An adaptive control method for intelligent inspection model of large-scale chicken farms

Abstract

The present invention belongs to the field of intelligent inspection and modern livestock breeding technology, and proposes an adaptive control method for an intelligent inspection model of a large-scale chicken farm, specifically: identifying the adjustment factors of the intelligent inspection model and their corresponding control nodes in the chicken farm, and then obtaining its strategy coverage according to the distribution of the control nodes in the chicken farm, and then regularly monitoring the adjustment signals of the control nodes and forming control density characteristics, and performing command redundancy analysis according to the control density characteristics and strategy coverage to form redundant settlement values, and finally regulating the control nodes in combination with the redundant settlement values. Effectively quantify the risk of individual welfare differentiation caused by the dense control instructions brought about by the complexity of the model of the intelligent inspection system itself, reduce the probability of inducing adaptive differences, achieve the purpose of maintaining the stable health status of the chickens and reducing the risk of disease, and effectively guarantee the production performance and benefits of the chicken farm.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent inspection and modern livestock breeding, and specifically relates to an adaptive control method of an intelligent inspection model of a large-scale chicken farm. Background Art

[0002] The intelligent inspection model can greatly reduce the use of human resources in livestock and poultry farming, and its farming control is often more scientific, so its farming efficiency, including weight gain rate, is better than traditional manual control or simple PID control; however, the current application of intelligent inspection models in livestock and poultry farming mostly tends to be large-scale livestock such as cattle and sheep. This is because in the initial stage of the application of intelligent inspection models in livestock and poultry farming, the breeding group itself has advantages such as controllable number and obvious individual characteristics. When large-scale chicken farms apply intelligent inspection models, due to the large number of groups, it is relatively difficult to quantify individual differences. Therefore, the obtained intelligent inspection model is prone to a large number of control influences. Factors; on the one hand, it is because the number of the above-mentioned chickens is huge, and individual differences cause frequent fluctuations in the collected behavioral data or frequent triggering of regulatory feedback. On the other hand, intelligent inspections usually use multiple monitored environmental variables and use the coordinated relationship between environmental variables to control each other. Therefore, there will be conflicting instructions in a short period of time, causing the control influencing factors to swing within a fluctuation range. This swing is often the source of a large number of control influencing factors; for example, the ventilation instruction requires turning on the fan or increasing air flow, while the temperature control demand in the same period requires reducing the entry of cold air from the outside to maintain a higher indoor temperature. There are conflicting instructions between the two in a short period of time.

[0003] Control influencing factors refer to the control instructions that regulate or adjust the breeding conditions in large-scale chicken farms, including light, temperature, ventilation, and feeding. Such intensive control instructions can easily lead to a sudden increase in the risk of welfare conditions differentiation between individuals in large-scale chicken farms, which will eventually cause a blow to the health of the chickens, increase the risk of disease, and significantly reduce the production performance of the chicken farm, affecting the benefits of the chicken farm. For example, there is a problem of temperature gradient decrease with distance in the temperature welfare of individuals far away from the temperature control source and individuals close to the temperature control source. When the temperature conditions frequently adjust the far-end individuals to accumulate the adaptability of relatively low temperature conditions, while the close-end individuals accumulate the adaptability of relatively high temperature conditions. This difference in adaptability is the source of the sudden increase in the differentiation of temperature welfare conditions. Therefore, each control node needs to form an adaptive weight adjustment based on its layout position and adjustment complexity. Summary of the invention

[0004] The purpose of the present invention is to propose an adaptive control method for an intelligent inspection model of a large-scale chicken farm to solve one or more technical problems existing in the prior art and at least provide a beneficial choice or create conditions.

[0005] In order to achieve the above object, according to one aspect of the present invention, a method for adaptively controlling an intelligent inspection model of a large-scale chicken farm is provided, the method comprising the following steps:

[0006] S1, identifying the adjustment factors of the intelligent inspection model and their corresponding control nodes in the chicken farm;

[0007] S2, obtains the strategy coverage according to the distribution of control nodes in the chicken farm;

[0008] S3, regularly monitor the adjustment signal of the control node and form a control density feature;

[0009] S4, performing instruction redundancy analysis according to control density characteristics and strategy coverage to form a redundant settlement value;

[0010] S5, regulate the control node in combination with the redundant settlement value.

[0011] Furthermore, in step S1, the method for identifying the control nodes of the intelligent inspection model in the chicken farm is: the intelligent inspection system includes several adjustment factors, and the adjustment factors include one or more of temperature control, feed amount control, ventilation control, brightness control and air humidity control; each adjustment factor is composed of several control nodes corresponding to the function, wherein the control nodes are the same type of control devices arranged at different positions in the chicken farm.

[0012] Furthermore, in step S2, the method for obtaining the strategy coverage of the control node according to the distribution of the control node in the chicken farm is: respectively obtaining the distance values ​​between a control node and the Nnei control nodes closest to it, and obtaining the average value of each distance value as the strategy coverage of the control node; wherein Nnei is the preset number of adjacent nodes.

[0013] The above method is suitable for control nodes with relatively balanced distribution in chicken farms. This method of obtaining strategy coverage will also make the control strategy more accurate. However, when the distribution balance of control nodes is not obvious, it is easy to cause underfitting, resulting in insufficient final control accuracy and decreased regulation effect. Therefore, an alternative solution is proposed for the optimization of strategy coverage for unevenly distributed control nodes.

[0014] Preferably, in step S2, the method for obtaining the policy coverage of the control nodes according to their distribution in the chicken farm can be replaced by: taking any control node as the current coverage point, obtaining the distance between the current coverage point and all other control nodes to form a distance sequence RCV.Ls, sorting the distance sequence from small to large, and calculating its policy coverage SRDg as:

[0015] ; RCV.Ls(k1) represents the k1th element of the distance sequence; rk(k1) represents the ranking value of the k1th element in the distance sequence; mean{} is the mean value function, and k1 is the sequence number of the element in the distance sequence.

[0016] Further, in step S3, the method of regularly monitoring the adjustment signal of the control node and forming the control density feature is: the preset signal monitoring interval TG is 20-60 minutes, and each control node obtains all adjustment signals within the signal monitoring interval period, wherein the adjustment signal includes a power value and a timestamp;

[0017] The control density feature is constructed once every TG. The control density feature includes adjustment amplitude and amplitude modulation density. The adjustment amplitude is the extreme difference of the power value within the TG period. The absolute value of the difference in power value between any adjustment signal and the previous adjustment signal is recorded as its adjustment degree. The average value of all adjustment degrees within the TG period is recorded as the amplitude modulation density.

[0018] Further, in step S4, the method of performing instruction redundancy analysis to form a redundant settlement value according to the control density characteristics and the policy coverage is: at the same time, if the adjustment amplitude of a control node is larger than the median value of the adjustment amplitudes of all control nodes, then the time is defined as the amplitude overflow point of the control node, and the adjustment reference interval is set according to the amplitude overflow point;

[0019] S403, within the adjustment reference interval, for the same control node, the adjustment amplitudes at all times are normalized to form an amplitude normalization value, and the product of the amplitude modulation density and the amplitude normalization value at any time is the adjustment modulus aud;

[0020] The principle of obtaining the regulation modulus here is actually to amplify the values ​​of the regulation amplitude at all times that are biased towards the maximum value through the amplitude normalization, thereby increasing the weighting of the data of the modulation density value biased towards the maximum value and reducing the weight of the smaller value. In the process of mathematical analysis, it can be understood as achieving a classification effect based on the bias of the maximum and minimum values ​​in the set. This enables it to dynamically and effectively reflect the control density state of the intelligent inspection system on a specific control node, thereby providing mathematical support for the application of the regulation modulus in time series analysis to observe the regulation effect, and thus can identify the trend and periodicity of the regulation behavior over time.

[0021] S404, obtain the adjustment modulus at each moment in the adjustment reference interval and form a modulus sequence, the number of elements in the sequence is recorded as nST; filter and obtain each element in the modulus sequence that is greater than the upper quartile and less than the lower quartile, and form two sequences recorded as the first modulus sequence FST.Ls and the second modulus sequence SST.Ls; calculate the current modulus offset Cmpau according to the first modulus sequence and the second modulus sequence of the adjustment reference interval, Cmpau = max{|avg(FST.Ls)-aud p |,|avg(SST.Ls)-aud p |};aud p Represents the adjustment modulus at the current moment, avg() is the average value function;

[0022] S405, respectively obtain the ratio of the maximum value to the minimum value in the first modulus sequence and the second modulus sequence, record them as the first modulus order F.gd and the second modulus order S.gd, and calculate the redundancy settlement value of the control node according to the current modulus offset, the first modulus order and the second modulus order.

[0023] Further, the method for setting the adjustment reference interval according to the amplitude overflow point is: S401, starting from the current moment, traversing each moment in the reverse time direction, defining the control node having at least one amplitude overflow point between the current moment and the traversed moment as an overflow condition node, searching for the amplitude overflow point closest to the current moment in any overflow condition node, and the number of moments between the amplitude overflow point and the current moment is recorded as the overflow distance;

[0024] S402, the overflow distances corresponding to each overflow condition node constitute a set Odls, the median value and interquartile range of the set Odls are recorded as Odls.Mid and Odls.IQR respectively, when the number of overflow condition nodes at the traversal moment is equal to the total number of control nodes, or when the number of moments between the traversal moment and the current moment is greater than Odls.Mid + Odls.IQR, the traversal is stopped, and the traversal moment is used as the traversal end point, and the time period between the current moment and the traversal end point is defined as the adjustment reference interval.

[0025] Preferably, or alternatively, the reference interval is adjusted to be replaced by a preset time period size.

[0026] Since the calculation of the redundant settlement value is obtained by processing the adjustment reference interval corresponding to each moment, the risk of a sudden increase in welfare conditions between individuals in large-scale chicken farms that is easily caused by intensive control instructions can be effectively quantified. However, the division of the adjustment reference interval is too dependent on the adjustment amplitude, which will lead to the problem of oversensitivity in the analysis process of the adjustment reference interval, especially in the period when most control nodes show that the amplitude overflow points occur more intensively. This problem is more prominent. However, the existing technology cannot effectively compensate for this oversensitivity phenomenon. In order to eliminate this influence, the present invention proposes a more preferred solution as follows:

[0027] Preferably, in step S4, the method of performing instruction redundancy analysis to form a redundant settlement value according to the control density characteristics and the policy coverage can be replaced by: monitoring the control nodes at equal intervals within the current C_TMF period, recording the policy coverage corresponding to each control node as SRDg, and the number of detection moments as Tn;

[0028] For any control node, the corresponding control density characteristic binary values ​​obtained at any detection time are the adjustment amplitude and the amplitude modulation density. The product of the adjustment amplitude and the amplitude modulation density is denoted as the frequency modulation entropy value Fremo. The average and variance of all frequency modulation entropy values ​​obtained by any control node in the monitoring interval are denoted as Avemo and Varmo. The order frequency modulation factor of the control node is calculated through the strategy coverage and the frequency modulation entropy value.

[0029] The calculation principle and purpose of the order frequency modulation factor is to obtain the contribution rate of each control node at any detection time, so as to accurately obtain the fluctuation trend of the control command over time and realize adaptive weight setting for the layout position and adjustment complexity of each control node.

[0030] At any detection moment, the weighted average of the adjustment amplitudes of all control nodes is recorded as the order amplitude modulation; the binary group consisting of the detection moment and its corresponding order amplitude modulation constitutes the initial time series set, and the initial time series set is processed by a high-pass filter to form an iterative time series set, and the maximum value of the iterative time series set is recorded as the high-frequency singular inflection point, and the period between any high-frequency singular inflection point and the first high-frequency singular inflection point in the reverse time direction is recorded as a sub-amplitude modulation analysis domain;

[0031] The purpose of high-pass filter processing on the initial time series set is to amplify the short-term fluctuations in the initial time series set;

[0032] For the same control node, traverse and obtain the Pearson correlation coefficient of the control density characteristics at two consecutive detection moments in the sub-amplitude modulation analysis domain. When the obtained Pearson correlation coefficient is greater than 0.8, the detection moment with the minimum amplitude modulation density in the two detection moments is eliminated. The set of detection moments in the sub-amplitude modulation analysis domain after elimination is called the iteration analysis domain set. The ratio of the number of elements in the iteration analysis domain to the number of detection moments in its amplitude modulation analysis domain is called the high-order divergence ratio Hdira.

[0033] For any detection moment in the iterative analysis domain set, the control density feature binary value corresponding to a control node and the strategy coverage are recorded to form a control feature vector. The cosine similarity of the control feature vectors corresponding to any two control nodes is calculated. The upper quartile of all cosine similarities in the iterative analysis domain set is recorded as the similarity threshold. If the cosine similarity between any two control nodes is greater than the similarity threshold, then the two control nodes are said to have similar influences. The redundancy settlement value of the control node is calculated based on the frequency modulation entropy value, the high-order divergence ratio and the order frequency modulation factor.

[0034] Beneficial effects: Since the redundant settlement value is calculated based on the load change of the regulation signal obtained by each control node and the coverage radius of each control node, the risk of individual welfare differentiation caused by the intensive regulation instructions brought about by the model complexity of the intelligent inspection system itself can be effectively extracted, thereby providing a mathematical reference for the control state of intensive instructions, and using the evaluation of the cumulative risk of welfare differentiation to provide a more stable and sustainable chicken farm environment regulation optimization method for the health status of the flock, thereby reducing the side effects of multi-model joint regulation brought about by the model diversity of the intelligent inspection system itself on environmental control in the chicken farm.

[0035] Further, in step S5, the method for regulating the control node in combination with the redundancy settlement value is: the median value of the redundancy settlement value of each control node at the same time is recorded as the settlement value level, and when the redundancy settlement value of the control node at any time is greater than the settlement value level, it is defined that high redundancy occurs at that time;

[0036] For any control node, the corresponding redundant settlement values ​​at each time when high redundancy occurs are constructed into a sequence and recorded as high redundancy settlement sequence HRSS, where the median, maximum value and interquartile range are recorded as HRSS.mid, HRSS.max and HRSS.iqr respectively;

[0037] If high redundancy occurs at the current moment, its settlement value level is recorded as PVl: when PVl>HRSS.mid+HRSS.iqr, the allocation weight is set to 0.85-0.89, and the default value is 0.85; otherwise, when PVl>HRSS.max, the allocation weight is set to 0.9-0.94, and the default value is 0.9; otherwise, when PVl>HRSS.mid, the allocation weight is set to 0.95-0.99, and the default value is 0.95; otherwise, the allocation weight is set to 1; in the next signal monitoring interval, whenever the control node obtains an adjustment signal, the product of the power value of the adjustment signal and the allocation weight is used as the power value in the actual control process.

[0038] Preferably, all undefined variables in the present invention, if not clearly defined, can be manually set thresholds.

[0039] The present invention also provides an adaptive control system of an intelligent inspection model of a large-scale chicken farm, the adaptive control system of an intelligent inspection model of a large-scale chicken farm comprises: a processor, a memory, and a computer program stored in the memory and executable on the processor, the processor implements the steps in the adaptive control method of the intelligent inspection model of a large-scale chicken farm when executing the computer program, the adaptive control system of the intelligent inspection model of a large-scale chicken farm can be run in computing devices such as desktop computers, laptop computers, PDAs, and cloud data centers, and the executable system may include, but is not limited to, a processor, a memory, and a server cluster, and the processor executes the computer program to run in the following system units:

[0040] A node identification unit is used to identify the adjustment factors of the intelligent inspection model and their corresponding control nodes in the chicken farm;

[0041] A strategy coverage calculation unit is used to obtain the strategy coverage of the control nodes according to their distribution in the chicken farm;

[0042] A regulation signal monitoring unit, used for regularly monitoring the regulation signal of the control node and forming a control density feature;

[0043] An instruction redundancy analysis unit, used for performing instruction redundancy analysis according to control density characteristics and strategy coverage to form a redundancy settlement value;

[0044] The policy control unit is used to control the control node in combination with the redundant settlement value.

[0045] The beneficial effects of the present invention are as follows: the present invention provides an adaptive control method for an intelligent inspection model of a large-scale chicken farm, which quantifies the risk of individual welfare differentiation caused by the intensive control instructions brought about by the complexity of the model of the intelligent inspection system itself according to the load changes of the adjustment signals obtained by each control node and the coverage radius of each control node, reduces the probability of inducing adaptive differences, and thus provides a mathematical reference for the control state of intensive instructions. The evaluation of the cumulative risk of welfare differentiation is used to provide a more stable and sustainable chicken farm environment adjustment and optimization method for the health status of the chickens, thereby achieving the purpose of maintaining the stability of the health status of the chickens and reducing the risk of disease, and effectively ensuring the production performance and benefits of the chicken farm. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The above and other features of the present invention will become more obvious by describing in detail the embodiments shown in the accompanying drawings. The same reference numerals in the accompanying drawings of the present invention represent the same or similar elements. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work. In the accompanying drawings:

[0047] Figure 1 Shown is a flow chart of an adaptive control method of an intelligent inspection model for a large-scale chicken farm;

[0048] Figure 2 Shown is a structural diagram of an adaptive control system for an intelligent inspection model of a large-scale chicken farm. DETAILED DESCRIPTION

[0049] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention, so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0050] like Figure 1 The figure shows a flow chart of an adaptive control method of an intelligent inspection model for a large-scale chicken farm. Figure 1 To illustrate an adaptive control method of an intelligent inspection model of a large-scale chicken farm according to an embodiment of the present invention, the method comprises the following steps:

[0051] Example 1

[0052] S1, identifying the adjustment factors of the intelligent inspection model and their corresponding control nodes in the chicken farm;

[0053] S2, obtains the strategy coverage according to the distribution of control nodes in the chicken farm;

[0054] S3, regularly monitor the adjustment signal of the control node and form a control density feature;

[0055] S4, performing instruction redundancy analysis according to control density characteristics and strategy coverage to form a redundant settlement value;

[0056] S5, regulate the control node in combination with the redundant settlement value.

[0057] Furthermore, in step S1, the method for identifying the control nodes of the intelligent inspection model in the chicken farm is: the intelligent inspection system includes several adjustment factors, and the adjustment factors include one or more of temperature control, feed amount control, ventilation control, brightness control and air humidity control; each adjustment factor is composed of several control nodes corresponding to the function, wherein the control nodes are the same type of control devices arranged at different positions in the chicken farm.

[0058] Specifically, the control equipment used for temperature control includes thermostats, air conditioning equipment or heaters; the control equipment used for feed feeding amount control includes feeding controllers or quantitative feeders; the control equipment used for ventilation volume control includes fans and ventilation controllers thereof; the control equipment used for brightness control includes roller blind systems and LED dimming systems; the control equipment used for air humidity control includes humidifiers or dehumidifiers;

[0059] The essence of the adjustment factor is a control item used to regulate the environmental attributes in a large-scale chicken farm. Each control variable is jointly controlled by several adjustment models in the intelligent inspection system, and each adjustment model can control one or more adjustment factors. The control influence factor emphasized in the background technology belongs to the subclass of the adjustment factor. The difference is that the control influence factor is subordinate to the control node, while the adjustment factor summarizes all its corresponding control nodes.

[0060] The intelligent inspection system usually includes several adjustment models. For any adjustment factor, each control node belongs to the minimum control unit of the adjustment model. At the same time, each control node receives adjustment signals from different adjustment models. The adjustment signal is in the form of percentage value, or when the adjustment signal is in the form of target power, the target power is subtracted from the current power, and the ratio of the difference to the current power is the adjustment signal in the form of percentage value converted from the target power. The adjustment model belongs to the deep learning, reinforcement learning model or PID feedback control model. Each adjustment model is provided and optimized by the supplier of the intelligent inspection model, runs on the server, and performs intelligent inspection and regulation of large-scale chicken farms in real time.

[0061] Furthermore, in step S2, the method for obtaining the strategy coverage of the control node according to the distribution of the control node in the chicken farm is: respectively obtaining the distance values ​​between a control node and the Nnei control nodes closest to it, and obtaining the average value of each distance value as the strategy coverage of the control node; wherein Nnei is the preset number of adjacent nodes.

[0062] The control nodes selected in this step are all control nodes under the same adjustment factor; the default value and minimum value of the adjacent node quantity are both 1, and the adjacent node quantity is less than 20% of the total number of control nodes.

[0063] The power value is used to measure the workload of the control node. In addition to the power value, the workload can be replaced by any of the following: feeding amount, temperature value, brightness value, ventilation amount, heat generation, water vapor output or condensation amount, depending on the adjustment factor.

[0064] The above method is suitable for control nodes with relatively balanced distribution in chicken farms. This method of obtaining strategy coverage will also make the control strategy more accurate. However, when the distribution balance of control nodes is not obvious, it is easy to cause underfitting, resulting in insufficient final control accuracy and decreased regulation effect. Therefore, an alternative solution is proposed for the optimization of strategy coverage for unevenly distributed control nodes.

[0065] Preferably, in step S2, the method for obtaining the policy coverage of the control nodes according to their distribution in the chicken farm can be replaced by: taking any control node as the current coverage point, obtaining the distance between the current coverage point and all other control nodes to form a distance sequence RCV.Ls, sorting the distance sequence from small to large, and calculating its policy coverage SRDg as:

[0066] ; RCV.Ls(k1) represents the k1th element of the distance sequence; rk(k1) represents the ranking value of the k1th element in the distance sequence; mean{} is the mean value function, and k1 is the sequence number of the element in the distance sequence.

[0067] Further, in step S3, the method of regularly monitoring the adjustment signal of the control node and forming the control density feature is: the preset signal monitoring interval TG is 20-60 minutes, and each control node obtains all adjustment signals within the signal monitoring interval period, wherein the adjustment signal includes a power value and a timestamp;

[0068] The control density feature is constructed every TG. The control density feature includes the adjustment amplitude and the amplitude modulation density. The adjustment amplitude is the range of the power value in the TG period. The absolute value of the difference between the power value of any adjustment signal and the previous adjustment signal is recorded as its adjustment degree. The average value of all adjustment degrees in the TG period is recorded as the amplitude modulation density.

[0069] The regulation signal is sent by the regulation model to each control node;

[0070] The amplitude modulation density can be incorporated into the maintenance feature weight, so that the control density feature has a reference to the signal maintenance time and enhances the sensitivity of instruction redundancy analysis.

[0071] Preferably, the calculation method of the amplitude modulation density adds the signal maintenance time weight, specifically: define the maintenance degree of any adjustment signal as the time distance between the timestamps of the adjustment signal and the next adjustment signal in the time direction, then the amplitude modulation density is the weighted average of all adjustment degrees in the TG time period, wherein the weight of the weighted average is the maintenance degree of the adjustment signal.

[0072] Further, in step S4, the method of forming a redundant settlement value by performing instruction redundancy analysis based on control density characteristics and strategy coverage is: taking the 3-7 natural days before the current moment as the monitoring interval; at the same moment, if the adjustment amplitude of a control node is larger than the median value of the adjustment amplitudes of all control nodes, then define the moment as the amplitude overflow point of the control node, S401; starting from the current moment, traverse each moment in the reverse time direction, define the control node with at least one amplitude overflow point between the current moment and the traversed moment as the overflow condition node, search for the amplitude overflow point closest to the current moment in any overflow condition node, and the number of moments between the amplitude overflow point and the current moment is recorded as the overflow distance;

[0073] S402, the overflow distances corresponding to each overflow condition node constitute a set Odls, the median value and interquartile range of the set Odls are recorded as Odls.Mid and Odls.IQR respectively, when the number of overflow condition nodes at the traversal time is equal to the total number of control nodes, or when the number of moments between the traversal time and the current moment is greater than Odls.Mid + Odls.IQR, the traversal is stopped, and the traversal moment is used as the traversal end point, and the time period between the current moment and the traversal end point is defined as the adjustment reference interval;

[0074] S403, for the same control node, normalize the adjustment amplitudes at all times to form an amplitude normalization value, and the product of the amplitude modulation density and the amplitude normalization value at any time is the adjustment modulus aud;

[0075] S404, obtain the adjustment modulus at each moment in the adjustment reference interval and form a modulus sequence, the number of elements in the sequence is recorded as nST; filter and obtain each element in the modulus sequence that is greater than the upper quartile and less than the lower quartile, and form two sequences recorded as the first modulus sequence FST.Ls and the second modulus sequence SST.Ls; calculate the current modulus offset Cmpau according to the first modulus sequence and the second modulus sequence of the adjustment reference interval, Cmpau = max{|avg(FST.Ls)-audp |,|avg(SST.Ls)-aud p |};aud p Represents the adjustment modulus at the current moment, avg() is the average value function;

[0076] S405, respectively obtain the ratio of the maximum value to the minimum value in the first modulus sequence and the second modulus sequence, record them as the first modulus order F.gd and the second modulus order S.gd, and calculate the redundancy settlement value Redr of the control node according to the current modulus offset, the first modulus order and the second modulus order:

[0077] ;

[0078] in is the rounding symbol, lg() is the logarithmic function with base 10, and max{} is the maximum value function.

[0079] Further, in step S5, the method for regulating the control node in combination with the redundancy settlement value is: the median value of the redundancy settlement value of each control node at the same time is recorded as the settlement value level, and when the redundancy settlement value of the control node at any time is greater than the settlement value level, it is defined that high redundancy occurs at that time;

[0080] For any control node, the corresponding redundant settlement values ​​at each time when high redundancy occurs are constructed into a sequence and recorded as high redundancy settlement sequence HRSS, where the median, maximum value and interquartile range are recorded as HRSS.mid, HRSS.max and HRSS.iqr respectively;

[0081] The length of the HRSS sequence is required to be no more than 100 and no less than 50. When the length of the HRSS exceeds 100, the redundant settlement values ​​corresponding to the 100 moments closest to the current moment are preferentially selected;

[0082] If high redundancy occurs at the current moment, its settlement value level is recorded as PVl: when PVl>HRSS.mid+HRSS.iqr, the allocation weight is set to 0.85-0.89, and the default value is 0.85; otherwise, when PVl>HRSS.max, the allocation weight is set to 0.9-0.94, and the default value is 0.9; otherwise, when PVl>HRSS.mid, the allocation weight is set to 0.95-0.99, and the default value is 0.95; otherwise, the allocation weight is set to 1; in the next signal monitoring interval, whenever the control node obtains an adjustment signal, the product of the power value of the adjustment signal and the allocation weight is used as the power value in the actual control process.

[0083] The preset signal monitoring interval TG is 20-60 minutes. The next signal monitoring interval means that if the time length between the current moment and the first moment in the reverse time direction at which a redundant settlement value is formed is less than the signal monitoring interval, then this part of time belongs to the next signal monitoring interval corresponding to the last moment at which a redundant settlement value is formed.

[0084] Example 2

[0085] This embodiment 2 replaces the method of performing instruction redundancy analysis based on control density characteristics and policy coverage to form a redundant settlement value based on embodiment 1, specifically:

[0086] Further, in step S4, the method of performing instruction redundancy analysis to form a redundant settlement value based on the control density characteristics and strategy coverage can be replaced by: taking the 3-7 natural days before the current time as the monitoring interval; monitoring the control nodes at equal intervals within the current C_TMF period, that is, obtaining an adjustment amplitude and amplitude modulation density, and recording the detection time scale as the detection time; recording the corresponding strategy coverage of each control node as SRDg, and the number of detection times as Tn;

[0087] For any control node, the corresponding control density characteristic binary values ​​obtained at any detection time are the adjustment amplitude and the amplitude modulation density. The product of the adjustment amplitude and the amplitude modulation density is the frequency modulation entropy value Fremo. The average and variance of all frequency modulation entropy values ​​obtained by any control node in the monitoring interval are Avemo and Varmo. The order frequency modulation factor Facor of the control node is calculated by the strategy coverage and the frequency modulation entropy value:

[0088] ;

[0089] Where k1 is the cumulative variable, Fremo k1 、Fremo k1+1 The entropy values ​​of the k1th and k1+1th frequency modulations in the frequency modulation set obtained by any control node, ln() is a logarithmic function with the natural constant e as the base;

[0090] At any detection moment, the weighted average of the adjustment amplitudes of all control nodes is the order amplitude modulation, where the weight of each control node is the order frequency modulation factor; the binary group consisting of the detection moment and its corresponding order amplitude modulation constitutes the initial time series set, and the initial time series set is processed by a high-pass filter to form an iterative time series set, and the maximum value of the iterative time series set is recorded as the high-frequency singular inflection point, and the period between any high-frequency singular inflection point and the first high-frequency singular inflection point in the reverse time direction is recorded as a sub-amplitude modulation analysis domain;

[0091] For the same control node, traverse and obtain the Pearson correlation coefficient of the control density feature at two consecutive detection moments in the sub-amplitude modulation analysis domain. When the obtained Pearson correlation coefficient is greater than 0.8, the detection moment with the minimum amplitude modulation density in the two detection moments is eliminated. The traversal starts from the detection moment closest to the current moment. If the amplitude modulation densities of the two are the same, the two detection moments are retained at the same time. The set composed of the detection moments in the sub-amplitude modulation analysis domain after elimination is recorded as the iterative analysis domain set. The ratio of the number of elements in the iterative analysis domain to the number of detection moments in its amplitude modulation analysis domain is recorded as the high-order divergence ratio Hdira.

[0092] For any detection moment in the iterative analysis domain set, the control density feature binary value and strategy coverage corresponding to a control node are recorded to form a control feature vector, that is, the control feature vector is composed of the adjustment amplitude, the amplitude modulation density and the strategy coverage. The cosine similarity of the control feature vectors corresponding to any two control nodes is calculated, and the upper quartile of all cosine similarities in the iterative analysis domain set is recorded as the similarity threshold. If the cosine similarity of any two control nodes is greater than the similarity threshold, then the two control nodes are said to have similar influences. The redundant settlement value Rsval of the control node is calculated based on the frequency modulation entropy value, the high-order divergence ratio and the order frequency modulation factor:

[0093] ;

[0094] Where Wn k2 is the number of control nodes with similar influence to the control node in the k2th sub-amplitude modulation resolution domain, Un is the number of sub-amplitude modulation resolution domains in the monitoring interval, Facor k3 、Fremo k3 are the order frequency modulation factor and frequency modulation entropy value corresponding to the k3th control node with similar influence to the current control node in the k2th sub-amplitude modulation analysis domain, respectively. Fn and Dl are the number of initial detection moments and the number of eliminated moments contained in the k2th sub-amplitude modulation analysis domain, respectively. and Hdira k2 is the number of combinations and high-order divergence ratio corresponding to the k2th sub-modulation analytic domain.

[0095] The embodiment of the present invention provides an adaptive control system of an intelligent inspection model for a large-scale chicken farm, such as Figure 2 The figure shows a structural diagram of an adaptive control system of an intelligent inspection model of a large-scale chicken farm of the present invention. The adaptive control system of an intelligent inspection model of a large-scale chicken farm of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the adaptive control method embodiment of the intelligent inspection model of a large-scale chicken farm are implemented.

[0096] The system comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to run in the following units of the system:

[0097] A node identification unit is used to identify the adjustment factors of the intelligent inspection model and their corresponding control nodes in the chicken farm;

[0098] A strategy coverage calculation unit is used to obtain the strategy coverage of the control nodes according to their distribution in the chicken farm;

[0099] A regulation signal monitoring unit, used for regularly monitoring the regulation signal of the control node and forming a control density feature;

[0100] An instruction redundancy analysis unit, used for performing instruction redundancy analysis according to control density characteristics and strategy coverage to form a redundancy settlement value;

[0101] The policy control unit is used to control the control node in combination with the redundant settlement value.

[0102] The adaptive control system of the intelligent inspection model of a large-scale chicken farm can be run on computing devices such as desktop computers, laptops, PDAs and cloud servers. The adaptive control system of the intelligent inspection model of a large-scale chicken farm, the executable system may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the example is only an example of an adaptive control system of an intelligent inspection model of a large-scale chicken farm, and does not constitute a limitation on the adaptive control system of an intelligent inspection model of a large-scale chicken farm. It may include more or fewer components than the example, or a combination of certain components, or different components. For example, the adaptive control system of the intelligent inspection model of a large-scale chicken farm may also include input and output devices, network access devices, buses, etc.

[0103] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the adaptive control system operating system of the intelligent inspection model of a large-scale chicken farm, and uses various interfaces and lines to connect the various parts of the adaptive control system operating system of the intelligent inspection model of a large-scale chicken farm.

[0104] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the adaptive control system of the intelligent inspection model of the large-scale chicken farm by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0105] Although the description of the present invention has been quite detailed and has been described in particular with respect to several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, so as to effectively cover the intended scope of the present invention. In addition, the present invention is described above with the embodiments foreseeable by the inventors, and its purpose is to provide a useful description, and those non-substantial changes to the present invention that are not currently foreseen may still represent equivalent changes of the present invention.

Claims

1. An adaptive control method for an intelligent inspection model of a large-scale chicken farm, characterized in that: The method comprises the following steps: S1, identifying the adjustment factors of the intelligent inspection model and their corresponding control nodes in a chicken farm; S2, obtains the strategy coverage according to the distribution of control nodes in the chicken farm; S3, regularly monitor the adjustment signal of the control node and form a control density feature; S4, performing instruction redundancy analysis according to control density characteristics and strategy coverage to form a redundant settlement value; S5, regulating the control node in combination with the redundant settlement value; In step S2, the method for obtaining the policy coverage of the control node according to its distribution in the chicken farm is: respectively obtaining the distance values ​​between a control node and the Nnei control nodes closest to it, and obtaining the average value of each distance value as the policy coverage of the control node; wherein Nnei is the preset number of adjacent nodes; In step S3, the method for regularly monitoring the adjustment signal of the control node and forming the control density feature is: the preset signal monitoring interval TG is 20-60 minutes, and each control node obtains all the adjustment signals within the signal monitoring interval, wherein the adjustment signal includes a power value and a timestamp; The control density feature is constructed every TG. The control density feature includes the adjustment amplitude and the amplitude modulation density. The adjustment amplitude is the range of the power value in the TG period. The absolute value of the difference between the power value of any adjustment signal and the previous adjustment signal is recorded as its adjustment degree. The average value of all adjustment degrees in the TG period is recorded as the amplitude modulation density. In step S4, the method for forming a redundant settlement value by performing instruction redundancy analysis based on the control density characteristics and the policy coverage is: at the same time, if the adjustment amplitude of a control node is larger than the median value of the adjustment amplitudes of all control nodes, then the time is defined as the amplitude overflow point of the control node, and the adjustment reference interval is set according to the amplitude overflow point; S403, for the same control node, normalize the adjustment amplitudes at all times within the adjustment reference interval to form an amplitude normalization value, and the product of the amplitude modulation density and the amplitude normalization value at any time is the adjustment modulus aud; S404, obtain the adjustment modulus at each moment in the adjustment reference interval and form a modulus sequence, the number of elements in the sequence is recorded as nST; filter and obtain each element in the modulus sequence that is greater than the upper quartile and less than the lower quartile, and form two sequences recorded as the first modulus sequence FST.Ls and the second modulus sequence SST.Ls; calculate the current modulus offset Cmpau according to the first modulus sequence and the second modulus sequence of the adjustment reference interval, Cmpau = max{|avg(FST.Ls)-aud p |,|avg(SST.Ls)-aud p |};aud p Represents the adjustment modulus at the current moment, avg() is the average value function; S405, respectively obtain the ratio of the maximum value to the minimum value in the first modulus sequence and the second modulus sequence, record them as the first modulus order F.gd and the second modulus order S.gd, and calculate the redundancy settlement value of the control node according to the current modulus offset, the first modulus order and the second modulus order; The method for setting the adjustment reference interval according to the amplitude overflow point is as follows: S401, starting from the current moment, traverse each moment in the reverse time direction, define the control node with at least one amplitude overflow point between the current moment and the traversed moment as the overflow condition node, search for the amplitude overflow point closest to the current moment in any overflow condition node, and the number of moments between the amplitude overflow point and the current moment is recorded as the overflow distance; S402, the overflow distances corresponding to each overflow condition node constitute a set Odls, the median value and interquartile range of the set Odls are recorded as Odls.Mid and Odls.IQR respectively, when the number of overflow condition nodes at the traversal moment is equal to the total number of control nodes, or when the number of moments between the traversal moment and the current moment is greater than Odls.Mid + Odls.IQR, the traversal is stopped, and the traversal moment is used as the traversal end point, and the time period between the current moment and the traversal end point is defined as the adjustment reference interval.

2. The adaptive control method of the intelligent inspection model of a large-scale chicken farm according to claim 1 is characterized in that: In step S1, the method for identifying the control nodes of the intelligent inspection model in a chicken farm is: the intelligent inspection system includes several adjustment factors, and the adjustment factors include one or more of temperature control, feed amount control, ventilation control, brightness control and air humidity control; each adjustment factor is composed of several control nodes corresponding to the function, wherein the control nodes are the same type of control devices arranged at different positions in the chicken farm.

3. The adaptive control method of the intelligent inspection model of a large-scale chicken farm according to claim 1 is characterized in that: Adjust the reference interval to replace it with the preset time period size.

4. The adaptive control method of the intelligent inspection model of a large-scale chicken farm according to claim 1 is characterized in that: In step S4, the method of performing instruction redundancy analysis to form a redundant settlement value according to the control density characteristics and the policy coverage can be replaced by: monitoring the control nodes at equal intervals within the current monitoring interval, recording the policy coverage corresponding to each control node as SRDg, and the number of time points as Tn; For any control node, the corresponding control density characteristic binary values ​​obtained at any time are the adjustment amplitude and the amplitude modulation density. The product of the adjustment amplitude and the amplitude modulation density is the frequency modulation entropy value Fremo. The average and variance of all the frequency modulation entropy values ​​obtained by any control node in the monitoring interval are Avemo and Varmo. The order frequency modulation factor of the control node is calculated by the strategy coverage and the frequency modulation entropy value. At any moment, the weighted average of the adjustment amplitudes of all control nodes is recorded as the order amplitude modulation. The binary group consisting of the moment and its corresponding order amplitude modulation constitutes the initial time series set. The initial time series set is processed by a high-pass filter to form an iterative time series set. The maximum value of the iterative time series set is recorded as the high-frequency singular inflection point. The period between any high-frequency singular inflection point and the first high-frequency singular inflection point in the reverse time direction is recorded as a sub-amplitude modulation analysis domain. For the same control node, traverse and obtain the Pearson correlation coefficient of the control density characteristics at two consecutive moments in the sub-amplitude modulation analysis domain. When the obtained Pearson correlation coefficient is greater than 0.8, the moment with the minimum amplitude modulation density in the two moments is eliminated. The set of moments in the sub-amplitude modulation analysis domain after elimination is called the iteration analysis domain set. The ratio of the number of elements in the iteration analysis domain to the number of moments in its amplitude modulation analysis domain is called the high-order divergence ratio Hdira. At any moment in the iterative analysis domain set, the control density characteristic binary value corresponding to a control node and the strategy coverage constitute the control characteristic vector, and the cosine similarity of the control characteristic vectors corresponding to any two control nodes is calculated. The upper quartile of all cosine similarities in the iterative analysis domain set is recorded as the similarity threshold. If the cosine similarity of any two control nodes is greater than the similarity threshold, then the two control nodes are said to have similar influences. The redundancy settlement value of the control node is calculated based on the frequency modulation entropy value, the high-order divergence ratio and the order frequency modulation factor.

5. The adaptive control method of the intelligent inspection model of a large-scale chicken farm according to claim 1 is characterized in that: In step S5, the method for regulating the control node in combination with the redundancy settlement value is: the median value of the redundancy settlement value of each control node at the same time is recorded as the settlement value level, and when the redundancy settlement value of the control node at any time is greater than the settlement value level, it is defined that high redundancy occurs at that time; For any control node, the corresponding redundant settlement values ​​at each time when high redundancy occurs are constructed into a sequence and recorded as high redundancy settlement sequence HRSS, where the median, maximum value and interquartile range are recorded as HRSS.mid, HRSS.max and HRSS.iqr respectively; If high redundancy occurs at the current moment, its settlement value level is recorded as PVl: when PVl>HRSS.mid+HRSS.iqr, the allocation weight is set to 0.85-0.89, and the default value is 0.85; otherwise, when PVl>HRSS.max, the allocation weight is set to 0.9-0.94, and the default value is 0.9; otherwise, when PVl>HRSS.mid, the allocation weight is set to 0.95-0.99, and the default value is 0.95; otherwise, the allocation weight is set to 1; in the next signal monitoring interval, whenever the control node obtains an adjustment signal, the product of the power value of the adjustment signal and the allocation weight is used as the power value in the actual control process.

6. An adaptive control system of an intelligent inspection model for a large-scale chicken farm, characterized in that: The adaptive control system of the intelligent inspection model of a large-scale chicken farm includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the adaptive control method of the intelligent inspection model of a large-scale chicken farm described in any one of claims 1 to 5 are implemented. The adaptive control system of the intelligent inspection model of a large-scale chicken farm runs on desktop computers, laptop computers, PDAs, and computing devices in cloud data centers.

Images