CN112288021A - Medical wastewater monitoring data quality control method, device and system - Google Patents
Medical wastewater monitoring data quality control method, device and system Download PDFInfo
- Publication number
- CN112288021A CN112288021A CN202011203309.2A CN202011203309A CN112288021A CN 112288021 A CN112288021 A CN 112288021A CN 202011203309 A CN202011203309 A CN 202011203309A CN 112288021 A CN112288021 A CN 112288021A
- Authority
- CN
- China
- Prior art keywords
- data
- sequence
- abnormal
- algorithm
- monitoring data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/18—Water
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Chemical & Material Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Food Science & Technology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a quality control method, a device and a system for medical wastewater monitoring data, wherein the method comprises the steps of collecting monitoring data of the whole medical wastewater treatment process; classifying and identifying the acquired monitoring data to obtain different data types; and matching the corresponding abnormal detection algorithm and parameters according to the identified data type to obtain abnormal point data. The method is based on the monitoring data of the existing medical wastewater whole-process big data platform, and intelligent quality control is realized. The abnormal point positions are accurately found out by utilizing the algorithm, the workload of data quality auditing specialists is reduced, the problem that manual auditing is easy to miss is solved, and the data quality auditing efficiency is improved; compare than artifical anomaly detection of tradition, based on the unique waste water treatment discharge pattern of different hospitals, carry out intelligent modeling and match, can realize the automatic identification and the inspection of different grade type monitoring data based on the different monitoring index of every hospital, intelligence efficient realizes the anomaly detection of different indexes.
Description
Technical Field
The invention relates to the field of medical wastewater supervision, in particular to a medical wastewater monitoring data quality control method, device and system.
Background
In the field of medical wastewater supervision, most of the medical wastewater is subjected to water quality inspection by adopting a manual sampling mode at present, and most of the medical wastewater has serious defects in intermediate treatment.
For the current situation that the water quality inspection and sampling of the wastewater are still in a manual monitoring stage, some operation and maintenance mechanisms begin to adopt an automatic monitoring technology to carry out automatic detection on the operation and maintenance treatment of the medical wastewater. Compared with other automatic monitoring projects, the automatic monitoring of the medical wastewater has the advantages that the water quality is changed greatly due to special environment, data can be influenced by more factors, for example, the water quality in the medical wastewater is poor, the wastewater contains more large-particle impurities, the natural factor interference of a discharge port and the interference factor are considered, and the occurrence of the factors is very easy to cause abnormal data generation and influence on the data quality. Abnormal data may possibly affect the result of data mining analysis performed by technicians, and misjudgment may be generated in severe cases. Traditional sampling data, because the environment stability in the sampling process can be better guaranteed to the mode of adopting artifical sampling, the data bulk of artifical sampling is less moreover, consequently often can accomplish the basic quality control that detects data through the mode of artifical inspection affirmation. At present, the method of combining manual review and on-site investigation is still remained in the aspect of data quality control in most domestic small and medium-sized data service type enterprises. If the abnormal data of the small sample is checked, manual review can be effective in a short time, but the manual review inevitably causes omission along with the increase of the data volume.
The traditional medical wastewater data quality inspection often depends on technicians, so that the requirements on the professional level and the service familiarity of the technicians are high. However, in many cases, enterprise data quality auditors are trained and then put on duty, and differences in professional levels of technicians have certain influence on data auditing results, and it cannot be guaranteed that data quality is completely over-critical. In addition, most data monitored by data service type enterprises may have more dimensions, and as time increases, the data dimensions increase, more manpower and material resources need to be invested in numerical quality auditing, and the cost also increases. However, the manual auditing is not absolutely necessary, and even if more manpower is invested, it is difficult to continuously and uninterruptedly find out all the anomalies.
The major difficulties encountered by enterprises in data quality auditing are as follows:
firstly, certain requirements are required for the professional level of relevant data auditors, professional knowledge is required, data sensitivity is high, and the business is familiar, so that a large amount of energy and material resources are required to be invested in the early stage of an enterprise for professional and business training.
And secondly, when large data are faced, a large amount of manpower time is required to be invested for auditing and on-site rechecking and troubleshooting. In order to pursue timeliness, data quality auditing needs to be completed in a short time. However, omission is extremely easy to occur manually in the case of rapid review of data.
And thirdly, as the time increases, the data scale is inevitably increased, the original manpower and material resources cannot meet the requirement of data quality examination, and if the data quality examination is continuously carried out in a manual examination mode, the cost is also increased.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a medical wastewater monitoring data quality control method, device and system, so that intelligent quality control is realized on the monitoring data of the whole medical wastewater treatment process, abnormal point data are accurately and efficiently found, and operation and maintenance personnel can timely perform field maintenance.
In order to achieve the purpose, the technical scheme of the invention is as follows:
in a first aspect, an embodiment of the present invention provides a medical wastewater monitoring data quality control method, including:
collecting monitoring data of the whole medical wastewater treatment process;
classifying and identifying the acquired monitoring data to obtain different data types;
and matching the corresponding abnormal detection algorithm and parameters according to the identified data type to obtain abnormal point data.
Further, the medical wastewater monitoring data quality control method further comprises the following steps:
and pushing the obtained abnormal point data to the target object.
Further, classifying and identifying the acquired monitoring data to obtain different data types;
the method comprises the steps that fast Fourier transform is applied to acquired monitoring data, different acquired data signals are transformed from a time domain to a frequency domain form, whether the acquired data are periodic or not is confirmed according to the energy distribution condition of frequency in a spectrogram after data transformation, and the period of a data sequence is obtained according to the N/frequency of the acquired data, namely the acquired data are identified as periodic data;
and for non-periodic data, the unit root test in the time sequence data is passed, if the data sequence passes the test, the data sequence is identified as non-periodic stationary data, and if the data sequence does not pass the test, the data sequence is identified as non-periodic irregular fluctuating data.
Further, for aperiodic stationarity data, adopting an orphan forest algorithm to identify abnormal data in operation and maintenance:
when constructing the solitary forest subtree, firstly sampling a data set by a system sampling method to construct a sub-forest, and forming the sub-forest into a base forest abnormity detector; judging the abnormal condition of data entering a sliding window through a base forest abnormal detector; and (4) according to the history, judging whether the data volume to be detected and the abnormal rate of the sliding window data exceed a threshold value, and if so, determining abnormal point data.
Further, for the periodic data, a Prophet-AE-LSTM integration algorithm is adopted to monitor the abnormality of the operation and maintenance data:
predicting the selected historical dependence data by using a Prophet algorithm to obtain a predicted value;
solving a Prophet predicted value and an actual residual error so as to carry out white noise detection on a predicted residual error sequence;
and if the residual sequence passes through white noise test, indicating that the related information in the sequence is extracted, comparing the Prophet predicted value with the model prediction confidence interval, and if the Prophet predicted value exceeds the prediction confidence interval, determining the predicted point as an abnormal data point.
If the residual sequence does not pass white noise test, indicating that the related information remained in the sequence is not extracted, extracting the predicted residual and carrying out next-stage training, namely predicting the residual by adopting a multi-feature superimposed AE-LSTM algorithm; splicing the features learned in the AE algorithm encoder with other related collected features, and then inputting the spliced features into an LSTM model for sequence prediction of a residual error item; compressing lag term variables of an original sequence by adopting an AE algorithm, splicing the compressed characteristics with other collected related variables and prophet predicted values to form new input characteristics, training residual terms still containing information, and obtaining predicted values of the residual terms;
and adding the Prophet predicted value and the residual predicted value obtained in the previous step to obtain a final predicted value, and if the predicted value exceeds the experience threshold value, obtaining abnormal point data.
The expression of the Prophet algorithm is as follows:
y(t)＝g(t)+s(t)+h(t)+εt
wherein g (t) represents a trend term in the time series, s (t) represents a period term in the time series, h (t) represents the potential influence brought by holidays, epsilontAnd representing a model error term reflecting an abnormal variation not represented in the model.
Further, the EMD-AE-LSTM algorithm is adopted to carry out anomaly prediction on the non-periodic irregular fluctuation data:
firstly, decomposing original sequence data by using EMD (empirical mode decomposition) to obtain a component IMF (intrinsic mode function) of each sequence;
respectively constructing EM-LSTM network prediction for each IMF component on the basis of EMD empirical mode decomposition, splicing the learned characteristics of an AE algorithm encoder from sequence lag variables with other related acquisition characteristics, and inputting the spliced characteristics into an LSTM model for sequence prediction;
and further comparing residual errors obtained according to the prediction result and the actual value with a threshold value, and if the residual errors exceed the threshold value, obtaining abnormal point data.
In a second aspect, an embodiment of the present invention provides a medical wastewater monitoring data quality control system, including:
the data acquisition module is used for acquiring monitoring data of the whole medical wastewater treatment process;
the data classification module is used for classifying and identifying the acquired monitoring data to obtain different data types;
the data analysis module is used for matching a corresponding abnormal detection algorithm and parameters according to the identified data type to obtain abnormal point data;
and the data pushing module is used for pushing the obtained abnormal point data to the target object.
In a third aspect, the embodiment of the present invention provides a medical wastewater monitoring data quality control device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.
In a fourth aspect, the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the computer program implements the steps of the method described above.
Compared with the prior art, the invention has the beneficial effects that:
the medical wastewater monitoring data quality control method provided by the embodiment is based on the monitoring data of the existing medical wastewater whole-process big data platform, and intelligent quality control is realized. The abnormal point positions are accurately found out by utilizing the algorithm, the workload of data quality auditing specialists is reduced, the problem that manual auditing is easy to miss is solved, and the data quality auditing efficiency is improved; compare than artifical anomaly detection of tradition, based on the unique waste water treatment discharge pattern of different hospitals, carry out intelligent modeling and match, can realize the automatic identification and the inspection of different grade type monitoring data based on the different monitoring index of every hospital, intelligence efficient realizes the anomaly detection of different indexes to adapt to different grade type data, realize many indexes and detect simultaneously, can discover fast that data takes place the unusual condition simultaneously.
Drawings
FIG. 1 is a flow chart of a medical wastewater monitoring data quality control method according to embodiment 1 of the present invention;
FIG. 2 is a graphical illustration of stationarity data;
FIG. 3 is a schematic of periodic data;
FIG. 4 is a schematic diagram of non-periodic irregularity data;
FIG. 5 is a flow chart of the EMD-AE-LSTM algorithm;
FIG. 6 is a schematic diagram of the structure of an LSTM network;
FIG. 7 is a schematic diagram illustrating the construction of a medical wastewater monitoring data quality control system according to example 2 of the present invention;
fig. 8 is a schematic composition diagram of a medical wastewater monitoring data quality control device according to embodiment 3 of the present invention.
Detailed Description
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1:
referring to fig. 1 to 6, the method for quality control of medical wastewater monitoring data provided by the present embodiment specifically includes:
101. monitoring data of the whole medical wastewater treatment process are obtained based on IOT (Internet of Things) monitoring equipment of the whole hospital wastewater treatment process.
102. And classifying and identifying the acquired monitoring data to obtain different data types.
103. Matching a corresponding anomaly detection algorithm and parameters according to the identified data type to obtain anomaly point data; namely, the automatic identification and inspection of different types of monitoring data can be realized based on different monitoring indexes of each hospital, and the anomaly detection of different indexes can be realized intelligently and efficiently, so that the method is suitable for different types of data and realizes the simultaneous detection of multiple indexes.
Therefore, the method realizes intelligent quality control based on the monitoring data of the existing medical wastewater whole-process big data platform. The abnormal point positions are accurately found out by using an algorithm, the workload of data quality auditing specialists is reduced, the problem that manual auditing is easy to miss is solved, and the data quality auditing efficiency is improved; compare than artifical anomaly detection of tradition, based on the unique waste water treatment discharge pattern of different hospitals, carry out intelligent modeling and match, can realize the automatic identification and the inspection of different grade type monitoring data based on the different monitoring index of every hospital, intelligence efficient realizes the anomaly detection of different indexes to adapt to different grade type data, realize many indexes and detect simultaneously, can discover fast that data takes place the unusual condition simultaneously.
As a preferable aspect of this embodiment, the method for quality control of medical wastewater monitoring data further includes:
104. pushing the obtained abnormal point data to relevant engineering inspection personnel; therefore, the abnormal starting time and the abnormal ending time cannot be accurately mastered, technicians can conveniently analyze the abnormal data by combining with business experience, the abnormal reason is found, and the operation and maintenance personnel are guided to carry out field maintenance in time.
Specifically, in this embodiment, the monitoring data in step 101 is device monitoring data of a historical time period adjacent to the time period to be measured, and a monitoring data set of hospital data is generated; the time interval to be measured is a continuous time interval, the field of the data set mainly comprises indexes such as date, hospital type, water discharge, residual chlorine and ph, the data set comprises historical time interval monitoring data, and the sampling frequency of each parameter is 2 hours.
Specifically, in this embodiment, the step 102 includes:
as shown in fig. 2-4, because the types of the data collected in step 101 are complicated, for different data in this complicated case, in this example, fast fourier transform (fft) is firstly applied to transform different collected data signals from time domain to frequency domain, and from the energy distribution of frequency in the spectrogram after data transformation, whether the collected data has periodicity is determined, and meanwhile, the period of the data sequence is obtained from the collection number N/frequency (HZ), i.e. the data sequence is identified as periodic data; secondly, for the non-periodic data, the unit root inspection (ADF) in the time sequence data is passed, if the data sequence passes the inspection, the non-periodic stationary data is identified, and if the data sequence does not pass the inspection, the non-periodic irregular fluctuating data (i.e. the irregular toggle data in fig. 1) is identified.
Specifically, in this embodiment, the step 103 includes:
for the non-periodic stationary data, the embodiment identifies abnormal data in operation and maintenance by using the solitary forest algorithm. When constructing the solitary forest subtree, firstly sampling a data set by a system sampling method to construct a sub-forest, and forming the sub-forest into a base forest abnormity detector; judging the abnormal condition of data entering a sliding window through a base forest abnormal detector; according to the history, whether the data volume to be detected and the abnormal rate of the sliding window data exceed a threshold value or not is judged as abnormal point data; selecting a model updating strategy with a smaller or larger updating proportion; and calculating the abnormal rate difference value of each sub-forest and the base forest based on the updated data set, removing the sub-forests with larger difference values, constructing a plurality of sub-forests for supplement, forming a new base forest abnormal detector, and realizing the continuous update of the solitary forest algorithm.
For the periodic data, the embodiment adopts a Prophet-AE (-automatic encoder) -LSTM integration algorithm to perform anomaly monitoring on the operation and maintenance data, and specifically includes the following steps:
firstly, predicting selected historical dependence data by using a Prophet algorithm to obtain a predicted value;
then, the predicted value of Prophet and the actual residual error are obtained, and white noise test is further carried out on the predicted residual error sequence.
If the residual error passes through white noise inspection, the fact that the related information in the sequence is extracted is shown, namely the Prophet predicted value is compared with the model prediction confidence interval, and if the Prophet predicted value exceeds the prediction confidence interval, the prediction point is abnormal.
If the residual error does not pass white noise test, indicating that the related information remained in the sequence is not extracted, extracting the predicted residual error for next-stage training, namely predicting the residual error by adopting an AE-LSTM algorithm (self-encoder-long short-term memory model) with multi-feature superposition; the AE algorithm is a typical feature representation learning method, and a hidden layer in the network can accurately and efficiently represent the core features of sequence data. In the model of this embodiment, the features learned in the encoder are spliced with other related collected features, and then input into the LSTM model for sequence prediction of the residual term. In this embodiment, an AE algorithm is mainly used to compress lag term variables of an original sequence, and the compressed features are spliced with other acquired related variables and prophet predicted values to form new input features, and residual terms still containing information are trained to obtain predicted values of the residual terms.
And finally, adding the Prophet predicted value and the residual predicted value obtained in the previous step to obtain a final predicted value. If the predicted value exceeds a certain model experience threshold, the pushing is abnormal.
For non-periodic irregular fluctuation data, the embodiment adopts an idea based on decomposition-integration to firstly predict the data to be detected by adopting an EMD-AE-LSTM algorithm, and identifies abnormal data in the type according to an idea of comparing a residual error between a predicted value and an actual value with a threshold value. Firstly, decomposing original sequence data by using EMD (empirical mode decomposition) to obtain a component IMF (intrinsic mode function) of each sequence; respectively constructing EM-LSTM network prediction for each IMF component on the basis of EMD empirical mode decomposition, splicing the features learned by an AE algorithm encoder from sequence lag variables with other related collected features, and inputting the spliced features into an LSTM model for sequence prediction; and further comparing residual errors obtained according to the prediction result and the actual value with a threshold value, and if the residual errors exceed the threshold value, obtaining abnormal point data. The basic idea of Empirical Mode Decomposition (EMD) is to reduce sequence complexity by decomposing an original sequence into a plurality of IMF components with lower complexity, i.e., an EMD algorithm can decompose a complex signal into a plurality of Intrinsic Mode Functions (IMFs) and 1 Trend sequence. The expression is as follows:
wherein s istAs the original signal, imfi(t) is the ith IMF component, rn(t) is a trend term.
As shown in fig. 5, the present embodiment separately constructs an AE-LSTM network prediction model for relatively smooth but still highly fluctuating IMF components based on empirical mode decomposition. The accuracy and the effectiveness of the prediction model are guaranteed for the next step of comparison with the residual error threshold value while the accuracy of the prediction model is improved.
Specifically, for the Prophet model for predicting time series in the above expression, the rules such as trend, periodicity, change of special date and the like in the data series can be captured, and the future can be predicted on the basis of the rule change based on historical data and the assumption that the user will continue the rules in the future. The expression of the Prophet model is as follows:
y(t)＝g(t)+s(t)+h(t)+εt
wherein g (t) represents a trend term in the time series, s (t) represents a period term in the time series, h (t) represents the potential influence brought by holidays, epsilontAnd representing a model error term reflecting an abnormal variation not represented in the model.
For the self-encoder (AE) in the above expression, which is a neural network that makes output values equal to input values using a back-propagation algorithm, the input is first compressed into a latent spatial representation, and the output is then reconstructed from this representation. The self-encoder consists of two parts: first, the encoder, which can compress the input into a latent spatial representation, can be represented by an encoding function h ═ f (x); second, the decoder, which can reconstruct the input from the potential spatial representation, can be represented by the decoding function r ═ g (h). By training the auto-encoder with output values equal to input values, the potential tokens h will have value attributes, which are also expected intrinsic features, which is also a typical representation learning and feature learning method and is often applied in the case of actual training samples without labels.
For the LSTM model in the above expression, i.e. the long-short term memory network model, the model is a special recurrent neural network that is specific to the sequence problem. Compared with the common recurrent neural network, the LSTM network introduces the concept of "gate" instead of the traditional neuron, so that it can control the long-term memory state. The structure of the LSTM network is schematically shown in fig. 6.
In an LSTM network, a "gate" is a screening structure of information, i.e., an information weight. The method carries out dot product operation on input information through a weight matrix, then controls an output value in a [0,1] interval through a sigmoid function, outputs '0' to represent that all information is discarded, and outputs '1' to represent that the information is completely passed. The LSTM model consists mainly of an input gate, a forgetting gate, an output gate, and cell states. Wherein the input gate is used for controlling how much information input at the current moment can be added into the cell state; the forgetting gate determines how much information in the cell state at the last moment can be transmitted to the current moment; the output gate finally outputs the result based on the cell states updated by the forgoing gate and the input gate. The cell state mainly records the current input, the state of the hidden layer at the previous moment, the cell state at the previous moment and the information in the gate structure. The specific calculation steps of the LSTM algorithm model are as follows:
ft＝σ(Wf·[ht-1,xt]+bf)
it＝σ(Wi·[ht-1,xt]+bi)
ot＝σ(Wo·(ht-1,xt)+bo)
ht＝ot·tanh(Ct)
wherein h ist-1Is the output at time t-1 of the hidden layer, xtIs an input at time t, WfIs the weight matrix of the forgetting gate, bfIs the offset of the forgetting gate, σ () is the sigmoid function; wiIs a weight matrix of the input gate, biIs the offset of the input gate;
is candidate information extracted from the input layer at time t, WcIs a weight matrix of the input layer, bcIs the offset of the input gate; ctIs cell unit status information at time t, Ct-1Is cell unit status information at time t-1; woIs a weight matrix of output gates, boIs the offset of the output gate; otIs the output gate information weight, htIs the final output information. In summary, the LSTM network may implement the extraction of information through three "gates" such thatThe long-term information is easier to transmit in the network, and the problem of gradient disappearance caused by long-distance dependence in the traditional recurrent neural network is avoided.
When the method is specifically applied to abnormality detection, the method comprises the following steps:
(1) model training, pattern recognition: based on unique monitoring data of each hospital, model training is conducted on the administration operation and maintenance data of the hospital, and the conversion from field monitoring and manual auditing to intelligent identification of data inspection operation and maintenance is achieved;
(2) model deployment, exception identification: the method comprises the steps that data are collected through networking and loaded to a trained intelligent recognition model, the model firstly carries out pattern matching based on data characteristics, and then carries out abnormal knowledge alarm based on the data pattern through deviation degree;
(3) and remote checking of the alarm information: the medical data for intelligent operation and maintenance is subjected to 24-hour online automatic intelligent quality inspection, and operation and maintenance personnel only need to check and judge alarm information to realize intelligent control and comprehensive information intelligentization.
(4) Model retraining: the intelligent control algorithm comprises a retraining module, and retraining can be performed based on a new data model based on a set retraining threshold value (15 days, 30 days and 60 days), so that the model can be matched with a new mode of data to perform abnormal recognition detection, and comprehensive automatic data quality is realized.
In conclusion, compared with the prior art, the medical wastewater monitoring data quality control method provided by the embodiment has the following technical advantages:
1. the data quality is guaranteed in 24 hours, a large amount of manpower and material resources are saved, and the abnormity detection rate is improved.
2. By adopting the intelligent detection algorithm, the 24-hour quality control guarantee of the monitoring data can be realized, and through carrying out mode identification processing on the automatic monitoring data, quality testing personnel do not need to pass field inspection and 24-hour manual visual data examination, the algorithm can automatically push out the suspected abnormal condition, and only needs to carry out investigation and analysis on the abnormal pushing result, so that manpower and material resources are greatly saved;
3. by adopting the intelligent detection algorithm, the data validity of the real-time data can be detected, the traditional quality inspection time in days is shortened to 2 hours, and the effective rate of operation and maintenance is greatly improved;
4. the traditional data anomaly detection is only used for monitoring data of a certain type by using a specific algorithm, but in practical application, different data indexes of different industries are possibly inconsistent, so that a single specific algorithm cannot be well suitable for anomaly detection of different types of data, and the medical wastewater monitoring quality control intelligent algorithm system well solves the problem, realizes automatic classification of medical wastewater data, and automatically calls a corresponding algorithm to realize intelligent detection.
Example 2:
referring to fig. 7, the present embodiment provides a medical wastewater monitoring data quality control system, which includes:
a data acquisition module 701, configured to acquire monitoring data of an overall medical wastewater treatment process based on an IOT (Internet of Things) monitoring device for the overall hospital wastewater process, where the monitoring data is device monitoring data of a historical time period adjacent to a time period to be detected, and generate a monitoring data set of hospital data; the time interval to be measured is a continuous time interval, the field of the data set mainly comprises indexes such as date, hospital type, water discharge, residual chlorine and ph, the data set comprises historical time interval monitoring data, and the sampling frequency of each parameter is 2 hours.
And the data classification module 702 is configured to perform classification and identification on the acquired monitoring data to obtain different data types.
And the data analysis module 703 is configured to match a corresponding anomaly detection algorithm and parameter according to the identified data type to obtain anomaly point data.
And a data pushing module 704, configured to push the obtained abnormal point data to the target object.
Specifically, the working principle of the data classification module includes:
because the types of data acquired by the data acquisition module are complex and numerous, aiming at different data under the complex condition, the data classification module of the embodiment firstly applies fast fourier transform (fft) to convert different acquired data signals into a frequency domain form from a time domain, and determines whether the acquired data has periodicity from the energy distribution condition of frequency in a spectrogram after data conversion, and meanwhile, the acquired data is identified as periodic data by acquiring the period of a data sequence by an acquisition number N/frequency (HZ); secondly, for the non-periodic data, the unit root inspection (ADF) in the time sequence data is passed, if the data sequence passes the inspection, the data sequence is identified as non-periodic stationary data, and if the data sequence does not pass the inspection, the data sequence is identified as non-periodic irregular fluctuating data.
Specifically, in this embodiment, the working principle of the data analysis module includes:
for the non-periodic stationary data, the embodiment identifies abnormal data in operation and maintenance by using the solitary forest algorithm. When constructing the solitary forest subtree, firstly sampling a data set by a system sampling method to construct a sub-forest, and forming the sub-forest into a base forest abnormity detector; judging the abnormal condition of data entering a sliding window through a base forest abnormal detector; according to the history, whether the data volume to be detected and the abnormal rate of the sliding window data exceed a threshold value or not is judged as abnormal point data; selecting a model updating strategy with a smaller or larger updating proportion; and calculating the abnormal rate difference value of each sub-forest and the base forest based on the updated data set, removing the sub-forests with larger difference values, constructing a plurality of sub-forests for supplement, forming a new base forest abnormal detector, and realizing the continuous update of the solitary forest algorithm.
For the periodic data, the embodiment adopts a Prophet-AE (-automatic encoder) -LSTM integration algorithm to perform anomaly monitoring on the operation and maintenance data, and specifically includes the following steps:
firstly, predicting selected historical dependence data by using a Prophet algorithm to obtain a predicted value;
then, the predicted value of Prophet and the actual residual error are obtained, and white noise test is further carried out on the predicted residual error sequence.
If the residual sequence passes white noise inspection, the related information in the sequence is extracted, namely the Prophet predicted value is compared with the model prediction confidence interval, and if the Prophet predicted value exceeds the prediction confidence interval, the prediction point is abnormal.
If the residual error does not pass white noise test, indicating that the related information remained in the sequence is not extracted, extracting the predicted residual error for next-stage training, namely predicting the residual error by adopting an AE-LSTM algorithm (self-encoder-long short-term memory model) with multi-feature superposition; the AE algorithm is a typical feature representation learning method, and a hidden layer in the network can accurately and efficiently represent the core features of sequence data. In the model of the invention, the features learned in the encoder are spliced with other related collected features and then input into the LSTM model for sequence prediction of residual error terms. According to the invention, an AE algorithm is mainly adopted to compress lag term variables of an original sequence, and the compressed characteristics are spliced with other collected related variables and prophet predicted values to form new input characteristics to train residual terms still containing information, so as to obtain predicted values of the residual terms.
And finally, adding the Prophet predicted value and the residual predicted value obtained in the previous step to obtain a final predicted value. If the predicted value exceeds a certain model experience threshold, the push is abnormal
For non-periodic irregular fluctuation data, the embodiment adopts an idea based on decomposition-integration to firstly predict the data to be detected by adopting an EMD-AE-LSTM algorithm, and identifies abnormal data in the type according to an idea of comparing a residual error between a predicted value and an actual value with a threshold value. Firstly, decomposing original sequence data by using EMD (empirical mode decomposition) to obtain a component IMF (intrinsic mode function) of each sequence; respectively constructing EM-LSTM network prediction for each IMF component on the basis of EMD empirical mode decomposition, splicing the features learned by an AE algorithm encoder from sequence lag variables with other related collected features, and inputting the spliced features into an LSTM model for sequence prediction; and further comparing residual errors obtained according to the prediction result and the actual value with a threshold value, and if the residual errors exceed the threshold value, obtaining abnormal point data. The basic idea of Empirical Mode Decomposition (EMD) is to reduce sequence complexity by decomposing an original sequence into a plurality of IMF components with lower complexity, i.e., an EMD algorithm can decompose a complex signal into a plurality of Intrinsic Mode Functions (IMFs) and 1 Trend sequence. The expression is as follows:
wherein s istAs the original signal, imfi(t) is the ith IMF component, rn(t) is a trend term.
As shown in fig. 5, the present embodiment separately constructs an AE-LSTM network prediction model for relatively smooth but still highly fluctuating IMF components based on empirical mode decomposition. The accuracy and the effectiveness of the prediction model are guaranteed for the next step of comparison with the residual error threshold value while the accuracy of the prediction model is improved.
For the Prophet model for predicting the time series in the expression, the rules such as trend, periodicity, change of special dates and the like in the data series can be captured, and the future can be predicted on the basis of the rule change based on historical data and the assumption that the rules are continued by the user in the future. The expression of the Prophet model is as follows:
y(t)＝g(t)+s(t)+h(t)+εt
wherein g (t) represents a trend term in the time series, s (t) represents a period term in the time series, h (t) represents the potential influence brought by holidays, epsilontAnd representing a model error term reflecting an abnormal variation not represented in the model.
For the self-encoder (AE) in the above expression, which is a neural network that makes output values equal to input values using a back-propagation algorithm, the input is first compressed into a latent spatial representation, and the output is then reconstructed from this representation. The self-encoder consists of two parts: first, the encoder, which can compress the input into a latent spatial representation, can be represented by an encoding function h ═ f (x); second, the decoder, which can reconstruct the input from the potential spatial representation, can be represented by the decoding function r ═ g (h). By training the auto-encoder with output values equal to input values, the potential tokens h will have value attributes, which are also expected intrinsic features, which is also a typical representation learning and feature learning method and is often applied in the case of actual training samples without labels.
For the LSTM model in the above expression, i.e. the long-short term memory network model, the model is characterized by being a special recurrent neural network that is specific to the sequence problem. Compared with the common recurrent neural network, the LSTM network introduces the concept of "gate" instead of the traditional neuron, so that it can control the long-term memory state. The structure of the LSTM network is schematically shown in fig. 6.
In an LSTM network, a "gate" is a screening structure of information, i.e., an information weight. The method carries out dot product operation on input information through a weight matrix, then controls an output value in a [0,1] interval through a sigmoid function, outputs '0' to represent that all information is discarded, and outputs '1' to represent that the information is completely passed. The LSTM model consists mainly of an input gate, a forgetting gate, an output gate, and cell states. Wherein the input gate is used for controlling how much information input at the current moment can be added into the cell state; the forgetting gate determines how much information in the cell state at the last moment can be transmitted to the current moment; the output gate finally outputs the result based on the cell states updated by the forgoing gate and the input gate. The cell state mainly records the current input, the state of the hidden layer at the previous moment, the cell state at the previous moment and the information in the gate structure. The specific calculation steps of the LSTM algorithm model are as follows:
ft＝σ(Wf·[ht-1,xt]+bf)
it＝σ(Wi·[ht-1,xt]+bi)
ot＝σ(Wo·(ht-1,xt)+bo)
ht＝ot·tanh(Ct)
wherein h ist-1Is the output at time t-1 of the hidden layer, xtIs an input at time t, WfIs the weight matrix of the forgetting gate, bfIs the offset of the forgetting gate, σ () is the sigmoid function; wiIs a weight matrix of the input gate, biIs the offset of the input gate;
is candidate information extracted from the input layer at time t, WcIs a weight matrix of the input layer, bcIs the offset of the input gate; ctIs cell unit status information at time t, Ct-1Is cell unit status information at time t-1; woIs a weight matrix of output gates, boIs the offset of the output gate; otIs the output gate information weight, htIs the final output information. In conclusion, the LSTM network can extract information through three gates, so that long-term information is easier to transmit in the network, and the problem of gradient disappearance caused by long-distance dependence in the traditional recurrent neural network is solved.
Example 3:
referring to fig. 8, the medical wastewater monitoring data quality control apparatus provided in this embodiment includes a processor 801, a memory 802, and a computer program 803, such as a medical wastewater monitoring data quality control processing program, stored in the memory 801 and executable on the processor 801. The processor 801, when executing the computer program 803, implements the steps of embodiment 1 described above, such as the steps shown in fig. 1. Alternatively, the processor 801 implements the functions of the modules in embodiment 2 when executing the computer program 803.
Illustratively, the computer program 803 may be partitioned into one or more modules/units that are stored in the memory 802 and executed by the processor 801 to implement the present invention. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program 803 in the medical wastewater monitoring data quality control device. For example, the computer program 803 may be partitioned into a data acquisition and data classification module.
The medical wastewater monitoring data quality control device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing equipment. The medical wastewater monitoring data quality control device can include, but is not limited to, a processor 801 and a memory 802. It will be understood by those skilled in the art that fig. 5 is merely an example of a medical wastewater monitoring data quality control device and does not constitute a limitation of the medical wastewater monitoring data quality control device, and may include more or fewer components than those shown, or some components in combination, or different components, for example, the medical wastewater monitoring data quality control device may further include input and output devices, network access devices, buses, etc.
The Processor 801 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete Gate or transistor logic device, discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 802 may be an internal storage element of the medical wastewater monitoring data quality control device, such as a hard disk or a memory of the medical wastewater monitoring data quality control device. The memory 802 may also be an external storage device of the medical wastewater monitoring data quality control device, such as a plug-in hard disk, a Smart Memory Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, which are provided on the medical wastewater monitoring data quality control device. Further, the memory 802 may also include both an internal storage unit and an external storage device of the medical wastewater monitoring data quality control apparatus. The memory 802 is used to store the computer program and other programs and data required by the medical wastewater monitoring data quality control device. The memory 802 may also be used to temporarily store data that has been output or is to be output.
Example 4:
the present embodiment provides a computer-readable storage medium, which stores a computer program that, when executed by a processor, implements the steps of the method of embodiment 1.
The computer-readable medium can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Further, the computer readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
The above embodiments are only for illustrating the technical concept and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention accordingly, and not to limit the protection scope of the present invention accordingly. All equivalent changes or modifications made in accordance with the spirit of the present disclosure are intended to be covered by the scope of the present disclosure.
Claims (10)
1. A medical wastewater monitoring data quality control method is characterized by comprising the following steps:
collecting monitoring data of the whole medical wastewater treatment process;
classifying and identifying the acquired monitoring data to obtain different data types;
and matching the corresponding abnormal detection algorithm and parameters according to the identified data type to obtain abnormal point data.
2. The medical wastewater monitoring data quality control method according to claim 1, further comprising:
and pushing the obtained abnormal point data to the target object.
3. The medical wastewater monitoring data quality control method according to claim 1, wherein the classifying and identifying the collected monitoring data to obtain different data types comprises:
the method comprises the steps that fast Fourier transform is applied to acquired monitoring data, different acquired data signals are converted from a time domain to a frequency domain form, whether the acquired data are periodic or not is confirmed according to the energy distribution condition of frequency in a spectrogram after data conversion, and the period of a data sequence is obtained according to the N/frequency of the acquired data, namely the acquired data are identified as periodic data;
and for non-periodic data, the unit root test in the time sequence data is passed, if the data sequence passes the test, the data sequence is identified as non-periodic stationary data, and if the data sequence does not pass the test, the data sequence is identified as non-periodic irregular fluctuating data.
4. The medical wastewater monitoring data quality control method according to claim 3, characterized in that for aperiodic stationarity data, an orphan forest algorithm is adopted to identify abnormal data in operation and maintenance:
when constructing the solitary forest subtree, firstly sampling a data set by a system sampling method to construct a sub-forest, and forming the sub-forest into a base forest abnormity detector; judging the abnormal condition of data entering a sliding window through a base forest abnormal detector; and judging whether the data volume to be detected and the abnormal rate of the sliding window data exceed a threshold value according to the history, and if so, determining the data are abnormal point data.
5. The medical wastewater monitoring data quality control method according to claim 3, wherein the operation and maintenance data is abnormally monitored by using a Prophet-AE-LSTM integrated algorithm for the periodic data:
predicting the selected historical dependence data by using a Prophet algorithm to obtain a predicted value;
solving a Prophet predicted value and an actual residual error so as to carry out white noise detection on a predicted residual error sequence;
if the residual sequence passes white noise inspection, indicating that related information in the sequence is extracted, comparing a Prophet predicted value with a model prediction confidence interval, and if the Prophet predicted value exceeds the prediction confidence interval, determining the predicted point as an abnormal data point;
if the residual sequence does not pass white noise test, indicating that the related information remained in the sequence is not extracted, extracting the predicted residual and carrying out next-stage training, namely predicting the residual by adopting a multi-feature superimposed AE-LSTM algorithm; splicing the learned characteristics in the AE algorithm encoder with other related collected characteristics, and then inputting the spliced characteristics into an LSTM model to perform sequence prediction of a residual error item; compressing lag term variables of an original sequence by adopting an AE algorithm, splicing the compressed characteristics with other collected related variables and prophet predicted values to form new input characteristics, and training residual terms still containing information to obtain predicted values of the residual terms;
and adding the Prophet predicted value and the residual predicted value obtained in the previous step to obtain a final predicted value, and if the predicted value exceeds an empirical threshold, obtaining abnormal point data.
6. The quality control method for medical wastewater monitoring data according to claim 3, wherein the EMD-AE-LSTM algorithm is used for abnormal prediction of the non-periodic irregular fluctuation data:
firstly, decomposing original sequence data by using EMD (empirical mode decomposition) to obtain a component IMF (intrinsic mode function) of each sequence;
respectively constructing EM-LSTM network prediction for each IMF component on the basis of EMD empirical mode decomposition, splicing the features learned by an AE algorithm encoder from sequence lag variables with other related collected features, and inputting the spliced features into an LSTM model for sequence prediction;
and further comparing residual errors obtained according to the prediction result and the actual value with a threshold value, and if the residual errors exceed the threshold value, obtaining abnormal point data.
7. The medical wastewater monitoring data quality control method according to claim 5, wherein the expression of the Prophet algorithm is as follows:
y(t)＝g(t)+s(t)+h(t)+εt
wherein g (t) represents a trend term in the time series, s (t) represents a period term in the time series, h (t) represents the potential influence brought by holidays, epsilontAnd representing a model error term reflecting an abnormal variation not represented in the model.
8. The medical wastewater monitoring data quality control system is characterized by comprising:
the data acquisition module is used for acquiring monitoring data of the whole medical wastewater treatment process;
the data classification module is used for classifying and identifying the acquired monitoring data to obtain different data types;
the data analysis module is used for matching a corresponding abnormal detection algorithm and parameters according to the identified data type to obtain abnormal point data;
and the data pushing module is used for pushing the obtained abnormal point data to the target object.
9. A medical wastewater monitoring data quality control device comprising a memory, a processor and a computer program stored in said memory and executable on said processor, wherein said processor when executing said computer program implements the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011203309.2A CN112288021B (en) | 2020-11-02 | 2020-11-02 | Medical wastewater monitoring data quality control method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011203309.2A CN112288021B (en) | 2020-11-02 | 2020-11-02 | Medical wastewater monitoring data quality control method, device and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112288021A true CN112288021A (en) | 2021-01-29 |
| CN112288021B CN112288021B (en) | 2022-04-29 |
Family
ID=74353906
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011203309.2A Active CN112288021B (en) | 2020-11-02 | 2020-11-02 | Medical wastewater monitoring data quality control method, device and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112288021B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113240162A (en) * | 2021-04-28 | 2021-08-10 | 南京天溯自动化控制系统有限公司 | Hospital energy consumption prediction method based on EEMD-Prophet algorithm |
| CN114722972A (en) * | 2022-06-01 | 2022-07-08 | 新华三人工智能科技有限公司 | Anomaly detection method and device |
| CN115495274A (en) * | 2022-11-15 | 2022-12-20 | 阿里云计算有限公司 | Exception handling method based on time sequence data, network equipment and readable storage medium |
| CN115662588A (en) * | 2022-10-24 | 2023-01-31 | 南京喜悦科技股份有限公司 | Regional medical disinfection supply-based full-flow information tracing system and method |
| CN116384158A (en) * | 2023-05-26 | 2023-07-04 | 广东合诚环境工程有限公司 | Sewage treatment equipment operation monitoring method and system based on big data |
| CN116881747A (en) * | 2023-09-06 | 2023-10-13 | 武汉华康世纪医疗股份有限公司 | Intelligent treatment method and system based on medical wastewater monitoring |
| CN117605539A (en) * | 2023-12-13 | 2024-02-27 | 济南福深兴安科技有限公司 | Intelligent early warning system and method for monitoring coal mine water damage |
| CN117689218A (en) * | 2024-02-04 | 2024-03-12 | 成都工喜科技有限公司 | Intelligent management and control system suitable for industrial enterprise production site environment risk |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107657288A (en) * | 2017-10-26 | 2018-02-02 | 国网冀北电力有限公司 | A kind of power scheduling flow data method for detecting abnormality based on isolated forest algorithm |
| CN110378392A (en) * | 2019-06-26 | 2019-10-25 | 华东师范大学 | A kind of indoor geriatric state's monitoring method based on LSTM-AE |
| CN110865929A (en) * | 2019-11-26 | 2020-03-06 | 携程旅游信息技术（上海）有限公司 | Abnormity detection early warning method and system |
| CN110888788A (en) * | 2019-10-16 | 2020-03-17 | 平安科技（深圳）有限公司 | Anomaly detection method and device, computer equipment and storage medium |
| CN111353482A (en) * | 2020-05-25 | 2020-06-30 | 天津开发区精诺瀚海数据科技有限公司 | LSTM-based fatigue factor recessive anomaly detection and fault diagnosis method |
| CN111352971A (en) * | 2020-02-28 | 2020-06-30 | 中国工商银行股份有限公司 | Bank system monitoring data anomaly detection method and system |
| US20200210393A1 (en) * | 2018-09-14 | 2020-07-02 | Verint Americas Inc. | Framework and method for the automated determination of classes and anomaly detection methods for time series |
| CN111639798A (en) * | 2020-05-26 | 2020-09-08 | 华青融天(北京)软件股份有限公司 | Intelligent prediction model selection method and device |
-
2020
- 2020-11-02 CN CN202011203309.2A patent/CN112288021B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107657288A (en) * | 2017-10-26 | 2018-02-02 | 国网冀北电力有限公司 | A kind of power scheduling flow data method for detecting abnormality based on isolated forest algorithm |
| US20200210393A1 (en) * | 2018-09-14 | 2020-07-02 | Verint Americas Inc. | Framework and method for the automated determination of classes and anomaly detection methods for time series |
| CN110378392A (en) * | 2019-06-26 | 2019-10-25 | 华东师范大学 | A kind of indoor geriatric state's monitoring method based on LSTM-AE |
| CN110888788A (en) * | 2019-10-16 | 2020-03-17 | 平安科技（深圳）有限公司 | Anomaly detection method and device, computer equipment and storage medium |
| CN110865929A (en) * | 2019-11-26 | 2020-03-06 | 携程旅游信息技术（上海）有限公司 | Abnormity detection early warning method and system |
| CN111352971A (en) * | 2020-02-28 | 2020-06-30 | 中国工商银行股份有限公司 | Bank system monitoring data anomaly detection method and system |
| CN111353482A (en) * | 2020-05-25 | 2020-06-30 | 天津开发区精诺瀚海数据科技有限公司 | LSTM-based fatigue factor recessive anomaly detection and fault diagnosis method |
| CN111639798A (en) * | 2020-05-26 | 2020-09-08 | 华青融天(北京)软件股份有限公司 | Intelligent prediction model selection method and device |
Non-Patent Citations (2)
| Title |
|---|
| 柳青秀等: "基于长短时记忆―自编码神经网络的风电机组性能评估及异常检测", 《计算机集成制造系统》 * |
| 赵英等: "基于LSTM-Prophet非线性组合的时间序列预测模型", 《计算机与现代化》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113240162A (en) * | 2021-04-28 | 2021-08-10 | 南京天溯自动化控制系统有限公司 | Hospital energy consumption prediction method based on EEMD-Prophet algorithm |
| CN113240162B (en) * | 2021-04-28 | 2022-03-01 | 南京天溯自动化控制系统有限公司 | Hospital energy consumption prediction method based on EEMD-Prophet algorithm |
| CN114722972A (en) * | 2022-06-01 | 2022-07-08 | 新华三人工智能科技有限公司 | Anomaly detection method and device |
| CN115662588A (en) * | 2022-10-24 | 2023-01-31 | 南京喜悦科技股份有限公司 | Regional medical disinfection supply-based full-flow information tracing system and method |
| CN115495274A (en) * | 2022-11-15 | 2022-12-20 | 阿里云计算有限公司 | Exception handling method based on time sequence data, network equipment and readable storage medium |
| CN115495274B (en) * | 2022-11-15 | 2023-03-07 | 阿里云计算有限公司 | Exception handling method based on time sequence data, network equipment and readable storage medium |
| CN116384158A (en) * | 2023-05-26 | 2023-07-04 | 广东合诚环境工程有限公司 | Sewage treatment equipment operation monitoring method and system based on big data |
| CN116384158B (en) * | 2023-05-26 | 2023-08-18 | 广东合诚环境工程有限公司 | Sewage treatment equipment operation monitoring method and system based on big data |
| CN116881747A (en) * | 2023-09-06 | 2023-10-13 | 武汉华康世纪医疗股份有限公司 | Intelligent treatment method and system based on medical wastewater monitoring |
| CN116881747B (en) * | 2023-09-06 | 2023-11-24 | 武汉华康世纪医疗股份有限公司 | Intelligent treatment method and system based on medical wastewater monitoring |
| CN117605539A (en) * | 2023-12-13 | 2024-02-27 | 济南福深兴安科技有限公司 | Intelligent early warning system and method for monitoring coal mine water damage |
| CN117689218A (en) * | 2024-02-04 | 2024-03-12 | 成都工喜科技有限公司 | Intelligent management and control system suitable for industrial enterprise production site environment risk |
| CN117689218B (en) * | 2024-02-04 | 2024-04-12 | 成都工喜科技有限公司 | Intelligent management and control system suitable for industrial enterprise production site environment risk |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112288021B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112288021B (en) | Medical wastewater monitoring data quality control method, device and system | |
| Russo et al. | Anomaly detection using deep autoencoders for in-situ wastewater systems monitoring data | |
| KR102149495B1 (en) | Optimization apparatus for training conditions of environmental prediction model and operating thereof | |
| CN112414694B (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN116451142A (en) | Water quality sensor fault detection method based on machine learning algorithm | |
| CN116416884B (en) | Testing device and testing method for display module | |
| CN113948159B (en) | Fault detection method, device and equipment for transformer | |
| CN115456107A (en) | Time series abnormity detection system and method | |
| CN117436569A (en) | Nuclear power equipment fault prediction and intelligent calibration method and system based on random forest | |
| KR20230021272A (en) | Bridge health monitoring system using artificial intelligence | |
| CN116597939A (en) | Medicine quality control management analysis system and method based on big data | |
| CN117060353A (en) | Fault diagnosis method and system for high-voltage direct-current transmission system based on feedforward neural network | |
| Wang et al. | Early warning of reciprocating compressor valve fault based on deep learning network and multi-source information fusion | |
| CN111125186A (en) | Data processing method and system based on questionnaire | |
| Dang et al. | seq2graph: Discovering dynamic non-linear dependencies from multivariate time series | |
| CN117933531A (en) | Distributed photovoltaic power generation power prediction system and method | |
| Bond et al. | A hybrid learning approach to prognostics and health management applied to military ground vehicles using time-series and maintenance event data | |
| CN115600695B (en) | Fault diagnosis method for metering equipment | |
| CN116631626A (en) | Patient clinical risk assessment method, device, equipment and medium | |
| Wang et al. | An Entropy-and Attention-Based Feature Extraction and Selection Network for Multi-Target Coupling Scenarios | |
| Cui et al. | Research on Fault Detection Technology of Air-to-Air Missile Based on Data Mining | |
| Colace et al. | Unsupervised Learning Techniques for Vibration-Based Structural Health Monitoring Systems Driven by Data: A General Overview | |
| Donets et al. | APPLICATION OF A DATA STRATIFICATION APPROACH IN COMPUTER MEDICAL MONITORING SYSTEMS. | |
| Zadkarami et al. | Big Data Analytics for Advanced Fault Detection in Wastewater Treatment Plants | |
| CN112187555B (en) | Real-time KPI data anomaly detection method and device based on machine learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |