deconstruct the data packet into multiple structured data records, align the data of different sources according to the time stamps in the structured data records to generate aggregated time series data containing internal temperature, external temperature, solar radiation intensity and equipment thermal power.

[0036] A model parameter identification unit configured to receive the aggregated time series data generated by the data correlation aggregation unit, and to obtain equivalent heat capacity and equivalent total heat exchange coefficient representing thermodynamic characteristics of the facility by solving an energy balance equation of the facility based on a system identification algorithm, the energy balance equation being specifically in the form of:

[0037]

[0038] wherein C eq is the equivalent heat capacity to be solved, K total is the equivalent total heat exchange coefficient to be solved, ΔT in is a change of the internal temperature of the facility in a sampling time interval, T in (t) is the internal temperature of the facility at time t, T out (t) is the external temperature of the facility at time t, Δt is a time interval of data sampling, P dev (t) is the heat power generated by the regulating device at time t, G(t) is the solar radiation intensity at time t, and A is a preset effective receiving area of solar radiation.

[0039] A model verification and update triggering unit configured to receive the equivalent heat capacity and the equivalent total heat exchange coefficient solved by the model parameter identification unit, to calculate a mean square error representing model accuracy by using a set of independent verification data, and to determine according to a model update determination condition.

[0040] If the model update determination condition is established, a model update instruction is generated to trigger re-computation of the model parameter identification unit.

[0041] If the model update determination condition is not established, the equivalent heat capacity and the equivalent total heat exchange coefficient are output.

[0042] The model update determination condition is MSE>ε max , wherein MSE is the mean square error, and the calculation formula is:

[0043]

[0044] , wherein M is the number of data points in the verification data set, T pred,j is the predicted temperature of the model for the jth data point, T actual,j is the actual temperature of the jth data point, ε max is a preset maximum allowed mean square error threshold.

[0045] The response prediction function generation unit is configured to receive the equivalent heat capacity and the equivalent overall heat exchange coefficient output by the model verification and update trigger unit when the model update determination condition is not met, so as to construct the thermal response model, which is used to predict the change trajectory of the internal environment temperature of the facility over time under given external environment conditions and device control inputs.

[0046] Preferably, in the adaptive regulation engine, further comprising:

[0047] The regulation target acquisition unit is configured to receive the thermal response model constructed by the response prediction function generation unit in the thermal inertia modeling module, and receive the real-time environment data collected by the environment perception module, and then call the optimal environment parameter range corresponding to the current growth stage of the crops in the crop growth demand curve;

[0048] The alternative strategy generation unit is configured to receive the optimal environment parameter range called by the regulation target acquisition unit and the real-time environment data acquired, and first perform a regulation trigger determination. When the real-time environment data is not within the optimal environment parameter range, generate an alternative control strategy set containing multiple different device operation combinations according to all available regulation devices and settable operating states in the system, and the alternative control strategy set must contain a zero strategy of not performing any operation.

[0049] The strategy deduction and cost quantification unit is configured to receive the thermal response model acquired by the regulation target acquisition unit, and traverse each alternative control strategy in the alternative control strategy set generated by the alternative strategy generation unit, deduce the environment parameter change trajectory and predicted energy consumption of each alternative control strategy within a preset time in the future by calling the thermal response model, and then calculate the comprehensive cost function value corresponding to each alternative control strategy by a comprehensive cost function, and the specific form of the comprehensive cost function is:

[0050] J = w env · D env + w energy · C energy ,

[0051] wherein J is the comprehensive cost function value, w env and w energy are weight coefficients contained in the strategy weight adjustment signal received by the energy saving feedback module, C energy is the predicted energy consumption, D env is the deviation degree of the environment parameter from the optimal environment parameter range, and the calculation formula is:

[0052]

[0053] Wherein, H is a preset deduction time length, T pred (t) is a predicted temperature of the environmental parameter change trajectory at time t, T target is a median temperature of the optimal environmental parameter range;

[0054] An optimal instruction decision unit is configured to receive the comprehensive cost function values of all candidate control strategies calculated by the strategy deduction and cost quantification unit, determine a final control strategy according to an optimal strategy determination condition, and encapsulate the final control strategy into the device control instruction, wherein the optimal strategy determination condition is:

[0055] The candidate control strategy corresponding to the comprehensive cost function value with the minimum value is determined as the final control strategy.

[0056] Preferably, the energy-saving feedback module further comprises:

[0057] An execution result acquisition unit is configured to receive the device control instruction encapsulated and output by the optimal instruction decision unit in the adaptive regulation and control engine, extract the prediction result from the device control instruction, and acquire actual environmental data and actual energy consumption data collected by the environmental perception module within a preset time period during and after execution of the device control instruction.

[0058] A deviation quantification calculation unit is configured to receive the prediction result extracted by the execution result acquisition unit and the acquired actual environmental data and actual energy consumption data, calculate a model prediction error value and an energy consumption prediction error value through a model prediction error formula and an energy consumption prediction error formula, wherein the model prediction error formula is:

[0059]

[0060] Wherein, E model is a model prediction error value, T is a length of instruction execution, T pred (t) is a predicted temperature at time t in the prediction result, T actual (t) is an actual temperature at time t in the actual environmental data;

[0061] The energy consumption prediction error formula is:

[0062] E energy = C actual -C pred ,

[0063] Wherein, E energy is an energy consumption prediction error value, C actual is total energy consumption in the actual energy consumption data, and C predto predict the total predicted energy consumption in the result;

[0064] a feedback trigger determination unit configured to receive the model prediction error value and the energy consumption prediction error value calculated by the bias quantification calculation unit, and determine by a model calibration determination condition and a strategy adjustment determination condition;

[0065] if the model calibration determination condition is met, a model update trigger flag is generated;

[0066] if the strategy adjustment determination condition is met, a weight adjustment trigger flag is generated;

[0067] the model calibration determination condition is: E model > δ model ,

[0068] wherein, δ model is a preset model error threshold;

[0069] the strategy adjustment determination condition is: |E energy | > δ energy ,

[0070] wherein, δ energy is a preset energy consumption error threshold;

[0071] a feedback signal generation unit configured to receive the model update trigger flag and the weight adjustment trigger flag generated by the feedback trigger determination unit;

[0072] when the model update trigger flag is received, the model calibration signal is generated;

[0073] when the weight adjustment trigger flag is received, a new set of weight coefficients is calculated by a weight update algorithm and encapsulated into the strategy weight adjustment signal, and the specific form of the weight update algorithm is:

[0074] w ′ env = w env + α (E model - β · E energy ),

[0075] wherein, w ′ env is the updated environmental bias degree weight coefficient, w env is the weight coefficient before updating, E model is the model prediction error value, E energy is the energy consumption prediction error value, α is a preset learning rate, and β is a preset balance factor.

[0076] The application provides a digital agricultural management system based on the Internet of Things.

[0077] 1. The application adopts an adaptive regulation engine to dynamically generate device control instructions according to real-time environmental data and a preset crop growth demand curve, realizes dynamic optimization and accurate execution of the control strategy, and solves the problems of rigid control strategy, inability to actively adapt to changes in the external environment and crop growth stage requirements, compared with the existing control scheme using fixed threshold or rigid time program.

[0078] 2. The application adopts a technical scheme in which the adaptive regulation engine built-in comprehensive cost function quantifies the environmental deviation degree and the predicted energy consumption when making decisions, achieves the technical effect of minimizing the energy consumption of the regulation process while meeting the crop growth environment requirements, and solves the problem of high operating cost caused by the lack of energy consumption evaluation and constraints in the regulation process, compared with the technical scheme in the prior art in which the control target is only to meet the environmental parameter standard.

[0079] 3. The application adopts a thermal inertia modeling module to construct a thermal response model, which can quantify the internal correlation between the control behavior and environmental changes, and is called by the adaptive regulation engine to predict the control consequences, achieves the effect of early intervention and accurate regulation, and solves the problem of delayed control behavior, easy to cause environmental parameter overshoot and fluctuation and cause additional energy loss, compared with the existing technical scheme which lacks prediction ability and can only passively respond to environmental changes.

[0080] 4. The application adopts an energy-saving feedback module to compare the predicted and actual results after the execution of the control instructions, generates model calibration and strategy weight adjustment signals, and feeds them back to the upstream module, achieves the effect of system self-correction and continuous optimization, and solves the problem of the system being unable to automatically adapt to the dynamic requirements of the entire life cycle of crops and requiring frequent manual intervention to set parameters, compared with the existing fixed parameter control scheme. BRIEF DESCRIPTION OF DRAWINGS

[0081] Figure 1 The figure is a system framework diagram of the application. DETAILED DESCRIPTION

[0082] To enable those skilled in the art to understand the application scheme, the technical solutions in the embodiments of the application will be described in detail below with reference to the drawings of the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, other embodiments obtained by those skilled in the art without creative labor should be within the scope of the application.

[0083] The application will be described in detail below with reference to the drawings:

[0084] Embodiment:

[0085] Please refer to the attached Figure 1 The embodiment of the present application provides a digital agricultural management system based on Internet of Things, which comprises:

[0086] An environment perception module is configured to collect real-time environment data inside and outside a target agricultural facility and real-time energy consumption data of a regulating device inside the facility, and to collect the real-time environment data and the real-time energy consumption data to form historical environment data and historical energy consumption data;

[0087] The environment perception module comprises an edge computing gateway deployed on site, which is responsible for preliminary cleaning and formatting of raw data collected by sensors, and then uploading structured data;

[0088] The environment perception module further comprises:

[0089] A raw data acquisition unit is configured to be deployed at sensors and energy consumption monitoring devices inside and outside the target agricultural facility, and to periodically acquire raw data points containing measurement values;

[0090] A data validity discrimination unit is configured to receive raw data points acquired by the raw data acquisition unit, and to filter the raw data points through validity determination conditions, wherein raw data points that do not meet the validity determination conditions are discarded, and raw data points that meet the conditions are confirmed as valid data points, and the validity determination conditions are:

[0091] The measurement value of the raw data point x i must fall within a dynamic threshold interval calculated from data in the nearest time window, and the calculation formula of the dynamic threshold interval is:

[0092] [μ-kσ, μ+kσ], wherein x i is the measurement value of the current raw data point, μ is the arithmetic mean of the measurement values of all raw data points in the nearest time window, σ is the standard deviation of the measurement values of all raw data points in the nearest time window, and k is a preset confidence coefficient greater than zero;

[0093] A data smoothing filter unit is configured to receive valid data points confirmed by the data validity discrimination unit, and to calculate a string of continuous valid data points using a moving average algorithm to generate a smoothed filter value, and the calculation formula of the moving average algorithm is:

[0094] wherein, is the smoothed filter value generated at time t, N is the number of valid data points contained in the moving average window, and x t-i is the measurement value of the valid data point at time t-i;

[0095] a structured packaging unit configured to receive the smoothed filtering values generated by the data smoothing filtering unit, combine each smoothed filtering value with a unique sensor identifier and a collection timestamp to generate a structured data record;

[0096] a data collection uploading unit configured to combine a plurality of structured data records generated by the structured packaging unit into a data package, and upload the data package;

[0097] a thermal inertia modeling module configured to receive historical environmental data and historical energy consumption data formed by the environmental perception module, and construct a thermal response model containing facility thermodynamic parameters in combination with preset structural parameters of a target agricultural facility;

[0098] The thermal response model takes the equivalent heat capacity and the equivalent total heat exchange coefficient of the target agricultural facility as core variables, and is a mathematical model for quantifying and predicting the delay effect and change rate of the internal environmental temperature of the facility caused by the control behavior;

[0099] The thermal inertia modeling module further comprises:

[0100] a data correlation aggregation unit configured to receive the data package uploaded by the data collection uploading unit in the environmental perception module, deconstruct the data package into a plurality of structured data records, and align the data from different sources according to the timestamps in the structured data records to generate aggregated time series data containing the internal temperature of the facility, the external temperature, the solar radiation intensity, and the device heat power;

[0101] a model parameter identification unit configured to receive the aggregated time series data generated by the data correlation aggregation unit, and obtain the equivalent heat capacity and the equivalent total heat exchange coefficient representing the thermodynamic characteristics of the facility by solving the energy balance equation of the facility through a system identification algorithm, the specific form of the energy balance equation being:

[0102]

[0103] wherein, C eq is the equivalent heat capacity to be solved, K total is the equivalent total heat exchange coefficient to be solved, ΔT in is the change of the internal temperature of the facility within a sampling time interval, T in (t) is the internal temperature of the facility at time t, T out (t) is the external temperature of the facility at time t, Δt is the data sampling time interval, P dev (t) is the heat power generated by the control device at time t, G(t) is the solar radiation intensity at time t, and A is the preset effective solar radiation receiving area;

[0104] The model verification and update triggering unit is configured to receive the equivalent heat capacity and the equivalent total heat exchange coefficient solved by the model parameter identification unit, calculate the mean square error (MSE) representing the accuracy of the model by using a set of independent verification data, and determine whether the model update condition is met or not;

[0105] If the model update condition is met, a model update instruction is generated to trigger the model parameter identification unit to recalculate;

[0106] If the model update condition is not met, the equivalent heat capacity and the equivalent total heat exchange coefficient are outputted;

[0107] The model update condition is MSE>∈ max , wherein the MSE is the mean square error, and the calculation formula is:

[0108]

[0109] , wherein M is the number of data points in the verification data set, T pred,j is the predicted temperature of the model for the jth data point, T actual,j is the actual temperature of the jth data point, and ∈ max is the preset maximum allowed mean square error threshold;

[0110] The response prediction function generation unit is configured to receive the equivalent heat capacity and the equivalent total heat exchange coefficient outputted by the model verification and update triggering unit when the model update condition is not met, so as to construct a thermal response model, which is used to predict the change trajectory of the internal environment temperature of the facility over time under given external environmental conditions and device control inputs;

[0111] The adaptive control engine is configured to receive the real-time environmental data collected by the environmental perception module, call the thermal response model constructed by the thermal inertia modeling module, and generate a device control instruction containing the predicted results of the operation of the control device based on a preset crop growth demand curve;

[0112] The crop growth demand curve is a data set storing the optimal environmental parameter ranges corresponding to different growth stages of crops, and the adaptive control engine calls the corresponding optimal environmental parameter range as the control target according to the current growth stage of the crops;

[0113] The adaptive control engine further comprises:

[0114] The control target acquisition unit is configured to receive the thermal response model constructed by the response prediction function generation unit in the thermal inertia modeling module, and receive the real-time environmental data collected by the environmental perception module, and then call the optimal environmental parameter range corresponding to the current growth stage of the crops from the crop growth demand curve;

[0115] An alternative strategy generation unit configured to receive the optimal environment parameter range called by the regulation target acquisition unit and the acquired real-time environment data, first perform a regulation trigger determination, and when the real-time environment data is not within the optimal environment parameter range, generate an alternative control strategy set containing multiple different device operation combinations according to all available regulation devices and settable operating states in the system, the alternative control strategy set necessarily containing a zero strategy of not performing any operation;

[0116] A strategy deduction and cost quantification unit configured to receive the thermal response model acquired by the regulation target acquisition unit, and traverse each alternative control strategy in the alternative control strategy set generated by the alternative strategy generation unit, deduce the environmental parameter change trajectory and predicted energy consumption of each alternative control strategy within a preset time in the future by calling the thermal response model, and then calculate the comprehensive cost function value corresponding to each alternative control strategy by a comprehensive cost function, the specific form of the comprehensive cost function being:

[0117] J = w env ·D env +w energy ·C energy ,

[0118] wherein J is the comprehensive cost function value, w env and w energy are weight coefficients contained in the strategy weight adjustment signal received by the energy-saving feedback module, C energy is the predicted energy consumption, D env is the deviation degree of the environmental parameter from the optimal environment parameter range, and the calculation formula is:

[0119]

[0120] wherein H is a preset deduction time length, T pred (t) is the predicted temperature of the environmental parameter change trajectory at time t, and T target is the median temperature of the optimal environment parameter range;

[0121] An optimal instruction decision unit configured to receive the comprehensive cost function values of all alternative control strategies calculated by the strategy deduction and cost quantification unit, determine a final control strategy according to an optimal strategy determination condition, and then encapsulate the final control strategy into a device control instruction, the optimal strategy determination condition being:

[0122] selecting the value minimum from all comprehensive cost function values, and determining the alternative control strategy corresponding to the value minimum comprehensive cost function value as the final control strategy;

[0123] The energy-saving feedback module is configured to obtain actual environment data and actual energy consumption data collected by the environment perception module after execution of the device control instruction, compare the data collected by the environment perception module with the predicted results contained in the device control instruction, generate a model calibration signal fed back to the thermal inertia modeling module and a strategy weight adjustment signal fed back to the adaptive regulation engine according to deviations generated by the comparison;

[0124] The adaptive regulation engine generates the device control instruction in the following manner: simulating and evaluating a plurality of candidate control strategies through the comprehensive cost function, and selecting a candidate control strategy that minimizes the value of the comprehensive cost function as the optimal device control instruction;

[0125] The comprehensive cost function is a function that quantifies deviations of environment parameters from optimal target intervals and predicted energy consumption generated by executing the candidate control strategy in a weighted summation manner;

[0126] The model calibration signal generated by the energy-saving feedback module is used to adjust the values of the equivalent heat capacity and the equivalent total heat exchange coefficient in the thermal response model;

[0127] The strategy weight adjustment signal generated by the energy-saving feedback module is used to adjust the weight coefficients of the deviations of the environment parameters from the optimal target intervals and the predicted energy consumption in the comprehensive cost function.

[0128] The energy-saving feedback module further comprises:

[0129] The execution result acquisition unit is configured to receive the device control instruction packaged and output by the optimal instruction decision unit in the adaptive regulation engine, extract the predicted results from the device control instruction, and obtain actual environment data and actual energy consumption data collected by the environment perception module during and after execution of the device control instruction for a preset time period;

[0130] The deviation quantification calculation unit is configured to receive the predicted results extracted by the execution result acquisition unit and the obtained actual environment data and actual energy consumption data, calculate model prediction error values and energy consumption prediction error values through a model prediction error formula and an energy consumption prediction error formula, the model prediction error formula being:

[0131]

[0132] E model is the model prediction error value, T is the duration of instruction execution, T pred (t) is the predicted temperature at time t in the predicted results, T actual (t) is the actual temperature at time t in the actual environment data;

[0133] The energy consumption prediction error formula is:

[0134] Eenergy = C actual - C pred ,

[0135] wherein, E energy is the energy consumption prediction error value, C actual is the total energy consumption in the actual energy consumption data, C pred is the total predicted energy consumption in the prediction result;

[0136] a feedback trigger determination unit configured to receive the model prediction error value and the energy consumption prediction error value calculated by the bias quantification calculation unit, and determine by a model calibration determination condition and a strategy adjustment determination condition;

[0137] if the model calibration determination condition is met, a model update trigger flag is generated;

[0138] if the strategy adjustment determination condition is met, a weight adjustment trigger flag is generated;

[0139] the model calibration determination condition is: E model > δ model ,

[0140] wherein, δ model is a preset model error threshold;

[0141] the strategy adjustment determination condition is: |E energy | > δ energy ,

[0142] wherein, δ energy is a preset energy consumption error threshold;

[0143] a feedback signal generation unit configured to receive the model update trigger flag and the weight adjustment trigger flag generated by the feedback trigger determination unit;

[0144] when the model update trigger flag is received, a model calibration signal is generated;

[0145] when the weight adjustment trigger flag is received, a new set of weight coefficients is calculated by a weight update algorithm and encapsulated into a strategy weight adjustment signal, and the specific form of the weight update algorithm is:

[0146] w ′ env = w env + α (E model - β · E energy ),

[0147] wherein, w ′ env is the updated environmental bias degree weight coefficient, w env is the environmental bias degree weight coefficient before updating, E modelE is a model prediction error value energy E is an energy consumption prediction error value, a is a preset learning rate, and β is a preset balance factor.

[0148] The environment perception module provides a highly reliable and standardized data basis for the digital agricultural management system through the built-in full-process data processing unit from raw data collection to structured packaging. The module is a simple information collection terminal and the first pass of data quality. Through dynamic threshold effectiveness discrimination and moving average smoothing filtering algorithms, the module converts the original signals full of noise and abnormalities in the field environment into structured data streams that are accurate, stable, and can be directly used for advanced analysis, fundamentally ensuring the accuracy of subsequent model construction and the effectiveness of decision control, and laying a solid data cornerstone for realizing precision agricultural management.

[0149] The thermal inertia modeling module gives the digital agricultural management system the ability to deeply understand the physical characteristics of specific agricultural facilities through its built-in data correlation, parameter identification, and model verification unit. The module refines and sublimates discrete historical data into physical models that can describe the dynamic response law of heat, enabling the system to upgrade from perception to prediction. By constructing an energy balance equation containing equivalent heat capacity and equivalent overall heat transfer coefficient, the module provides the system with the ability to deduce the environmental changes caused by different control strategies, which is the core of realizing forward-looking and preventive regulation and completely changes the mode of traditional control systems with lagging response.

[0150] The adaptive control engine constitutes the wisdom decision core of the digital agricultural management system through its built-in control target acquisition, alternative strategy generation, and optimal instruction decision unit. The module can consider the dynamic needs of crops at different life cycles, the thermal inertia prediction model of the facility, and the energy consumption cost as a whole, use a comprehensive cost function to quickly deduce and quantitatively evaluate a large number of alternative strategies, and finally select a unique control instruction that takes into account the best growth conditions for crops and the lowest operating cost. Each adjustment behavior is a precise decision made from a global optimal perspective, fundamentally improving resource utilization efficiency and agricultural output efficiency.

[0151] The energy-saving feedback module injects the vitality of self-evolution and continuous optimization into the digital agricultural management system through the built-in execution result acquisition, deviation quantification calculation, and feedback signal generation unit. After each control cycle, the module quantitatively evaluates the accuracy of the current model and strategy by rigorously comparing the differences between prediction and reality, and can automatically generate model calibration signals and strategy weight adjustment signals, forming a closed-loop learning mechanism to ensure that the system can automatically adapt to environmental seasonal changes, facility aging, and even crop type replacement, achieving intelligent and adaptive management.

[0152] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A digital agricultural management system based on the Internet of Things, characterized in that, The digital agricultural management system includes: The environmental sensing module is configured to collect real-time environmental data inside and outside the target agricultural facility, as well as real-time energy consumption data of the control equipment inside the facility, and to aggregate the real-time environmental data and real-time energy consumption data to form historical environmental data and historical energy consumption data. The thermal inertia modeling module is configured to receive historical environmental data and historical energy consumption data generated by the environmental perception module, and combine them with the pre-set structural parameters of the target agricultural facility to construct a thermal response model that includes the facility's thermodynamic parameters. The adaptive control engine is configured to receive real-time environmental data collected by the environmental perception module, call the thermal response model constructed by the thermal inertia modeling module, and generate equipment control commands containing the predicted results of the control equipment operation based on the preset crop growth demand curve. The energy-saving feedback module is configured to acquire the actual environmental data and actual energy consumption data collected by the environmental sensing module after the equipment control command is executed. It compares the data collected by the environmental sensing module with the prediction results contained in the equipment control command. Based on the deviation generated by the comparison, it generates a model calibration signal fed back to the thermal inertia modeling module and a strategy weight adjustment signal fed back to the adaptive control engine.

2. The Internet of Things-based digital agricultural management system according to claim 1, characterized in that, The environmental perception module includes an edge computing gateway deployed on-site. The edge computing gateway is responsible for initially cleaning and formatting the raw data collected by the sensors before uploading the structured data.

3. The Internet of Things-based digital agricultural management system according to claim 1, characterized in that, The thermal response model is a mathematical model that uses the equivalent heat capacity and equivalent total heat transfer coefficient of the target agricultural facility as core variables to quantify the delayed effect and rate of change of the control behavior on the internal ambient temperature of the facility.

4. The Internet of Things-based digital agricultural management system according to claim 1, characterized in that, The crop growth requirement curve is a data set that stores the optimal environmental parameter range corresponding to different growth stages of the crop. The adaptive control engine calls the corresponding optimal environmental parameter range as the control target based on the current growth stage of the crop.

5. A digital agricultural management system based on the Internet of Things according to claim 1, characterized in that, The adaptive control engine generates device control commands by simulating and evaluating multiple alternative control strategies through a comprehensive cost function, and selecting the alternative control strategy that minimizes the value of the comprehensive cost function as the optimal device control command. The comprehensive cost function is a function that comprehensively quantifies the deviation between environmental parameters and the optimal target interval, as well as the predicted energy consumption generated by executing the alternative control strategy, through a weighted summation method.

6. A digital agricultural management system based on the Internet of Things according to claim 1, characterized in that, The model calibration signal generated by the energy-saving feedback module is used to adjust the values ​​of equivalent heat capacity and equivalent total heat transfer coefficient in the thermal response model. The strategy weight adjustment signal generated by the energy-saving feedback module is used to adjust the weight coefficients of environmental parameters and the optimal target interval, as well as the predicted energy consumption, in the comprehensive cost function.

7. A digital agricultural management system based on the Internet of Things according to claim 1, characterized in that, The environment perception module further includes: The raw data acquisition unit is configured to be deployed at sensors and energy consumption monitoring equipment inside and outside the target agricultural facility to periodically acquire raw data points containing measurement values; The data validity verification unit is configured to receive the raw data points acquired by the raw data acquisition unit, and to filter the raw data points according to validity judgment conditions. Raw data points that do not meet the validity judgment conditions are discarded, while raw data points that meet the conditions are confirmed as valid data points. The validity judgment conditions are: Original data point x i The measured value must fall within a dynamic threshold interval calculated from the data within the most recent time window. The formula for calculating the dynamic threshold interval is: [μ-kσ,μ+kσ], where x i σ is the measurement value of the current raw data point, μ is the arithmetic mean of the measurements of all raw data points within the most recent time window, σ is the standard deviation of the measurements of all raw data points within the most recent time window, and k is a preset confidence coefficient greater than zero. The data smoothing filtering unit is configured to receive valid data points confirmed by the data validity verification unit, and to calculate a smoothing filter value by using a moving average algorithm on a series of consecutive valid data points. The calculation formula for the moving average algorithm is as follows: in, Let x be the smoothed filtered value generated at time t, N be the number of valid data points included in the moving average window, and x be the value of the smoothed filter. t-i These are the measurements of the valid data points at time ti; The structured encapsulation unit is configured to receive the smoothed filter values ​​generated by the data smoothing filter unit, and associate and combine each smoothed filter value with a unique sensor identifier and a collection timestamp to generate a structured data record. The data aggregation and uploading unit is configured to combine multiple structured data records generated by the structured encapsulation unit into a data packet and upload the data packet.

8. A digital agricultural management system based on the Internet of Things according to claim 1, characterized in that, The thermal inertia modeling module further includes: The data association and aggregation unit is configured to receive the data packet uploaded by the data collection and uploading unit in the environmental perception module, decompose the data packet into multiple structured data records, and align the data from different sources according to the timestamps in the structured data records to generate aggregated time series data including facility internal temperature, external temperature, solar radiation intensity and equipment thermal power. The model parameter identification unit is configured to receive the aggregated time series data generated by the data association and aggregation unit, and iteratively solve the energy balance equation of the facility using a system identification algorithm to obtain the equivalent heat capacity and equivalent total heat transfer coefficient characterizing the thermodynamic properties of the facility. The specific form of the energy balance equation is as follows: Among them, C eq For the equivalent heat capacity to be solved, K total Let ΔT be the equivalent overall heat transfer coefficient to be solved. in T represents the change in internal temperature of the facility over the sampling time interval. in (t) represents the internal temperature of the facility at time t, T out (t) represents the external temperature of the facility at time t, Δt represents the data sampling time interval, and P dev (t) represents the heat power generated by the control device at time t, G(t) represents the solar radiation intensity at time t, and A represents the preset effective solar radiation receiving area. The model verification and update triggering unit is configured to receive the equivalent heat capacity and the equivalent total heat transfer coefficient solved by the model parameter identification unit, calculate the mean square error characterizing the accuracy of the model using a set of independent verification datasets, and make a judgment based on the model update judgment conditions. If the model update determination condition is met, a model update instruction is generated to trigger the recalculation of the model parameter identification unit; If the model update judgment condition is not met, the equivalent heat capacity and the equivalent total heat transfer coefficient will be output. The model update determination condition is MSE>∈ max Where MSE is the mean squared error, calculated using the following formula: Where M is the number of data points in the validation dataset, and T pred,j Let T be the predicted temperature of the j-th data point by the model. actual,j The actual temperature of the j-th data point, ∈ max This is the preset maximum allowable mean square error threshold; The response prediction function generation unit is configured to receive the equivalent heat capacity and the equivalent total heat transfer coefficient output by the model verification and update triggering unit when the model update determination condition is not met, so as to construct the thermal response model. The thermal response model is used to predict the trajectory of the change of the internal ambient temperature of the facility over time under given external environmental conditions and equipment control input.

9. A digital agricultural management system based on the Internet of Things according to claim 1, characterized in that, The adaptive control engine further includes: The control target acquisition unit is configured to receive the thermal response model constructed by the response prediction function generation unit in the thermal inertia modeling module, and to receive the real-time environmental data collected by the environmental perception module, and then call the optimal environmental parameter range corresponding to the current growth stage of the crop from the crop growth demand curve. The alternative strategy generation unit is configured to receive the optimal environmental parameter range and the acquired real-time environmental data from the control target acquisition unit, first perform a control trigger determination, and when the real-time environmental data is not within the optimal environmental parameter range, generate a set of alternative control strategies containing multiple different combinations of device operation based on all available control devices and configurable operating states in the system. The set of alternative control strategies must include a zero strategy that does not perform any operation. The strategy deduction and cost quantification unit is configured to receive the thermal response model obtained by the control target acquisition unit, and to traverse each candidate control strategy in the candidate control strategy set generated by the candidate strategy generation unit. By calling the thermal response model, it deduces the environmental parameter change trajectory and predicted energy consumption of each candidate control strategy within a preset time period, and then calculates the comprehensive cost function value corresponding to each candidate control strategy through a comprehensive cost function. The specific form of the comprehensive cost function is as follows: J=w env ·D env +w energy ·C energy , Where J is the value of the comprehensive cost function, w env and w energy C represents the weighting coefficients included in the strategy weight adjustment signal received by the energy-saving feedback module. energy To predict energy consumption, D env The deviation of environmental parameters from the optimal environmental parameter range is calculated using the following formula: Where H is the preset simulation time length, and T pred (t) represents the predicted temperature at time t, where T is the trajectory of environmental parameter changes. target The median temperature within the optimal environmental parameter range; The optimal instruction decision unit is configured to receive the comprehensive cost function value of all alternative control strategies calculated by the strategy deduction and cost quantification unit, determine the final control strategy according to the optimal strategy determination criteria, and then encapsulate the final control strategy into the equipment control instruction. The optimal strategy determination criteria are: The candidate control strategy corresponding to the smallest comprehensive cost function value is selected from all the comprehensive cost function values ​​and determined as the final control strategy.

10. A digital agricultural management system based on the Internet of Things according to claim 1, characterized in that, The energy-saving feedback module further includes: The execution result acquisition unit is configured to receive the device control command encapsulated and output by the optimal instruction decision unit in the adaptive control engine, extract the prediction result from the device control command, and simultaneously acquire the actual environmental data and actual energy consumption data collected by the environmental perception module during and after the execution of the device control command. The deviation quantization calculation unit is configured to receive the prediction results extracted by the execution result acquisition unit, as well as the acquired actual environmental data and actual energy consumption data, and calculate the model prediction error value and the energy consumption prediction error value using the model prediction error formula and the energy consumption prediction error formula. The model prediction error formula is as follows: Among them, E model T represents the model prediction error value, and T represents the instruction execution time. pred (t) represents the predicted temperature at time t in the prediction results, where T is the predicted temperature. actual (t) represents the actual temperature at time t in the actual environmental data; The formula for the energy consumption prediction error is: E energy =C actual -C pred , Among them, E energy C represents the energy consumption prediction error value. actual C represents the total energy consumption in the actual energy consumption data. pred This represents the total predicted energy consumption in the forecast results. The feedback trigger judgment unit is configured to receive the model prediction error value and the energy consumption prediction error value calculated by the deviation quantization calculation unit, and make a judgment through the model calibration judgment condition and the strategy adjustment judgment condition. If the model calibration determination condition is met, a model update trigger flag is generated; If the aforementioned strategy adjustment determination condition is met, a weight adjustment trigger flag is generated; The model calibration determination condition is: E model >δ model , Where, δ model This is the preset model error threshold; The strategy adjustment determination condition is: |E energy |>δ energy , Where, δ energy This is the preset energy consumption error threshold; The feedback signal generation unit is configured to receive the model update trigger flag and the weight adjustment trigger flag generated by the feedback trigger determination unit; When the model update trigger flag is received, the model calibration signal is generated; When the weight adjustment trigger flag is received, a new set of weight coefficients is calculated using the weight update algorithm and encapsulated into the strategy weight adjustment signal. The specific form of the weight update algorithm is as follows: w ′ env =w env +α(E model -b·E energy ), Among them, w ′ env For the updated environmental deviation weighting coefficients, w env E represents the environmental deviation weighting coefficient before the update. model E represents the model prediction error. energy α is the energy consumption prediction error value, β is the preset learning rate, and β is the preset balance factor.

Images