Method for Controlling Service Field by Using Time Series Data

Abstract

The present invention relates to a service site control method using time-series data. The service site control method using time-series data executed through an operating server according to the present invention comprises: a first step of collecting and preprocessing designated D (1 ≤ D ≤ (S+C)) collected data among S (S ≥ 1) sensing data sensed in a time-series manner through S sensors provided at the service site and C (C ≥ 1) image data captured in a time-series manner through C cameras provided at the site, and configuring N (1 ≤ N ≤ D) time-series data modeled to be matchable within an allowable range with T (T ≥ 1) time-series models related to the time-series nature of a process or growth performed at the site; a second step of confirming t (1 ≤ t ≤ T) time-series models having periodicity matchable with a designated periodicity related to at least one process or growth performed at the site and n (1 ≤ n ≤ N) time-series data having periodicity matchable within a designated allowable range based on the N time-series data; and a frequency domain-based method corresponding to a designated frequency characteristic for the confirmed n time-series data. A third step of processing into time series data; a fourth step of configuring multiple analysis / prediction training data for M subjects, comprising M subject-specific N' (1≤N'≤N, n'⊂N') time series data in a specified structure, including M (M≥1) subject-specific n' (1≤n'≤n) time series data required for analysis or prediction of a designated subject among the n time series data processed in the frequency domain; and training M artificial intelligence modules provided for analysis or prediction of M subjects using the configured multiple analysis / prediction training data for M subjects; and after training the M artificial intelligence modules,A fifth step of verifying m feature information for analyzing or predicting m targets by inputting m time series data, including n' time series data for m targets among n time series data collected and preprocessed and processed into the frequency domain through processes corresponding to the first to third steps above, into m designated artificial intelligence modules; a sixth step of training a control artificial intelligence module for controlling devices equipped at the service site by configuring a plurality of control learning data that includes the verified m feature information and e control information set to control e (1 ≤ e ≤ E) devices related to the m feature information among E (E ≥ 1) devices equipped at the service site; a seventh step of verifying e control information for controlling e devices equipped at the site by inputting the m feature information into the control artificial intelligence module after training of the control artificial intelligence module, if m feature information for analyzing or predicting m targets is verified through m designated artificial intelligence modules through processes corresponding to the first to fifth steps above, and the e It includes an 8th step of performing a procedure to control e devices provided at the site using control information.

Description

Technology Field

[0001] The present invention relates to a method for controlling a service site using time-series data. The method involves preprocessing collected data collected in a time-series manner through various sensors or cameras equipped at a service site to model N (N≥1) time-series data, processing n (1≤n≤N) time-series data with a series of periodic characteristics into the frequency domain, configuring multiple analysis / prediction training data for each designated target among the n time-series data processed in the frequency domain, applying the configured training data to an artificial intelligence module for training, inputting the time-series data for each designated target among the collected, preprocessed, and frequency-domain-processed time-series data into the artificial intelligence module to analyze or predict m feature information for m targets, and configuring multiple control training data including control information set to control devices related to the feature information equipped at the service site to train a control artificial intelligence module for controlling devices equipped at the site. When m feature information for analyzing or predicting m targets is confirmed through the designated m artificial intelligence modules, the m feature information is input into the control artificial intelligence module to control e devices equipped at the site. This invention relates to a method for controlling e devices equipped at the site by verifying and utilizing e control information. Background Technology

[0002] Generally, first and second-generation smart farms are operated by setting the user's control settings for the greenhouse facilities and equipment to minimize the impact of various environmental changes outside the greenhouse and to provide an optimal cultivation and water-nutrient environment for the crops inside.

[0004] However, controlling greenhouse facilities and equipment to provide optimal nutrient and water conditions for crop cultivation is practically difficult due to the diverse structures and operating methods of the equipment; furthermore, since this relies largely on the user's knowledge and experience and capabilities regarding the target cultivation environment, significant disparities can occur among farms. Additionally, accurately responding to situations where crop damage is anticipated, such as pests, diseases, physiological disorders, and abnormal weather conditions, inevitably requires heavy reliance on the judgment and decision-making of experts (consultants).

[0006] Meanwhile, as prior art, Korean Published Patent Application No. 10-2021-015853 (published on December 31, 2021) relates to a system and method for generating a cultivation map of farmland for an artificial intelligence-based agricultural robot. It involves performing polygon rendering using aerial images, identifying the boundaries between cultivated lands, identifying the actual cultivable space according to the type of crop, generating a cultivation map, and providing it to a robotic agricultural machine, thereby enabling accurate cultivation work to be performed even in cultivable areas near the boundaries between cultivated lands.

[0008] However, the aforementioned prior art merely involves generating a cultivation map by performing polygon rendering using aerial images and accurately identifying the boundaries between cultivated fields through differences in polygon density, and then transmitting this map to a robotic farm machine to cultivate crops at precise locations; it was not capable of handling various environmental aspects for actual cultivation management in a smart farm (cultivation environment management, nutrient and water management, crop growth and physiology management, pest and disease and physiological disorder management, agricultural work management, yield and quality management, energy management, etc.).

[0010] Therefore, there is a need to explore new measures to meet the necessity and demand for non-face-to-face online consulting for domestic greenhouse farms amidst the global pandemic, while overcoming the situation where existing farmers highly prefer face-to-face consulting methods, resulting in relatively low participation and satisfaction with non-face-to-face online consulting services. The problem to be solved

[0012] The objective of the present invention is to preprocess collected data collected in a time-series manner through various sensors or cameras equipped at a service site to model N (N≥1) time-series data, process n (1≤n≤N) time-series data possessing a series of periodic characteristics into the frequency domain, construct multiple analysis / prediction training data for each designated target among the n time-series data processed in the frequency domain, apply the constructed training data to an artificial intelligence module for training, input the time-series data for each designated target among the collected, preprocessed, and frequency-domain processed time-series data into the artificial intelligence module to analyze or predict m feature information for m targets, construct multiple control training data including control information set to control devices related to the feature information equipped at the service site, and train a control artificial intelligence module for controlling devices equipped at the site, and when m feature information for analyzing or predicting m targets is confirmed through the designated m artificial intelligence modules, input the m feature information into the control artificial intelligence module to confirm and utilize e control information for controlling e devices equipped at the site. The invention provides a service site control method using time series data to control e equipped devices. means of solving the problem

[0013] A service site control method using time-series data executed through an operating server according to the present invention comprises: a first step of collecting and preprocessing designated D (1 ≤ D ≤ (S+C)) collected data among S (S ≥ 1) sensors equipped at a service site and C (C ≥ 1) image data captured in time-series through C (C ≥ 1) cameras equipped at the site, and configuring N (1 ≤ N ≤ D) time-series data modeled to be matchable within an allowable range with T (T ≥ 1) time-series models related to the time-series nature of a process or growth performed at the site; a second step of identifying t (1 ≤ t ≤ T) time-series models having periodicity matchable with a designated periodicity related to at least one process or growth performed at the site, and n (1 ≤ n ≤ N) time-series data having periodicity matchable within a designated allowable range, based on the N time-series data; a third step of processing the identified n time-series data into frequency domain-based time-series data corresponding to a designated frequency characteristic; and the frequency A fourth step of constructing multiple analysis / prediction training data for M subjects, comprising M time series data for M subjects (1≤N'≤N, n'⊂N') including time series data for M subjects (M≥1) required for analysis or prediction of a specified subject among n time series data processed into a domain, and training M artificial intelligence modules equipped for analysis or prediction of M subjects using the constructed multiple analysis / prediction training data for M subjects, and after training the M artificial intelligence modules,A fifth step of verifying m feature information for analyzing or predicting m targets by inputting m time series data, including n' time series data for m targets among n time series data collected and preprocessed and processed into the frequency domain through processes corresponding to the first to third steps above, into m designated artificial intelligence modules; a sixth step of training a control artificial intelligence module for controlling devices equipped at the service site by configuring a plurality of control learning data that includes the verified m feature information and e control information set to control e (1 ≤ e ≤ E) devices related to the m feature information among E (E ≥ 1) devices equipped at the service site; a seventh step of verifying e control information for controlling e devices equipped at the site by inputting the m feature information into the control artificial intelligence module after training of the control artificial intelligence module, if m feature information for analyzing or predicting m targets is verified through m designated artificial intelligence modules through processes corresponding to the first to fifth steps above, and the e It may include an 8th step of performing a procedure to control e devices equipped at the site using control information.

[0015] According to the present invention, the preprocessing may include a procedure for averaging the collected data having a specified periodicity among the D collected data to reveal the periodic pattern or the representativeness of a certain time interval.

[0017] According to the present invention, the preprocessing may include a procedure for correcting temporal influences corresponding to biases weighted by data from a previous time for data corresponding to a specified independent variable among the D collected data.

[0019] According to the present invention, the T time series models may include at least one time series model among a time series model that is matchable with the time series characteristics of a process performed through a device provided at a service site and a time series model that is matchable with the time series characteristics related to the growth of a crop cultivated at a service site.

[0021] According to the present invention, the N time series data may include time domain-based time series data based on the time series characteristics of collected data generated by time series sensing or time series shooting.

[0023] According to the present invention, the periodicity may include at least one periodicity among a periodicity corresponding to at least one process that is repeatedly performed at regular intervals through a device provided at a service site, a periodicity corresponding to a schedule related to at least one process to be performed through a device provided at a service site, a periodicity corresponding to a change in a daily interval (or Earth's rotation interval), a periodicity corresponding to a change in the trend of a designated period, a periodicity corresponding to a change in seasons, and a periodicity corresponding to a change in a yearly interval (or Earth's orbital interval).

[0025] According to the present invention, the t time series models may include a periodic time series model that includes a time series that matches the time series characteristics of a process or growth performed at the site, and includes a periodicity that matches the periodic characteristics of a process performed including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0027] According to the present invention, the second step may include analyzing the N time series data based on at least one periodic time series model to identify n time series data having periodicity that is matchable within a specified allowable range with at least one specified periodic time series model.

[0029] According to the present invention, the second step may include, in the case of time series data where the data variation due to time variation is slower than a specified reference value, applying an exponential smoothing method specified to the time series data and then identifying n time series data having periodicity that can be matched within a specified allowable range with at least one periodic time series model.

[0031] According to the present invention, the second step may include analyzing the N time series data based on at least one periodic time series model, and separately analyzing data corresponding to random elements or residual elements included in the time series data to identify n time series data having periodicity that can be matched within a specified allowable range with at least one periodic time series model.

[0033] According to the present invention, the frequency feature may include a feature of a model-based period (or frequency) corresponding to a time series model having periodicity.

[0035] According to the present invention, the frequency feature may include a data-based period (or frequency) identified by reading time-series data having periodicity.

[0037] According to the present invention, the frequency feature may include a model-based period (or frequency) feature corresponding to a time series model having periodicity, and a data-based period (or frequency) feature identified by reading time series data having periodicity and processing it statistically or through numerical analysis to identify the period (or frequency) feature.

[0039] According to the present invention, n time series data processed into the frequency domain may include time series data processed into the frequency domain corresponding to frequency characteristics that are matched within an allowable range and periodicity that matches the periodic characteristics of a process performed at the site including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0041] According to the present invention, the n' time series data may include at least one time series data necessary for the analysis or prediction of a designated target among n time series data processed into a frequency domain corresponding to a frequency characteristic that is matched within an allowable range and a periodicity that matches the periodic characteristics of a process performed at the site including a designated period or the periodic characteristics of growth affected by a designated change (or trend).

[0043] According to the present invention, the N' time series data includes n' time series data necessary for the analysis or prediction of a designated target among n time series data processed into the frequency domain, and may include i (1 ≤ i ≤ (Nn)) time series data necessary for the analysis or prediction of a designated target among (Nn) time series data.

[0045] According to the present invention, the fourth step may further include a step of preprocessing the N' time series data so that they can be used as training data for a designated artificial intelligence module.

[0047] According to the present invention, the preprocessing may include at least one procedure among a data modification procedure for changing a character or category attribute value of at least one time series data included in the N' time series data into numeric data, and a data scaling adjustment procedure for adjusting the range difference between each time series data included in the N' time series data to within a specified range.

[0049] According to the present invention, the preprocessing may include a procedure for processing to derive the result value of the next data through k (k≥1) previous data by equalizing the periodicity of at least one time series data included in the N' time series data into a fixed time period of a specified time range.

[0051] According to the present invention, the analysis / prediction learning data may include the n' time series data as data corresponding to observation features or feature vectors for artificial intelligence learning, and may include j (i≥1) data corresponding to observation results or labels associated with the n' time series data.

[0053] According to the present invention, the analysis / prediction learning data may include n' time series data and (n'+i) time series data including i (1≤i≤(Nn)) time series data required for the analysis or prediction of a designated target among (Nn) time series data as data corresponding to observation features or feature vectors for artificial intelligence learning, and may include j (i≥1) data corresponding to observation results or labels related to the designated time series data among the (n'+i) time series data.

[0055] According to the present invention, the fifth step may further include a step of preprocessing N' time series data for each of the M targets so that they can be input into the M learned artificial intelligence modules.

[0057] According to the present invention, the feature information may include feature information corresponding to the result of an artificial intelligence-based analysis or prediction using frequency domain-based time series data, and feature information corresponding to the result of an artificial intelligence-based analysis or prediction using multiple frequency domain-based time series data and time domain-based time series data.

[0059] According to the present invention, the sixth step may further include a step of performing a procedure to control e devices provided at the site using the e control information.

[0061] According to the present invention, the sixth step may further include: a step of verifying m actual measurement information for each target that is currently or after a certain period of time has elapsed, regarding the combination of m feature information related to the control of e devices provided at the site and e control information when e devices provided at the site are controlled using e control information; a step of generating feedback information corresponding to control information for improving efficiency related to the control of e devices by comparing and analyzing the m feature information and the verified m actual measurement information for each target; and a step of additionally training the control artificial intelligence module by configuring control learning data including the m feature information and the generated feedback information.

[0063] According to the present invention, the actual measurement information may include information on the actual measurement of at least one target among production volume, production quality, occurrence of an anomaly, occurrence of a failure, occurrence of environmental damage, and occurrence of pests and diseases, corresponding to the result of controlling e devices using the e control information.

[0065] According to the present invention, the feedback information may include e' control information that retrospectively adjusts e' (1 ≤ e' ≤ e) control information among e control information that controls e devices provided at the site to improve efficiency related to the control of e devices.

[0067] According to the present invention, the control learning data may include m feature information as data corresponding to an observed feature or feature vector for artificial intelligence learning, and j (i≥1) data corresponding to an observed result or label including e control information.

[0069] According to the present invention, the eighth step may further include: a step of verifying m actual measurement information for each target, measured from the present to after a certain period of time, regarding the combination of m feature information related to the control of e devices provided at the site and e control information when e devices provided at the site are controlled using e control information confirmed through the control artificial intelligence module; a step of comparing and analyzing the m feature information and the verified m actual measurement information for each target to generate feedback information corresponding to control information for improving efficiency related to the control of the e devices; and a step of configuring control learning data including the m feature information and the generated feedback information to further train the control artificial intelligence module. Effects of the invention

[0071] According to the present invention, a service site control method using time series data has the advantage of being able to predict future data, such as equipment preventive maintenance, marketing basis data, and environmental change prediction, by using previous values ​​in the form of past, present, and future.

[0073] In addition, according to the present invention, there is an advantage in that it is possible to predict abnormal cause values ​​for equipment predictive maintenance, crop growth / physiology prediction, pest and disease cause prediction, and IoT environment maintenance by detecting patterns in which periodic characteristics are repeated. Brief explanation of the drawing

[0075] FIG. 1 is a diagram illustrating the configuration of a service site control system using time series data according to the method of implementing the present invention. FIG. 2 is a flowchart illustrating the process of processing time series data, identified as having periodicity matching a specified periodicity related to a process or growth according to the method of implementing the present invention, into frequency domain-based time series data. FIG. 3 is a flowchart illustrating the process of analyzing or predicting feature information of a target by inputting time series data for each target into an artificial intelligence module trained for analysis or prediction according to the method of implementing the present invention. FIG. 4 is a flowchart illustrating the process of controlling a device at a site by training a control artificial intelligence module for controlling a device at a site according to the method of implementing the present invention, and then inputting feature information into the control artificial intelligence module to verify control information for controlling a device at a site. Specific details for implementing the invention

[0076] The operating principle of a preferred embodiment of the present invention will be described in detail below with reference to the attached drawings and description. However, the drawings and descriptions below relate to preferred embodiments among various methods for effectively explaining the features of the present invention, and the present invention is not limited to the drawings and descriptions below.

[0078] In other words, the following embodiments correspond to preferred union forms among the numerous embodiments of the present invention. It is explicitly stated that embodiments in which a specific configuration (or step) is omitted, embodiments in which a function implemented in a specific configuration (or step) is divided into specific configurations (or steps), embodiments in which a function implemented in two or more configurations (or steps) is integrated into a single configuration (or step), embodiments in which the order of operation of a specific configuration (or step) is changed, etc., are all included within the scope of the present invention, even if not separately mentioned in the following embodiments. Accordingly, it is clearly specified that various embodiments corresponding to subsets or complements based on the following embodiments may be divided retroactively to the filing date of the present invention.

[0080] Furthermore, in describing the present invention below, specific descriptions of related known functions or configurations will be omitted if it is determined that such detailed descriptions could unnecessarily obscure the essence of the invention. Additionally, the terms described below are defined considering their functions in the present invention, and these definitions may vary depending on the intentions or practices of the user or operator. Therefore, their definitions should be based on the overall content of the present invention.

[0082] Consequently, the technical concept of the present invention is determined by the claims, and the following embodiments are merely a means to efficiently explain the progressive technical concept of the present invention to those skilled in the art to which the present invention belongs.

[0084] Figure 1 is a diagram illustrating the configuration of a service site control system using time series data according to the method of implementing the present invention.

[0086] More specifically, the present Figure 1 describes a method for preprocessing time-series collected data through various sensors or cameras equipped at a service site to model N (N≥1) time-series data, processing n (1≤n≤N) time-series data with a series of periodic characteristics into the frequency domain, configuring multiple analysis / prediction training data for each designated target among the n time-series data processed in the frequency domain, applying the configured training data to an artificial intelligence module for training, inputting the time-series data for each designated target among the time-series data collected, preprocessed, and processed in the frequency domain into the artificial intelligence module to identify m feature information that analyzes or predicts m targets, configuring multiple control training data including control information set to control devices related to the feature information equipped at the service site, and training a control artificial intelligence module for controlling devices equipped at the site, wherein when m feature information that analyzes or predicts m targets is identified through the designated m artificial intelligence modules, the m feature information is input into the control artificial intelligence module to identify e control information for controlling e devices equipped at the site. This illustrates a system configuration for controlling e devices provided at the site by verifying and utilizing them. A person skilled in the art to which this invention pertains could infer various implementation methods of the system (e.g., methods in which some components are omitted, subdivided, or combined) by referring to and / or modifying Figure 1. However, the present invention comprises all such inferred implementation methods and its technical features are not limited to the implementation method illustrated in Figure 1.

[0088] The system of the present invention comprises an operating server (100) that inputs time-series data collected and configured from one or more sensors or cameras equipped at a service site into a plurality of artificial intelligence modules trained for the analysis or prediction of a designated target to identify feature information for analyzing or predicting the designated target, and trains a control artificial intelligence module equipped at the site by configuring control information for controlling a device equipped at the service site in relation to the feature information into a plurality of training data, and then inputs the feature information into the control artificial intelligence module to identify control information for controlling the device, and controls the device at the site using the identified control information; and a user terminal (150) that receives the result information from the operating server (100). Meanwhile, the operating server (100) may be implemented in at least one or a combination of two or more forms among an independent server form, a combination of two or more servers, and a server program installed and run on an existing server, and the present invention is not limited by the method of implementing the operating server (100).

[0090] The above user terminal (150) is a general term for a computer device used by a user who checks the result information provided through the above operating server (100) in relation to the above site, and may include at least one of a wired terminal such as a computer or laptop connected to a wired network and a wireless terminal such as a mobile phone, smartphone, tablet PC, or laptop connected to a wireless network.

[0092] Preferably, the user terminal (150) is equipped with and runs an application or program that performs at least one procedure designated for the function of receiving result information regarding a designated analysis or prediction from an operating server (100).

[0094] Meanwhile, referring to Figure 1, the operating server (100) comprises: a data configuration unit (105) that configures time series data collected and modeled through sensors or cameras provided at the service site; a data verification unit (110) that verifies time series data having a periodicity that matches a designated periodicity related to at least one process or growth performed at the site; a data processing unit (115) that processes time series data into frequency domain-based time series data; an artificial intelligence learning unit (120) that configures multiple learning data for analysis / prediction for each designated target including time series data processed in the frequency domain, and trains an artificial intelligence module for analysis or prediction for each target using the configured learning data; a feature information verification unit (125) that verifies feature information for analyzing or predicting the designated target by inputting the time series data for each designated target processed in the frequency domain into the artificial intelligence module; a control module learning unit (130) that configures multiple control learning data including control information for controlling a device at the site related to the feature information and trains a control artificial intelligence module for controlling a device provided at the site; and the When feature information that analyzes or predicts a designated target is confirmed through an artificial intelligence module, the system may be configured to include a control information confirmation unit (135) that confirms control information for controlling a device equipped at the site by inputting the confirmed feature information into the control artificial intelligence module, and a device control unit (140) that controls the device at the site using the confirmed control information.

[0096] According to Figure 1, the data composition unit (105) can collect and preprocess designated D (1 ≤ D ≤ (S+C)) collected data from among S (S ≥ 1) sensors equipped at the service site and C (C ≥ 1) cameras equipped at the site, and model N (1 ≤ N ≤ D) time series data that can be matched within an allowable range with T (T ≥ 1) time series models related to the time series of processes or growth performed at the site.

[0098] Typically, the preprocessing of the collected data may include performing at least one or more combinations of data cleaning, data transformation, data filtering, data integration, and data reduction.

[0100] Here, the data cleaning may include one or more combinations of the following: a procedure for filling missing data with default values ​​or replacing it with average / median values; a procedure for identifying and correcting or removing abnormal data corresponding to an abnormal state; and a procedure for smoothing noisy data.

[0102] In addition, the above data transformation may include a procedure for transforming the data into a form that is easy to analyze through one or more combinations of normalization, aggregation, summarization, and hierarchy generation.

[0104] In addition, the data filtering described above may include any one or a combination of two or more of an error data detection procedure, an error data correction procedure, an error data deletion procedure, a duplicate data verification procedure, a duplicate data correction procedure, and a duplicate data deletion procedure.

[0106] In addition, the above data integration may include a procedure for integrating similar data or linked data to facilitate data analysis.

[0108] In addition, the above data reduction may include a procedure to remove unused data among the collected data.

[0110] Meanwhile, according to the method of implementing the present invention, the preprocessing of the collected data may include a procedure for averaging the collected data having a designated periodicity among the D collected data to reveal the periodic pattern or the representativeness of a certain time interval.

[0112] In addition, according to the method of implementing the present invention, the preprocessing of the collected data may include a procedure for correcting the temporal influence corresponding to the bias weighted by the data of the previous time with respect to the data corresponding to a designated independent variable among the D collected data.

[0114] Meanwhile, T (T≥1) time series models related to the time series of processes or growth performed at the above-mentioned site may include at least one time series model among a time series model that is matchable with the time series characteristics of a process performed through a device provided at the service site and a time series model that is matchable with the time series characteristics related to the growth of crops cultivated at the agricultural site.

[0116] In addition, N (1 ≤ N ≤ D) time series data modeled to be matchable within the above allowable range may include time domain-based time series data based on the time series characteristics of collected data generated by time series sensing or time series shooting.

[0118] According to Figure 1, when N time series data are configured through the data configuration unit (105), the data verification unit (110) can verify t (1 ≤ t ≤ T) time series models having periodicity that is matchable with a specified periodicity related to at least one process or growth performed at the site, and n (1 ≤ n ≤ N) time series data having periodicity that is matchable within a specified allowable range based on the N time series data.

[0120] According to the method of implementing the present invention, the periodicity may include at least one periodicity among a periodicity corresponding to at least one process that is repeatedly performed at regular intervals through a device provided at a service site, a periodicity corresponding to a schedule related to at least one process to be performed through a device provided at a service site, a periodicity corresponding to a change in a daily interval (or Earth's rotation interval), a periodicity corresponding to a change in the trend of a designated period, a periodicity corresponding to a change in seasons, and a periodicity corresponding to a change in a yearly interval (or Earth's orbital interval).

[0122] Here, the periodicity may not include periodicity related to simple vibration of a device sensed through a conventional vibration sensor, and periodicity related to the frequency of electromagnetic waves sensed through an electromagnetic wave sensor.

[0124] According to the method of implementing the present invention, the t (1 ≤ t ≤ T) time series models having a periodicity that is matchable with a designated periodicity related to at least one process or growth may include a periodicity time series model that includes a time series that is matched with the time series characteristics of the process or growth performed at the site, and includes a periodicity that is matched with the periodic characteristics of the process performed including a designated period or the periodic characteristics of the growth affected by a designated change (or trend).

[0126] Additionally, the n (1 ≤ n ≤ N) time series data having the above periodicity may include time series data having a specified periodicity and a periodicity matched within a specified allowable range related to at least one process or growth performed at the site.

[0128] Additionally, the n time series data may include a periodic time series model that is matchable with a specified periodicity related to at least one process or growth performed at the site, and time series data having a periodicity that is matched within a specified allowable range.

[0130] Meanwhile, according to the method of implementing the present invention, the data verification unit (110) can analyze the N time series data based on at least one periodic time series model and verify n time series data having periodicity that can be matched within a specified allowable range with at least one specified periodic time series model.

[0132] In addition, according to the method of implementing the present invention, the data verification unit (110) can verify n time series data having periodicity that is matchable within a specified allowable range with at least one periodic time series model, after applying an exponential smoothing method specified for the time series data in the case of time series data where the data variation according to time variation is slower than a specified reference value.

[0134] In addition, according to the method of implementing the present invention, the data verification unit (110) analyzes the N time series data based on at least one periodic time series model, and separates and analyzes data corresponding to random elements or residual elements included in the time series data to verify n time series data having periodicity that can be matched within a specified allowable range with at least one periodic time series model.

[0136] According to Figure 1, when n (1 ≤ n ≤ N) time series data having periodicity are identified through the data verification unit (110), the data processing unit (115) can process the identified n time series data into frequency domain-based time series data corresponding to a specified frequency characteristic.

[0138] According to the method of implementing the present invention, the frequency feature may include a feature of a model-based period (or frequency) corresponding to a time series model having periodicity.

[0140] In addition, according to the method of implementing the present invention, the frequency feature may include a data-based period (or frequency) identified by reading time-series data having periodicity.

[0142] In addition, according to the method of implementing the present invention, the frequency feature may include a model-based period (or frequency) feature corresponding to a time series model having periodicity and a data-based period (or frequency) feature identified by reading time series data having periodicity and performing statistical processing (e.g., average, correction, etc.) or numerical analysis processing (e.g., average, correction, interpolation, etc.).

[0144] Meanwhile, according to the method of implementing the present invention, the n time series data processed into the frequency domain may include time series data processed into the frequency domain corresponding to frequency characteristics that are matchable within an allowable range and periodicity that matches the periodic characteristics of a process performed at the site including a designated period or the periodic characteristics of growth affected by a designated change (or trend).

[0146] In addition, the n time series data processed into the frequency domain do not include time series data related to the simple vibration of the device sensed through the vibration sensor or time series data related to the frequency of the electromagnetic wave sensed through the electromagnetic wave sensor.

[0148] Here, time series data related to the simple vibration of the device sensed through the vibration sensor or time series data related to the frequency of the electromagnetic wave sensed through the electromagnetic wave sensor are not included in n time series data processed into a frequency domain corresponding to frequency characteristics that match within an allowable range and periodicity that match the periodic characteristics of growth affected by the periodic characteristics of a process performed at the site including a specified period or a specified change (or trend).

[0150] In addition, time series data related to simple vibration of the device sensed through the vibration sensor or time series data related to the frequency of electromagnetic waves sensed through the electromagnetic wave sensor can be identified or managed as time series data corresponding to a separate frequency domain distinct from n time series data processed into a frequency domain corresponding to frequency characteristics that match the periodicity and frequency characteristics within an allowable range, which match the periodic characteristics of a process performed at the site including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0152] According to Figure 1, the artificial intelligence learning unit (120) can, when time series data is processed into frequency domain-based time series data through the data processing unit (115), construct multiple analysis / prediction learning data for M subjects, including M (M≥1) subject-specific n' (1≤n'≤n) subject-specific time series data required for analysis or prediction of a designated subject among the n subject-specific time series data processed into the frequency domain, and construct multiple analysis / prediction learning data for M subjects, including M subject-specific N' (1≤N'≤N, n'⊂N') subject-specific time series data in a designated structure, and train M artificial intelligence modules provided for analysis or prediction of M subjects using the constructed multiple analysis / prediction learning data for M subjects.

[0154] According to the method of implementing the present invention, the n' time series data may include at least one time series data required for a specified analysis or prediction among n time series data processed into a frequency domain corresponding to a frequency characteristic that is matched within an allowable range and a periodicity that matches the periodic characteristics of a process performed at the site including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0156] Meanwhile, the above N' time series data may include n' time series data required for a specified analysis or prediction among the n time series data processed into the frequency domain.

[0158] In addition, according to the method of implementing the present invention, the N' time series data may include n' time series data required for a specified analysis or prediction among n time series data processed into the frequency domain, and i (1 ≤ i ≤ (Nn)) time series data required for a specified analysis or prediction among (Nn) time series data.

[0160] Here, the (Nn) time series data may include time domain-based time series data.

[0162] Meanwhile, according to the method of implementing the present invention, the artificial intelligence learning unit (120) can preprocess the N' time series data so that they can be used as training data for a designated artificial intelligence module.

[0164] The above preprocessing may perform at least one or more combinations of data cleaning, data transformation, data filtering, data integration, and data reduction for at least one time series data included in the N' time series data.

[0166] According to the method of implementing the present invention, the preprocessing may include at least one procedure among a data modification procedure for changing a character or category attribute value of at least one time series data included in the N' time series data into numeric data, and a data scaling adjustment procedure for adjusting the range difference between each time series data included in the N' time series data to within a specified predetermined range.

[0168] Here, the preprocessing may include a procedure for processing to enable the derivation of the result value of the next data through k (k≥1) previous data by equalizing the periodicity of at least one time series data included in the N' time series data into a fixed time period of a specified time range.

[0170] Meanwhile, according to the method of implementing the present invention, the analysis / prediction learning data may include the n' time series data as data corresponding to observation features or feature vectors for artificial intelligence learning, and may include j (i≥1) data corresponding to observation results or labels associated with the n' time series data.

[0172] In addition, according to the method of implementing the present invention, the analysis / prediction learning data may include n' time series data and (n'+i) time series data including i (1≤i≤(Nn)) time series data required for a specified analysis or prediction among (Nn) time series data as data corresponding to an observation feature or feature vector for artificial intelligence learning, and may include j (i≥1) data corresponding to an observation result or label related to a specified time series data among the (n'+i) time series data. Here, the i time series data may include time domain-based time series data.

[0174] In addition, the i time series data mentioned above may include time domain-based time series data.

[0176] Meanwhile, according to the method of implementing the present invention, the M artificial intelligence modules may include an LSTM (Long Short-Term Memory) series algorithm optimized for time series analysis.

[0178] According to Figure 1, the feature information verification unit (125) can verify m feature information that analyzes or predicts m targets by training M artificial intelligence modules provided for analysis or prediction of M targets through the artificial intelligence learning unit (120), and then inputting N' time series data for m targets, including n' time series data for m targets among n time series data that are collected, preprocessed, and processed into the frequency domain, into the specified m artificial intelligence modules.

[0180] Here, the n' time series data for the specified analysis or prediction may include at least one time series data required for the specified analysis or prediction among n time series data processed into a frequency domain corresponding to frequency characteristics that are matched within an allowable range and periodicity matching the periodic characteristics of a process performed at the site including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0182] Additionally, the above N' time series data may include n' time series data required for a specified analysis or prediction among the n time series data processed into the frequency domain.

[0184] Additionally, the N' time series data may include n' time series data required for a specified analysis or prediction among the n time series data processed in the frequency domain, and i (1 ≤ i ≤ (Nn)) time series data required for a specified analysis or prediction among the (Nn) time series data. Here, the (Nn) time series data may include time domain-based time series data.

[0186] Here, the (Nn) time series data may include time domain-based time series data.

[0188] Meanwhile, according to the method of implementing the present invention, the feature information verification unit (125) can preprocess N' time series data for each of the M targets so that they can be applied to the learned M artificial intelligence modules.

[0190] Here, the preprocessing may perform at least one or more combinations of data cleaning, data transformation, data filtering, data integration, and data reduction for at least one time series data included in the N' time series data.

[0192] Additionally, the preprocessing may include at least one procedure among a data modification procedure that changes a character or category attribute value of at least one time series data included in the N' time series data into numeric data, and a data scaling adjustment procedure that adjusts the range difference between each time series data included in the N' time series data to within a specified range.

[0194] In addition, the above preprocessing may include a procedure for processing to derive the result value of the next data through k (k≥1) previous data by uniformizing the periodicity of at least one time series data included in the N' time series data into a fixed time period of a specified time range.

[0196] And, the feature information verification unit (125) can verify m feature information that analyzes or predicts m targets by inputting the preprocessed N' time series data for m targets into m artificial intelligence modules for analysis or prediction for m targets.

[0198] Meanwhile, the above feature information may include result information corresponding to AI-based analysis or prediction using frequency domain-based time series data, and result information corresponding to AI-based analysis or prediction using multiple frequency domain-based time series data and time domain-based time series data.

[0200] Here, the above feature information may not include result information corresponding to artificial intelligence-based analysis or prediction using only time domain-based time series data.

[0202] Meanwhile, according to the method of implementing the present invention, the feature information verification unit (125) can store the verified m feature information in a designated storage medium and provide the verified m feature information to a designated user terminal (140) associated with the site.

[0204] Here, the above feature information may include at least one feature information among the result of analyzing an anomaly occurrence, the result of predicting an anomaly occurrence, the result of predicting a future state, and the result of feeding back the control or adjustment of at least some processes performed at the site.

[0206] According to Figure 1, the control module learning unit (130) can train a control artificial intelligence module for controlling devices equipped at the service site by configuring a plurality of control learning data that includes the m identified feature information and e (1 ≤ e ≤ E) devices related to the m feature information among the E (E ≥ 1) devices equipped at the service site, when m feature information is identified or predicted through the feature information verification unit (125).

[0208] According to the method of implementing the present invention, the control module learning unit (130) can provide the identified m feature information to a designated expert terminal (not shown separately) to verify e control information set to control e devices related to the m feature information among the E devices provided at the service site by an expert.

[0210] Here, the expert may include at least one of a person holding a degree in a field related to the service site, a person who has obtained a designated certification in a field related to the service site, or a person who has experience in a field related to the service site for a designated period of time or longer.

[0212] According to the method of implementing the present invention, the control module learning unit (130) can perform a procedure to control e devices provided at the site using the e control information.

[0214] In addition, according to the method of implementing the present invention, the control module learning unit (130) can, when controlling e devices provided at the site using e control information, check m actual measurement information for each target that is currently or after a certain period of time has elapsed regarding the combination of m feature information related to the control of e devices provided at the site and the e control information, and compare and analyze the m feature information and the m actual measurement information for each target that is checked to generate feedback information corresponding to control information for improving efficiency related to the control of e devices, thereby configuring control learning data including the m feature information and the generated feedback information to further train the artificial intelligence module for control.

[0216] Here, the control module learning unit (130) can receive actual measurement information for each of the m targets from a user terminal corresponding to the site.

[0218] In addition, the control module learning unit (130) receives data from a server that generates / manages data corresponding to actual measurement information for m targets or generates / manages data to be used to derive actual measurement information for m targets, and can verify the actual measurement information for m targets based on the received data.

[0220] In addition, the control module learning unit (130) can receive sensing data sensed through a sensor that senses data corresponding to actual measurement information for m targets among the sensors provided at the site, or data to be used to derive actual measurement information for m targets, and can verify actual measurement information for m targets based on the received sensing data.

[0222] In addition, the above-mentioned actual measurement information may include information on actual measurements of at least one target among production volume, production quality, occurrence of abnormalities, occurrence of failures, occurrence of environmental damage, and occurrence of pests and diseases, corresponding to the result of controlling e devices using the above-mentioned e control information.

[0224] In addition, the feedback information may include e' control information obtained by retrospectively adjusting designated e' (1 ≤ e' ≤ e) control information among e control information obtained by controlling e devices equipped at the site to improve efficiency related to the control of the e devices.

[0226] In addition, the control learning data may include m feature information as data corresponding to observed features or feature vectors for artificial intelligence learning, and j (i≥1) data corresponding to observed results or labels including feedback information.

[0228] Additionally, the control learning data may include m feature information as data corresponding to observation features or feature vectors for artificial intelligence learning, and j (i≥1) data corresponding to observation results or labels including e control information.

[0230] According to Drawing 1, the control information verification unit (135) can verify e control information for controlling e devices equipped at the site by inputting the m feature information into the control artificial intelligence module after learning the control artificial intelligence module for controlling the device equipped at the site through the control module learning unit (130), and then verifying m feature information that analyzes or predicts m targets through the designated m artificial intelligence modules.

[0232] According to Drawing 1, the device control unit (140) can perform a procedure to control the e devices provided at the site using the e control information when the e control information is confirmed through the control information confirmation unit (135).

[0234] According to the method of implementing the present invention, the device control unit (140) can, when controlling e devices provided at the site using e control information confirmed through the control artificial intelligence module, confirm m actual measurement information for each target that is currently or after a certain period of time has elapsed regarding the combination of m feature information related to the control of e devices provided at the site and the e control information, compare and analyze the m feature information and the confirmed m actual measurement information for each target to generate feedback information corresponding to control information for improving efficiency related to the control of e devices, and thereby configure control learning data including the m feature information and the generated feedback information to further train the control artificial intelligence module.

[0236] Here, the device control unit (140) can receive actual measurement information for each of the m targets from a user terminal corresponding to the site.

[0238] Additionally, the device control unit (140) can receive data from a server that generates / manages data corresponding to actual measurement information for m targets or generates / manages data to be used to derive actual measurement information for m targets, and can verify the actual measurement information for m targets based on the received data.

[0240] Additionally, the device control unit (140) can receive sensing data sensed through a sensor that senses data corresponding to actual measurement information for m targets among the sensors provided at the site, or data to be used to derive actual measurement information for m targets, and can verify actual measurement information for m targets based on the received sensing data.

[0242] Here, the actual measurement information may include information on at least one object among production volume, production quality, occurrence of anomalies, occurrence of failures, occurrence of environmental damage, and occurrence of pests and diseases, corresponding to the result of controlling e devices using the e control information.

[0244] In addition, the feedback information may include e' control information obtained by retrospectively adjusting designated e' (1 ≤ e' ≤ e) control information among e control information obtained by controlling e devices equipped at the site to improve efficiency related to the control of the e devices.

[0246] In addition, the control learning data may include m feature information as data corresponding to observed features or feature vectors for artificial intelligence learning, and j (i≥1) data corresponding to observed results or labels including feedback information.

[0248] Figure 2 is a flowchart illustrating the process of processing time series data, identified as having periodicity matching a specified periodicity related to a process or growth according to the method of implementing the present invention, into frequency domain-based time series data.

[0250] More specifically, Figure 2 illustrates a process of configuring time series data modeled by collecting through sensors or cameras provided at a service site in an operating server (100), verifying time series data having a periodicity that matches a designated periodicity related to at least one process or growth performed at the site, and processing the time series data into frequency domain-based time series data. A person skilled in the art to which the present invention belongs may infer various methods of implementing the above process (e.g., methods in which some steps are omitted or the order is changed) by referring to and / or modifying Figure 2. However, the present invention is formed by including all such inferred methods, and its technical features are not limited to the methods illustrated in Figure 2.

[0252] Referring to Figure 2, the operating server (100) collects D (1 ≤ D ≤ (S+C)) of the S sensing data sensed in a time series through S (S ≥ 1) sensors provided at the service site and C image data captured in a time series through C (C ≥ 1) cameras provided at the site (200).

[0254] Here, the preprocessing may include a procedure for averaging the collected data having a specified periodicity among the D collected data to reveal the periodic pattern or the representativeness of a certain time interval.

[0256] In addition, the above preprocessing may include a procedure for correcting temporal influences corresponding to biases weighted by data from the previous time for data corresponding to a specified independent variable among the D collected data.

[0258] When the above D collected data are collected, the operating server (100) preprocesses the collected D data and constructs N (1 ≤ N ≤ D) time series data that can be modeled to be matched within an allowable range with T (T ≥ 1) time series models related to the time series of processes or growth performed at the site (205).

[0260] Here, the T time series models may include at least one time series model among a time series model that is matchable with the time series characteristics of a process performed through a device provided at a service site and a time series model that is matchable with the time series characteristics related to the growth of a crop cultivated at an agricultural site.

[0262] In addition, the above N time series data may include time domain-based time series data based on the time series characteristics of collected data generated by time series sensing or time series shooting.

[0264] When the above N time series data are configured, the operating server (100) checks t (1 ≤ t ≤ T) time series models having a periodicity that can be matched with a specified periodicity related to at least one process or growth performed at the site, and n (1 ≤ n ≤ N) time series data having a periodicity that can be matched within a specified allowable range based on the above N time series data (210).

[0266] Here, the periodicity may include at least one periodicity among a periodicity corresponding to at least one process that is repeatedly performed at regular intervals through a device provided at a service site, a periodicity corresponding to a schedule related to at least one process to be performed through a device provided at a service site, a periodicity corresponding to a change in a daily interval (or Earth's rotation interval), a periodicity corresponding to a change in a trend of a designated period, a periodicity corresponding to a change in seasons, and a periodicity corresponding to a change in a yearly interval (or Earth's orbital interval).

[0268] In addition, the above t time series models may include a periodic time series model that includes a time series that matches the time series characteristics of the process or growth performed at the site, and includes a periodicity that matches the periodic characteristics of the process performed including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0270] According to the method of implementing the present invention, the operating server (100) can analyze the N time series data based on at least one periodic time series model and identify n time series data having periodicity that can be matched within a specified allowable range with at least one specified periodic time series model.

[0272] In addition, according to the method of implementing the present invention, in the case of time series data where the data variation due to time variation is slower than a specified reference value, the operating server (100) can apply an exponential smoothing method specified for the time series data and then identify n time series data having periodicity that can be matched within a specified allowable range with at least one periodic time series model.

[0274] When the above n time series data are confirmed, the operating server (100) processes the confirmed n time series data into frequency domain-based time series data corresponding to a specified frequency characteristic (215).

[0276] Here, the frequency feature may include a model-based period (or frequency) feature corresponding to a time series model having periodicity, or a data-based period (or frequency) identified by reading time series data having periodicity, or may include a feature of a period (or frequency) identified by statistically processing or numerically analyzing the features of a model-based period (or frequency) corresponding to a time series model having periodicity and the features of a data-based period (or frequency) identified by reading time series data having periodicity.

[0278] In addition, the n time series data processed into the frequency domain may include time series data processed into the frequency domain corresponding to frequency characteristics that are matchable within an allowable range and periodicity that matches the periodic characteristics of a process performed at the site including a specified period or the periodic characteristics of growth affected by a specified change (or trend).

[0280] Figure 3 is a flowchart illustrating the process of analyzing or predicting feature information of a target by inputting time-series data for each target into an artificial intelligence module trained for analysis or prediction according to the method of implementing the present invention.

[0282] More specifically, Figure 3 illustrates a process in which, after the process of Figure 2, an operating server (100) configures multiple learning data for analysis / prediction for each designated target, including time series data processed in the frequency domain, and trains an artificial intelligence module for analysis or prediction for each target using the configured learning data, and then inputs the time series data processed in the frequency domain for each designated target into the artificial intelligence module to verify feature information for analyzing or predicting the designated target. A person skilled in the art to which the present invention belongs may infer various methods of implementing the above process (e.g., methods in which some steps are omitted or the order is changed) by referring to and / or modifying Figure 3; however, the present invention is formed by including all such inferred methods, and its technical features are not limited to the methods illustrated in Figure 3.

[0284] Referring to Figure 3, the operating server (100) processes the time series data into frequency domain-based time series data corresponding to a specified frequency characteristic through the process of Figure 2, and then

[0286] Preprocessing is performed to construct multiple analysis / prediction training data for M subjects, including M subjects' n' (1 ≤ N' ≤ N, n' ⊂ N') time series data for M subjects, which are necessary for the analysis or prediction of a specified subject among the n time series data processed in the frequency domain (300).

[0288] Here, the N' time series data includes n' time series data required for a specified analysis or prediction among the n time series data processed into the frequency domain, and may include i (1 ≤ i ≤ (Nn)) time series data required for a specified analysis or prediction among the (Nn) time series data.

[0290] Additionally, the preprocessing may include at least one procedure among a data modification procedure that changes a character or category attribute value of at least one time series data included in the N' time series data into numeric data, and a data scaling adjustment procedure that adjusts the range difference between each time series data included in the N' time series data to within a specified range.

[0292] In addition, the above preprocessing may include a procedure for processing to derive the result value of the next data through k (k≥1) previous data by uniformizing the periodicity of at least one time series data included in the N' time series data into a fixed time period of a specified time range.

[0294] When M time series data (1≤N'≤N, n'⊂N') for each of the M (M≥1) time series data (1≤N'≤N, n'⊂N') for each of the M time series data (100) for each of the M (M≥1) time series data (1≤N'≤N, n'⊂N') for each of the M305) for each of the M time series data (1≤N'≤N, n'⊂N') for each of the M time series data (1≤N'≤N, n'⊂N') for each of the M time series data (1≤N'≤N, n'⊂N') for each of the M time series data (1≤N'≤N, n'⊂N') for each of the M time series data (1≤N'≤N, n'⊂N') for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time series data for each of the M time

[0296] Here, the analysis / prediction learning data may include the n' time series data as data corresponding to observation features or feature vectors for artificial intelligence learning, and may include j (i≥1) data corresponding to observation results or labels associated with the n' time series data.

[0298] Additionally, the analysis / prediction learning data may include n' time series data and (n'+i) time series data including i (1≤i≤(Nn)) time series data required for a specified analysis or prediction among (Nn) time series data as data corresponding to observation features or feature vectors for artificial intelligence learning, and may include j (i≥1) data corresponding to observation results or labels related to the specified time series data among the (n'+i) time series data.

[0300] When the above multiple analysis / prediction learning data is configured, the operating server (100) uses the configured multiple analysis / prediction learning data for M targets to train M targets for analysis or prediction (310).

[0302] Afterward, the operating server (100) checks m time series data for m targets, including n' time series data for m targets that are collected and preprocessed through the process of Figure 2 and processed into a designated frequency domain, by inputting the n' time series data for m targets (1≤m≤M) into m artificial intelligence modules to analyze or predict m feature information for m targets (315).

[0304] Here, the above feature information may include feature information corresponding to the result of an AI-based analysis or prediction using frequency domain-based time series data, and feature information corresponding to the result of an AI-based analysis or prediction using multiple frequency domain-based time series data and time domain-based time series data.

[0306] Afterwards, the operating server (100) stores the identified m feature information in a designated storage medium (320) and provides the identified m feature information to a designated user terminal (140) associated with the site (325).

[0308] Figure 4 is a flowchart illustrating the process of controlling a device at a site by training a control artificial intelligence module for controlling a device at a site according to the method of implementing the present invention, inputting feature information into the control artificial intelligence module to verify control information for controlling the device at a site, and controlling the device at a site.

[0310] More specifically, Figure 4 illustrates a process in which, after the process of Figure 3, a plurality of control learning data including control information for controlling a field device related to feature information are configured in an operating server (100) to train a control artificial intelligence module for controlling a field device equipped with the feature information, and when feature information that analyzes or predicts a designated target is confirmed through the artificial intelligence module, the confirmed feature information is input into the control artificial intelligence module to confirm control information for controlling a field device equipped with the feature information, and the field device is controlled using the confirmed control information. A person skilled in the art to which the present invention belongs may infer various implementation methods for the above process (e.g., implementation methods in which some steps are omitted or the order is changed) by referring to and / or modifying Figure 4, but the present invention is formed by including all such inferred implementation methods, and its technical features are not limited only to the implementation method illustrated in Figure 4.

[0312] Referring to Figure 4, when the operating server (100) confirms m feature information that analyzes or predicts m targets through the process of Figure 3, it configures a plurality of control learning data (400) that includes the confirmed m feature information and e control information that is set to control e (1 ≤ e ≤ E) devices among E (E ≥ 1) devices equipped at the service site that are related to the m feature information.

[0314] Here, the operating server (100) provides the identified m feature information to a designated expert terminal, and the expert can verify e control information set to control e devices related to the m feature information among the E devices equipped at the service site.

[0316] When the above multiple control learning data are configured, the operating server (100) trains a control artificial intelligence module for controlling the device provided at the site (405).

[0318] Here, the operating server (100) can control e devices equipped at the site using the e control information, and when the e devices equipped at the site are controlled using the e control information, it can check m actual measurement information for each target that is currently or after a certain period of time has elapsed regarding the combination of m feature information related to the control of the e devices equipped at the site and the e control information, and compare and analyze the m feature information and the m actual measurement information for each target to generate feedback information corresponding to the control information for improving efficiency related to the control of the e devices, thereby configuring control learning data including the m feature information and the generated feedback information to further train the artificial intelligence module for control.

[0320] After the above-mentioned control artificial intelligence module is learned, the operating server (100) checks m feature information that analyzes m targets or predicts m targets through m artificial intelligence modules designated through the process of Figures 2 and 3 (410).

[0322] And, the operating server (100) inputs the m feature information into the control artificial intelligence module and checks the e control information for controlling the e devices equipped at the site (415).

[0324] When the above e control information is confirmed, the operating server (100) controls the e devices equipped at the site using the above e control information (420).

[0326] When e devices equipped at the site are controlled using e control information confirmed through the above-mentioned artificial intelligence module for control, the operating server (100) checks m actual measurement information for each target that is currently or after a certain period of time has e, regarding the combination of m feature information related to the control of e devices equipped at the site and e control information (425).

[0328] Here, the actual measurement information may include information on at least one object among production volume, production quality, occurrence of anomalies, occurrence of failures, occurrence of environmental damage, and occurrence of pests and diseases, corresponding to the result of controlling e devices using the e control information.

[0330] And, the operating server (100) compares and analyzes the m feature information and the m actual measurement information for each confirmed target to generate feedback information corresponding to the control information for improving efficiency related to the control of the e devices (430).

[0332] Here, the feedback information may include e' control information obtained by retrospectively adjusting designated e' (1 ≤ e' ≤ e) control information among e control information obtained by controlling e devices equipped at the site to improve efficiency related to the control of the e devices.

[0334] When the above feedback information is generated, the operating server (100) configures control learning data including the m feature information and the generated feedback information to further train the control artificial intelligence module (435).

[0336] Here, the control learning data may include m feature information as data corresponding to observed features or feature vectors for artificial intelligence learning, and j (i≥1) data corresponding to observed results or labels including feedback information. Explanation of the symbols

[0338] 100 : Operation Server 105 : Data Configuration Section 110: Data Verification Unit 115: Data Processing Unit 120 : Artificial Intelligence Learning Unit 125 : Feature Information Verification Unit 130: Control Module Learning Unit 135: Control Information Verification Unit 140 : Device control unit 145 : Device 150 : User terminal

Claims

Claim 1 A method executed through an operating server comprises: a first step of collecting and preprocessing D (1≤D≤(S+C)) collected data selected from a set of collected data consisting of S sensing data, which are time-series sensing data for each sensor sensed chronologically through S (S≥1) sensors provided at a service site, and C image data, which are time-series image data for each camera captured chronologically through C (C≥1) cameras provided at the site, to form N (1≤N≤D) time-series data modeled to be matchable within an allowable range with T (T≥1) time-series models related to the time-series nature of a process or growth performed at the site; and a second step of verifying t (1≤t≤T) time-series models having periodicity matchable with a designated periodicity related to at least one process or growth performed at the site, and n (1≤n≤N) time-series data having periodicity matchable within a designated allowable range, based on the N time-series data. A third step of processing the aforementioned identified n time series data into frequency domain-based time series data corresponding to specified frequency features; a fourth step of configuring multiple analysis / prediction training data for M targets of a specified structure, comprising n' (1 ≤ n' ≤ n) time series data required for analysis or prediction for M (M ≥ 1) targets requiring analysis or prediction at the service site among the n time series data processed into the frequency domain, and including N' (1 ≤ N' ≤ N, n' ⊂ N') time series data for M targets of the same structure, and training M artificial intelligence modules provided for analysis or prediction for M targets using the configured multiple analysis / prediction training data for M targets.Step 5: After training the M artificial intelligence modules, input N' time series data for each of the m targets, which includes n' time series data for each of the m targets collected and preprocessed through the processes corresponding to Steps 1 to 3 and processed into the frequency domain among the n time series data for each of the m (1 ≤ m ≤ M) targets, into the designated m artificial intelligence modules to analyze or predict m feature information for m targets; Step 6: construct a plurality of control learning data that includes the confirmed m feature information and e control information set to control e (1 ≤ e ≤ E) devices related to the m feature information among E (E ≥ 1) devices equipped at the service site, thereby training a control artificial intelligence module for controlling devices equipped at the site; if, after training the control artificial intelligence module, m feature information for analyzing or predicting m targets is confirmed through the designated m artificial intelligence modules through the processes corresponding to Steps 1 to 5, input the m feature information into the control artificial intelligence module to confirm e control information for controlling e devices equipped at the site. A service site control method using time-series data comprising: Step 7; and Step 8, performing a procedure to control e devices provided at the site using the e control information. Claim 2 A service site control method using time series data according to claim 1, wherein the preprocessing includes a procedure for averaging the collected data having a specified periodicity among the D collected data to reveal the periodic pattern or representativeness of a certain time interval. Claim 3 A service site control method using time series data according to claim 1, wherein the preprocessing includes a procedure for correcting temporal influences corresponding to biases weighted by previous time data for data corresponding to a specified independent variable among the D collected data. Claim 4 A method for controlling a service site using time series data according to claim 1, wherein the T time series models include at least one time series model among a time series model that is matchable with the time series characteristics of a process performed through a device provided at a service site and a time series model that is matchable with the time series characteristics related to the growth of a crop cultivated at a service site. Claim 5 A service site control method using time series data according to claim 1, characterized in that the N time series data include time domain-based time series data based on the time series characteristics of collected data generated by time series sensing or time series shooting. Claim 6 A service site control method using time series data according to claim 1, wherein the periodicity comprises at least one periodicity among a periodicity corresponding to at least one process that is repeatedly performed at regular intervals through a device provided at the service site, a periodicity corresponding to a schedule related to at least one process to be performed through a device provided at the service site, a periodicity corresponding to a change in a daily interval (or Earth's rotation interval), a periodicity corresponding to a change in a trend of a specified period, a periodicity corresponding to a change in seasons, and a periodicity corresponding to a change in a one-year interval (or Earth's orbital interval). Claim 7 A method for controlling a service site using time series data according to claim 1, wherein the t time series models include a periodic time series model that includes a time series that matches the time series characteristics of a process or growth performed at the site, and includes a periodicity that matches the periodic characteristics of a process performed including a specified period or the periodic characteristics of growth affected by a specified change (or trend). Claim 8 A service site control method using time series data according to claim 1, wherein the second step comprises analyzing the N time series data based on at least one periodic time series model to identify n time series data having periodicity that can be matched within a specified allowable range with at least one specified periodic time series model. Claim 9 A service site control method using time series data according to claim 1 or 8, wherein the second step comprises, in the case of time series data where the data variation due to time variation is slower than a specified reference value, applying an exponential smoothing method specified to the time series data and then identifying n time series data having periodicity that can be matched within a specified allowable range with at least one periodic time series model. Claim 10 A method for controlling a service site using time series data, wherein, in claim 1 or 8, the second step comprises analyzing the N time series data based on at least one periodic time series model, and separately analyzing data corresponding to random elements or residual / remainder elements included in the time series data to identify n time series data having periodicity that can be matched with at least one periodic time series model within a specified allowable range. Claim 11 A service site control method using time series data according to claim 1, wherein the frequency feature includes a model-based period (or frequency) feature corresponding to a time series model having periodicity. Claim 12 A service site control method using time series data, wherein, in claim 1, the frequency feature includes a data-based period (or frequency) confirmed by reading time series data having periodicity. Claim 13 A service site control method using time series data according to claim 1, wherein the frequency feature includes a model-based period (or frequency) feature corresponding to a time series model having periodicity and a data-based period (or frequency) feature confirmed by reading time series data having periodicity and processing the data-based period (or frequency) feature confirmed by statistical processing or numerical analysis processing. Claim 14 A service site control method using time series data according to claim 1, wherein n time series data processed into the frequency domain includes time series data processed into the frequency domain corresponding to frequency characteristics that are matched within an allowable range and periodicity that matches the periodic characteristics of a process performed at the site including a specified period or the periodic characteristics of growth affected by a specified change (or trend). Claim 15 A service site control method using time series data according to claim 1, wherein the n' time series data comprises at least one time series data necessary for the analysis or prediction of a designated target among n time series data processed into a frequency domain corresponding to a frequency characteristic that is matched within an allowable range and periodicity matching the periodic characteristics of a process performed at the site including a designated period or the periodic characteristics of growth affected by a designated change (or trend). Claim 16 A service site control method using time series data according to claim 1, wherein the N' time series data includes n' time series data necessary for analysis or prediction of a designated target among n time series data processed into the frequency domain, and i (1 ≤ i ≤ (Nn)) time series data necessary for analysis or prediction of a designated target among (Nn) time series data. Claim 17 A service site control method using time series data according to claim 1, wherein the fourth step further includes a step of preprocessing the N' time series data so that they can be used as training data for a designated artificial intelligence module. Claim 18 A service site control method using time series data according to claim 17, wherein the preprocessing comprises at least one procedure among a data modification procedure for changing a character or category attribute value of at least one time series data included in the N' time series data into numeric data, and a data scaling adjustment procedure for adjusting the range difference between each time series data included in the N' time series data to within a specified predetermined range. Claim 19 A service site control method using time series data according to claim 17, wherein the preprocessing includes a procedure for processing to equalize the periodicity of at least one time series data included in the N' time series data into a fixed time period of a specified time range so as to enable the derivation of the result value of the next data through k (k≥1) previous data. Claim 20 A method for controlling a service site using time series data according to claim 1, wherein the analysis / prediction learning data includes the n' time series data as data corresponding to observation features or feature vectors for artificial intelligence learning, and includes j (i≥1) data corresponding to observation results or labels related to the n' time series data. Claim 21 A method for controlling a service site using time series data according to claim 1, wherein the analysis / prediction learning data includes n' time series data and (n'+i) time series data including i (1≤i≤(Nn)) time series data necessary for the analysis or prediction of a designated target among (Nn) time series data, as data corresponding to an observation feature or feature vector for artificial intelligence learning, and includes j (i≥1) data corresponding to an observation result or label related to the designated time series data among the (n'+i) time series data. Claim 22 A service site control method using time series data according to claim 1, wherein the fifth step further includes a step of preprocessing N' time series data for each of the M targets so that they can be input into the learned M artificial intelligence modules. Claim 23 A service site control method using time series data according to claim 1, wherein the feature information includes feature information corresponding to the result of an artificial intelligence-based analysis or prediction using frequency domain-based time series data, and feature information corresponding to the result of an artificial intelligence-based analysis or prediction using multiple frequency domain-based time series data and time domain-based time series data. Claim 24 A service site control method using time series data according to claim 1, wherein the sixth step further comprises the step of performing a procedure to control e devices equipped at the site using the e control information. Claim 25 In claim 1, the sixth step further comprises: a step of verifying m actual measurement information for each target, measured from the present to after a certain period of time, regarding the combination of m feature information related to the control of e devices provided at the site and e control information when e devices provided at the site are controlled using e control information; a step of generating feedback information corresponding to control information for efficiency improvement related to the control of e devices by comparing and analyzing the m feature information and the verified m actual measurement information for each target; and a step of additionally training the control artificial intelligence module by configuring control learning data including the m feature information and the generated feedback information. Claim 26 A service site control method using time-series data according to claim 25, wherein the actual measurement information includes information on actual measurements of at least one target among production volume, production quality, occurrence of anomalies, occurrence of failures, occurrence of environmental damage, and occurrence of pests and diseases, corresponding to the result of controlling e devices using the e control information. Claim 27 A service site control method using time series data according to claim 25, wherein the feedback information comprises e' control information that retrospectively adjusts e' (1≤e'≤e) control information among e control information that controls e devices equipped at the site to improve efficiency related to the control of e devices. Claim 28 A service site control method using time series data according to claim 1, wherein the control learning data includes m feature information as data corresponding to an observation feature or feature vector for artificial intelligence learning, and includes j (i≥1) data corresponding to an observation result or label including e control information. Claim 29 In claim 1, the above-mentioned eighth step further comprises: a step of verifying m actual measurement information for each target, measured from the present to after a certain period of time, regarding the combination of m feature information related to the control of e devices provided at the site and e control information when e devices provided at the site are controlled using e control information confirmed through the control artificial intelligence module; a step of generating feedback information corresponding to control information for efficiency improvement related to the control of e devices by comparing and analyzing the m feature information and the verified m actual measurement information for each target; and a step of further training the control artificial intelligence module by configuring control learning data including the m feature information and the generated feedback information.

Images