CN115936060B - Substation capacitance temperature early warning method based on depth deterministic strategy gradient - Google Patents
Substation capacitance temperature early warning method based on depth deterministic strategy gradient Download PDFInfo
- Publication number
- CN115936060B CN115936060B CN202211699940.5A CN202211699940A CN115936060B CN 115936060 B CN115936060 B CN 115936060B CN 202211699940 A CN202211699940 A CN 202211699940A CN 115936060 B CN115936060 B CN 115936060B
- Authority
- CN
- China
- Prior art keywords
- substation
- early warning
- data
- capacitor
- temperature
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 239000003990 capacitor Substances 0.000 claims abstract description 75
- 230000009471 action Effects 0.000 claims abstract description 33
- 238000013528 artificial neural network Methods 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 44
- 239000010410 layer Substances 0.000 claims description 38
- 230000008569 process Effects 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 16
- 238000012544 monitoring process Methods 0.000 claims description 13
- 210000002569 neuron Anatomy 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 230000005855 radiation Effects 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 3
- 210000004205 output neuron Anatomy 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 239000002356 single layer Substances 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 abstract description 2
- 230000008901 benefit Effects 0.000 description 5
- 238000012806 monitoring device Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000002547 anomalous effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention relates to a substation capacitance temperature early warning method based on depth deterministic strategy gradient, which comprises the following steps: constructing a substation capacitance temperature early warning terminal comprising a prediction model based on a deep belief network and an early warning model based on a depth deterministic strategy gradient algorithm; establishing an index system affecting the temperature of the transformer substation capacitor, and predicting the operation temperature of the transformer substation capacitor through a deep belief network after finishing data preprocessing; and establishing a controller based on depth deterministic strategy gradient, taking the monitored index system affecting the capacitance temperature of the transformer substation as a state space set of the controller, taking the judgment action of the controller as an action space set, and judging the early warning level according to the action sequence number of the action space for early warning. According to the invention, the historical information data value is deeply dug by utilizing artificial intelligence, big data and other technologies, and the fault early warning of the transformer substation capacitance temperature is realized by applying the prediction capability of the neural network and the autonomous learning capability of the control model.
Description
Technical Field
The invention relates to the technical field of intelligent control of power grid power, in particular to a substation capacitance temperature early warning method based on depth deterministic strategy gradient.
Background
The transformer substation is used as a pivot point of a power grid system, plays an important role in power supply, and a capacitor bank in the transformer substation has a very important significance for normal operation of the transformer substation; when the capacitor is damaged, the reactive compensation and voltage regulation of the capacitor bank are difficult to normally realize, so that the power supply is affected to a certain extent.
At present, manual replacement is mostly adopted for the replacement of the transformer substation capacitor, but because the capacitor body is heavy and the damaged capacitor is possibly positioned at the top of the transformer substation angle steel frame, the risk of manual operation is high and the labor intensity is high; therefore, in order to avoid capacitor faults and frequent replacement, the early warning of the capacitor operation state of the transformer substation is enhanced, the temperature is gradually becoming a key operation index of the capacitor operation process of the transformer substation nowadays, when the electricity consumption peak arrives, the reactive compensation function of the capacitor is more obvious, the capacitor is easy to overheat due to the overheat, faults are caused, and explosion is caused, so that the problem that how to accurately early warn the temperature of the capacitor is the problem which must be considered at present.
It should be noted that the information disclosed in the above background section is only for enhancing understanding of the background of the present disclosure and thus may include information that does not constitute prior art known to those of ordinary skill in the art.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a substation capacitance temperature early warning method based on depth deterministic strategy gradient, and solves the problem that the prior art cannot accurately early warn the temperature of a capacitor.
The aim of the invention is achieved by the following technical scheme: a substation capacitance temperature early warning method based on depth deterministic strategy gradients, the temperature early warning method comprising:
s1, constructing a substation capacitance temperature early warning terminal comprising a prediction model based on a deep belief network and an early warning model based on a deep deterministic strategy gradient algorithm;
s2, establishing a meteorological condition set S of the transformer substation, a capacitor operation condition set A of the transformer substation and a capacitor real-time temperature value T of the capacitor 0 Predicted temperature value T of system capacitance 1 The method comprises the steps that (1) the operation temperature of a substation capacitor is predicted through a deep belief network after data preprocessing is completed;
s3, establishing a controller based on depth deterministic strategy gradient, and acquiring a meteorological condition set S, a substation capacitance operation condition set A, a transformer substation capacitance operation condition set,Capacitor real-time temperature value T 0 Predicted temperature value T of system capacitance 1 And the state space set is used as a state space set of the controller, the judging action of the controller is used as an action space set, and the early warning grade is judged according to the action sequence number of the action space to carry out early warning.
The data sources of the platform are a substation meteorological condition set S, a substation capacitor operation condition set A and a capacitor historical temperature value T which are detected by a substation monitoring device in a certain area -1 And a real-time temperature value T 0 Predicted temperature value T of system capacitance 1 . Firstly, in order to obtain real-time substation capacitance operation state information and environment information, a centralized mode is applied to design the architecture of the information acquisition terminal. Secondly, the network structure of the early warning control system comprises a monitoring layer, a data layer and a decision layer; the layers include state information and action instructions. And further, a storage method of system data and a model is determined, unified early warning standards of all capacitance systems in the transformer substation are determined, and online control of the digital platform on the transformer substation capacitance is realized.
The data preprocessing comprises the following steps: identifying and processing abnormal data in the initial data through longitudinal comparison processing and transverse comparison processing of the data; the substation meteorological condition set comprises real-time temperature, relative humidity, real-time air pressure, real-time radiation and real-time wind speed.
The longitudinal contrast treatment comprises:
according to the principle that environmental data of a transformer substation periodically changes and continuous data sets are similar, the method comprises the following steps ofComparing whether the variation amplitude of the data at the same time in continuous days is kept in a certain range, and if the variation amplitude is not in the range, considering the data to be abnormal, wherein L (t) is the average data value of the last days t; l (i, t) is the data value of the i-th class time t; delta (t) is an artificially specified anomaly threshold.
The transverse contrast process comprises:
setting the maximum range of data change by taking the data of adjacent moments as a reference through the formulaAnd judging whether the absolute value of the difference value before and after the data value exceeds a threshold value, and if so, considering the data value as abnormal data, wherein the absolute value of the difference value before and after the data value exceeds the threshold value. α (t) and β (t) are artificially specified anomaly thresholds.
The deep belief network is composed of multiple layers of nonlinear operation units, wherein the input of high-level features or factors is obtained from the output of low-level. It mainly includes deep belief networks (Deep Belief Network, DBN), convolutional neural networks, recurrent neural networks, etc. The training process of the DBN comprises pre-training and reverse fine tuning. Pretraining is essentially an unsupervised greedy layer-by-layer training: first training the lowest limited boltzmann machine (RBM); secondly, taking the next RBM layer as input data of the previous RBM layer, and training layer by layer from bottom to top to obtain an initial weight of the deep belief network; and finally, after the pre-training is finished, reversely fine-tuning the parameters of the network by utilizing the data with the labels, so as to complete the whole training process.
The prediction model comprises an h-layer neural network and u neurons in each layer, when the number u of the neurons in the hidden layers reaches a certain number, the single-layer hidden layer structure with h=1 is more convenient than the addition of more hidden layers to improve the training precision, so that the number u of the neurons in the hidden layers is subjected to trial-and-error through a trial-and-error method until the minimum number of the neurons meeting the error requirement is found; after the monitoring information is obtained, the capacitor operation temperature T of the future transformer substation is obtained quickly and accurately 1 。
The temperature predicted value of a certain time period (with a settable length) can be obtained by establishing a network containing n output neurons after the prediction process is completed. The invention selects a single output structure, namely, each time point (predicted point) corresponds to one neural network, and has the advantages of small network structure, high calculation speed, high training accuracy and the like. Before the deep belief network is put into use, a training process is needed to be carried out, and a historical data set (a meteorological condition set S, a substation capacitor operation condition set A and a capacitor historical temperature value T which are obtained through monitoring) -1 And real-time temperature value T 0 ) The training set is input into a deep belief network, and the deep belief network is fitted with a meteorological condition set S, a substation capacitor operation condition set A and a capacitor historical temperature value T -1 And real-time temperature value T 0 Data association between them. Based on the result, the test set data is input into the trained deep belief network, and the predicted temperature value T can be output 1 If the error meets the requirement with the actual value, the method can be put into use.
After training of the deep belief network is completed, a meteorological condition set S, a substation capacitor operation condition set A, a capacitor history and a real-time temperature value T are obtained through monitoring -1 And T is 0 As the input of the prediction, the prediction result T can be obtained 1 Thereby yielding a complete state set.
The depth deterministic strategy gradient meets the Markov decision process, obeys Markov properties, is obtained by fusing an Actor-Critic framework and a depth Q learning algorithm, inherits an independent target network in the depth Q learning algorithm, has experience playback capability and is obtained by a loss function formulaAnd parameter update formulaTo break the relevance between data and reduce the training difficulty of the model, wherein r is as follows i (s, a) represents the prize value r in state s and action a; />As a concentrated action value function, xi i Is a corresponding parameter; mu (mu) i (θ i ) Is a parameter of theta i Expression of N strategies, abbreviated μ i 。
The state space set of the controller is a transformer substation meteorological condition set S, a transformer substation capacitor operation condition set A and a capacitor historical temperature value T which are obtained through monitoring -1 And a real-time temperature value T 0 Predicted temperature value T of system capacitance 1 Expressed as state= { T h p r v I T -1 T 0 T 1 T is the ambient temperature, h is the ambient relative humidity, p is the substation air pressure, r is the substation relative radiation, v is the wind speed at the substation capacitor device, and I is the substation capacitor operation current;
the action sequence numbers in the action space set are respectively corresponding to 0-9, namely, are expressed as A= {0,1,2,3,4,5,6,7,8,9}, the larger the sequence number value is, the more urgent the situation is, 0 represents that the current temperature situation is good, and 9 represents that the operation needs to be stopped immediately and the capacitor needs to be replaced.
The capacitor temperature early warning method further comprises the following steps: the random trial-error learning process is a period of random trial-error learning process which is required to be accepted by each intelligent agent before use, and is called a pre-learning stage. In the early stage of pre-learning, the early warning terminal does not accumulate any experience, does not have intelligent control capability, and can obtain an optimal value function Q network only after receiving actions in various states. Therefore, before the system is put into use, functions with different amplitudes and different types are required to be overlapped to form a random data set, the random data set is used as a training data set to be input into the early warning terminal, a large number of trial and error learning training is carried out until the early warning terminal reaches convergence, and the training is finished.
The invention has the following advantages: a transformer substation capacitance temperature early warning method based on depth deterministic strategy gradient utilizes artificial intelligence, big data and other technologies to deeply dig historical information data value, and utilizes the prediction capability of a neural network and the autonomous learning capability of a control model to realize fault early warning of transformer substation capacitance temperature.
Drawings
FIG. 1 is a schematic flow chart of the present invention;
FIG. 2 is a schematic diagram of a network frame structure of the early warning terminal of the present invention;
FIG. 3 is a diagram of a deep belief network architecture of the present invention;
FIG. 4 is a trend graph of a pre-training device reward function of the present invention.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. Accordingly, the following detailed description of the embodiments of the present application, provided in connection with the accompanying drawings, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to be within the scope of the present application. The invention is further described below with reference to the accompanying drawings.
As shown in fig. 1, the invention specifically relates to a substation capacitance temperature early warning method based on depth deterministic strategy gradient, which comprises the following steps:
step 1, firstly, designing a substation capacitor temperature early warning terminal structure based on depth certainty strategy gradient aiming at the capacitor fault problem caused by overhigh temperature in the substation capacitor operation process, wherein the terminal structure mainly comprises a prediction model and an early warning model.
As shown in fig. 2, W1 is a decision layer, W2 is a data layer, W3 is a substation layer, and the substation capacitor temperature early-warning terminal based on the depth certainty strategy gradient is composed of a prediction model and an early-warning model respectively, wherein the prediction model is based on a depth belief network, and the early-warning model is based on a depth certainty strategy gradient algorithm. The data sources of the platform are a substation meteorological condition set S, a substation capacitor operation condition set A and a capacitor real-time temperature value T which are detected by a substation monitoring device in a certain area 0 Predicted temperature value T of system capacitance 1 . Firstly, in order to obtain real-time substation capacitance operation state information and environment information, a centralized mode is applied to design the architecture of the information acquisition terminal. Secondly, the network structure of the early warning control system comprises a monitoring layer, a data layer and a decision layer; the layers include state information and action instructions. Further, storage method for determining system data and modelAnd determining unified early warning standards of all capacitance systems in the transformer substation, and realizing online control of the digital platform on the transformer substation capacitance.
Step 2, in order to accurately predict the real-time temperature of the transformer substation, an index system which comprehensively influences the capacitance temperature of the transformer substation needs to be established, and the method comprises the following steps: meteorological condition set S of transformer substation, capacitor operation condition set A of transformer substation and capacitor real-time temperature value T 0 Predicted temperature value T of system capacitance 1 Therefore, through an index system, after data preprocessing is completed, the operation temperature of the substation capacitor can be predicted by using a deep belief network in deep learning.
The data preprocessing comprises longitudinal comparison processing and transverse comparison processing.
1) And (3) longitudinal comparison treatment: load longitudinal comparison refers to identifying abnormal data by comparing data at the same time for consecutive days. This is because the substation environment data changes periodically, and the continuous data sets are similar, i.e. the change amplitude should be kept within certain limits. If this range is exceeded, the data may be considered anomalous. The equation is as follows:
wherein L (t) is the average data value of the last few days t; l (i, t) is the data value of the i-th class time t; delta (t) is an artificially specified anomaly threshold.
2) Horizontal comparison treatment: the maximum range of data change is set with the data of adjacent time points as references. Then the data value may be considered erroneous data when the absolute value of the difference between the data value and the time instant exceeds a threshold value. The equation is as follows:
where α (t) and β (t) are artificially specified anomaly thresholds.
The deep neural network refers to a deep belief network model, which is composed of multiple layers of nonlinear operation units, wherein the input of high-level features or factors is obtained from the output of low-level. It mainly includes deep belief networks (Deep Belief Network, DBN), convolutional neural networks, recurrent neural networks, etc. The training process of the DBN comprises pre-training and reverse fine tuning. Pretraining is essentially an unsupervised greedy layer-by-layer training: first training the lowest limited boltzmann machine (RBM); secondly, taking the next RBM layer as input data of the previous RBM layer, and training layer by layer from bottom to top to obtain an initial weight of the deep belief network; and finally, after the pre-training is finished, reversely fine-tuning the parameters of the network by utilizing the data with the labels, so as to complete the whole training process.
As shown in fig. 3, the adjustment of the neural network structure can be discussed in terms of both network type and network depth. The depth of the network determines the generalization capability of the neural network, which includes the number of layers h of the neural network and the number of neurons per layer u. Therefore, when the number u of hidden layer neurons is sufficiently large, a single hidden layer structure of h=1 is much more convenient than increasing training accuracy by adding more hidden layers. Thus, the minimum number of neurons that meet the error requirement can be found by trial and error. Based on the method, a large amount of historical data is input to the input end of the prediction model, a corresponding prediction network can be obtained, and after the prediction model is put into use, the monitoring information can be acquired, and then the future substation capacitance operation temperature T can be rapidly and accurately acquired 1 。
The temperature predicted value of a certain time period (with a settable length) can be obtained by establishing a network containing n output neurons after the prediction process is completed. The invention selects a single output structure, namely, each time point (predicted point) corresponds to one neural network, and has the advantages of small network structure, high calculation speed, high training accuracy and the like. Before the deep belief network is put into use, a training process is needed to be carried out, and a historical data set (a meteorological condition set S, a substation capacitor operation condition set A and a capacitor historical temperature value T which are obtained through monitoring) 1 And real-time temperature value T 0 ) The training set is input into a deep belief network, and the deep belief network is fitted with a meteorological condition set S and a transformer substationCapacitor operation condition set A and capacitor historical temperature value T- 1 And real-time temperature value T 0 Data association between them. Based on the result, the test set data is input into the trained deep belief network, and the predicted temperature value T can be output 1 If the error meets the requirement with the actual value, the method can be put into use.
After the training of the deep belief network is completed, a meteorological condition set S, a substation capacitor operation condition set A, a capacitor history and a real-time temperature value T which are obtained by monitoring 1 And T is 0 As the input of the prediction, the prediction result T can be obtained 1 Thereby yielding a complete state set.
Step 3, further, in order to accurately and rapidly identify and early warn the real-time temperature, a controller based on depth deterministic strategy gradient is designed based on the operation scene of the transformer substation, and the meteorological condition set S of the transformer substation, the capacitor operation condition set A of the transformer substation and the capacitor real-time temperature value T of the capacitor detected by the monitoring device are sequentially carried out 0 Predicted temperature value T of system capacitance 1 The state space is defined as an operation space for determining the operation of the controller.
The depth deterministic strategy gradient meets the Markov decision process and obeys Markov properties, the depth deterministic strategy gradient algorithm is the fusion of an Actor-Critic framework and a depth Q learning algorithm, and inherits an independent target network in the depth Q learning algorithm, has experience playback capability, can break the relevance between data and reduces the training difficulty of a model. From this, the corresponding loss function formula and parameter update formula can be obtained as follows:
wherein r is i (s, a) represents the prize value r in state s and action a;as a concentrated action value function, xi is a corresponding parameter; mu (mu) i (θ i ) Is a parameter of theta i Expression of the N strategies at that time may be abbreviated as μ i 。
The controller state space set is a substation meteorological condition set S, a substation capacitor operation condition set A and a capacitor real-time temperature value T which are detected by the monitoring device 0 Predicted temperature value T of system capacitance 1 The method comprises the steps of carrying out a first treatment on the surface of the Defining a state space set as:
State＝{t h p r v I T 0 T 1 }
wherein t is the ambient temperature, h is the ambient relative humidity, p is the substation air pressure, r is the substation relative radiation, v is the wind speed near the substation capacitor device, and I is the substation capacitor operation current.
The combined Action space set Action of the substation capacitance temperature early warning terminal based on depth deterministic strategy gradient, namely the Action strategy adopted by the agent after decision should be the Action sequence numbers (early warning grades) of the early warning system in the problems described herein, respectively corresponding to sequence numbers 0-9, the larger the number is, the more urgent the situation is, 0 represents the good current temperature situation, and 9 represents the need of stopping operation and replacing the capacitor immediately; the set of action spaces may be expressed as:
A＝{0,1,2,3,4,5,6,7,8,9}
and 4, finally, setting a random data set formed by overlapping functions with different amplitudes and different types, inputting the random functions as a training data set into an intelligent body of the early warning terminal, and converging the intelligent body of the intelligent early warning terminal after a large number of trial and error training, so that the intelligent early warning terminal can be put into application.
As shown in FIG. 4, the pre-training phase is required before the intelligent controller is used, and the random trial-and-error learning process is a learning process of a section of random trial-and-error which each intelligent agent needs to accept before using. In the early stage of pre-learning, the controller has not accumulated any experience, does not have intelligent control capability, and can obtain an optimal value function Q network only after accepting actions in various states. Therefore, before the invention is put into use, a random data set formed by overlapping functions with different magnitudes and different types is required to be set, the random functions are used as training data sets to be input into the intelligent agent of the early warning terminal, and after a large number of trial and error training, the intelligent agent of the intelligent early warning terminal can be put into use after convergence. Therefore, compared with the traditional control means, the method can train by using the randomly generated data set, and the difficulty of acquiring the data training set is greatly reduced. It can also be seen that the agent substantially converged after about 600 rounds of iterations were completed, and completed and automatically stopped training at 739 rounds, with higher online learning capabilities. Therefore, the intelligent agent system has good convergence characteristic, and can smoothly perform rapid and accurate early warning tasks in actual application scenes.
In the regulation and control process, the information acquisition terminals between the substation capacitor and the early warning platform can communicate in a wired transmission or wireless communication 5G mode, so that the early warning platform can acquire state information from a substation system, and the action of maximizing the system benefit is judged by predicting the real-time running state, further, an early warning result is obtained, and the early warning result is sent to the substation control online platform for workers to quickly acquire corresponding information.
The invention sequentially completes state prediction, environmental state space acquisition and action space output by utilizing the depth deterministic strategy gradient formed by combining reinforcement learning and deep learning, thereby realizing rapid and accurate temperature early warning. The terminal device has online learning and experience playback capability, has good convergence characteristic and model adaptability, and can well cope with intelligent early warning scenes of abnormal temperature of the capacitance of the transformer substation.
The foregoing is merely a preferred embodiment of the invention, and it is to be understood that the invention is not limited to the form disclosed herein but is not to be construed as excluding other embodiments, but is capable of numerous other combinations, modifications and environments and is capable of modifications within the scope of the inventive concept, either as taught or as a matter of routine skill or knowledge in the relevant art. And that modifications and variations which do not depart from the spirit and scope of the invention are intended to be within the scope of the appended claims.
Claims (7)
1. A substation capacitance temperature early warning method based on depth deterministic strategy gradient is characterized in that: the temperature early warning method comprises the following steps:
s1, constructing a substation capacitance temperature early warning terminal comprising a prediction model based on a deep belief network and an early warning model based on a deep deterministic strategy gradient algorithm;
s2, establishing a meteorological condition set S of the transformer substation, a capacitor operation condition set A of the transformer substation and a capacitor real-time temperature value T of the capacitor 0 Predicted temperature value T of system capacitance 1 The method comprises the steps that (1) the operation temperature of a substation capacitor is predicted through a deep belief network after data preprocessing is completed;
s3, establishing a controller based on depth deterministic strategy gradient, and monitoring a meteorological condition set S, a substation capacitor operation condition set A and a capacitor historical temperature value T which are obtained -1 And a real-time temperature value T 0 Predicted temperature value T of system capacitance 1 As a state space set of the controller, taking the judging action of the controller as an action space set, and judging the early warning level according to the action sequence number of the action space for early warning;
the data preprocessing comprises the following steps: identifying and processing abnormal data in the initial data through longitudinal comparison processing and transverse comparison processing of the data; the substation meteorological condition set comprises real-time temperature, relative humidity, real-time air pressure, real-time radiation and real-time wind speed;
the longitudinal contrast treatment comprises:
according to the principle that environmental data of a transformer substation periodically changes and continuous data sets are similar, the method comprises the following steps ofComparing whether the variation amplitude of the data at the same time in consecutive days is kept within a certain range, if not, the data is considered abnormal, wherein L (t) is the average data of the last days tA value; l (i, t) is the data value of the i-th class time t; delta (t) is an artificially specified anomaly threshold;
the transverse contrast process comprises:
setting the maximum range of data change by taking the data of adjacent moments as a reference through the formulaWhether the absolute value of the difference between the front and rear moments of the data value exceeds a threshold value is judged, and if so, the data value is considered to be abnormal data, wherein alpha (t) and beta (t) are artificially designated abnormal threshold values.
2. The substation capacitance temperature early warning method based on depth deterministic strategy gradient according to claim 1, wherein the method comprises the following steps: the prediction model comprises an h-layer neural network and u neurons in each layer, when the number u of the neurons in the hidden layers reaches a certain number, the single-layer hidden layer structure with h=1 is more convenient than the addition of more hidden layers to improve the training precision, so that the number u of the neurons in the hidden layers is subjected to trial-and-error through a trial-and-error method until the minimum number of the neurons meeting the error requirement is found; after the monitoring information is obtained, the capacitor operation temperature T of the future transformer substation is obtained quickly and accurately 1 。
3. The substation capacitance temperature early warning method based on depth deterministic strategy gradient according to claim 1, wherein the method comprises the following steps: the temperature prediction value of a certain time period can be obtained by establishing a network containing n output neurons in the prediction process, a single output structure is selected, namely, each time point corresponds to one neural network, the deep belief network needs to be trained before being put into use, a historical data set is input into the deep belief network as a training set, and the deep belief network fits a meteorological condition set S, a substation capacitor operation condition set A and a capacitor historical temperature value T -1 And real-time temperature value T 0 Data relativity between the two, inputting the test set data into the trained deep belief network, and outputting the predicted temperature value T 1 。
4. The substation capacitance temperature early warning method based on depth deterministic strategy gradient according to claim 1, wherein the method comprises the following steps: after training of the deep belief network is completed, a meteorological condition set S, a substation capacitor operation condition set A, a capacitor history and a real-time temperature value T are obtained through monitoring -1 And T is 0 As the input of the prediction, the prediction result T can be obtained 1 Thereby yielding a complete state set.
5. The substation capacitance temperature early warning method based on depth deterministic strategy gradient according to claim 1, wherein the method comprises the following steps: the depth deterministic strategy gradient meets the Markov decision process, obeys Markov properties, is obtained by fusing an Actor-Critic framework and a depth Q learning algorithm, and is obtained by a loss function formulaAnd parameter update formulaTo break the relevance between data and reduce the training difficulty of the model, wherein r is as follows i (s, a) represents the prize value r in state s and action a; />As a concentrated action value function, xi i Is a corresponding parameter; mu (mu) i (θ i ) Is a parameter of theta i Expression of N strategies, abbreviated μ i 。
6. The substation capacitance temperature early warning method based on depth deterministic strategy gradient according to claim 1, wherein the method comprises the following steps: the state space set of the controller is a transformer substation meteorological condition set S, a transformer substation capacitor operation condition set A and a capacitor historical temperature value T which are obtained through monitoring -1 And a real-time temperature value T 0 Predicted temperature value T of system capacitance 1 Expressed as state= { T h p r v I T -1 T 0 T 1 T is the ambient temperature, h is the ambient relative humidity, p is the substation air pressure, r is the substation relative radiation, v is the wind speed at the substation capacitor device, and I is the substation capacitor operation current;
the action sequence numbers in the action space set are respectively corresponding to 0-9, namely, are expressed as A= {0,1,2,3,4,5,6,7,8,9}, the larger the sequence number value is, the more urgent the situation is, 0 represents that the current temperature situation is good, and 9 represents that the operation needs to be stopped immediately and the capacitor needs to be replaced.
7. The substation capacitance temperature early warning method based on depth deterministic strategy gradient according to any one of claims 1-6, wherein the method is characterized in that: the capacitor temperature early warning method further comprises the following steps: and setting functions with different amplitudes and different types to be overlapped to form a random data set, inputting the random data set as a training data set into the early warning terminal, performing a large number of trial and error learning training until the early warning terminal reaches convergence, and ending the training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211699940.5A CN115936060B (en) | 2022-12-28 | 2022-12-28 | Substation capacitance temperature early warning method based on depth deterministic strategy gradient |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211699940.5A CN115936060B (en) | 2022-12-28 | 2022-12-28 | Substation capacitance temperature early warning method based on depth deterministic strategy gradient |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115936060A CN115936060A (en) | 2023-04-07 |
| CN115936060B true CN115936060B (en) | 2024-03-26 |
Family
ID=86655749
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211699940.5A Active CN115936060B (en) | 2022-12-28 | 2022-12-28 | Substation capacitance temperature early warning method based on depth deterministic strategy gradient |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115936060B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116661317B (en) * | 2023-06-01 | 2024-04-12 | 六和电子（江西）有限公司 | Stepped temperature, time and pressure hot-pressing method for thin film capacitor |
| CN116470190B (en) * | 2023-06-19 | 2023-09-22 | 江西五十铃汽车有限公司 | Power battery pack heating method, system, computer and readable storage medium |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107907235A (en) * | 2017-07-14 | 2018-04-13 | 国网浙江省电力公司湖州供电公司 | A kind of substation capacitors automatic temperature measurement prior-warning device and method |
| CN109270442A (en) * | 2018-08-21 | 2019-01-25 | 西安工程大学 | High-voltage circuitbreaker fault detection method based on DBN-GA neural network |
| CN109284863A (en) * | 2018-09-04 | 2019-01-29 | 南京理工大学 | A kind of power equipment temperature predicting method based on deep neural network |
| WO2019141040A1 (en) * | 2018-01-22 | 2019-07-25 | 佛山科学技术学院 | Short term electrical load predication method |
| CN110175386A (en) * | 2019-05-21 | 2019-08-27 | 陕西科技大学 | Substation Electric Equipment temperature predicting method |
| CN110580543A (en) * | 2019-08-06 | 2019-12-17 | 天津大学 | Power load prediction method and system based on deep belief network |
| CN110763929A (en) * | 2019-08-08 | 2020-02-07 | 浙江大学 | Intelligent monitoring and early warning system and method for convertor station equipment |
| CN111028486A (en) * | 2019-12-26 | 2020-04-17 | 杭州电力设备制造有限公司 | High-temperature early warning method, device, equipment and storage medium for electrical equipment |
| CN111091143A (en) * | 2019-11-22 | 2020-05-01 | 国网新疆电力有限公司电力科学研究院 | Distribution transformer weight overload early warning method based on deep belief network and K-means clustering |
| CN112910288A (en) * | 2020-12-08 | 2021-06-04 | 上海交通大学 | Over-temperature early warning method based on inverter radiator temperature prediction |
| WO2021257128A2 (en) * | 2020-02-14 | 2021-12-23 | Cornell University | Quantum computing based deep learning for detection, diagnosis and other applications |
| CN113935463A (en) * | 2021-09-30 | 2022-01-14 | 南方电网数字电网研究院有限公司 | Microgrid controller based on artificial intelligence control method |
| WO2022062286A1 (en) * | 2020-09-24 | 2022-03-31 | 广东电网有限责任公司河源供电局 | Method and apparatus for controlling transformer, device, and storage medium |
| CN115130369A (en) * | 2022-05-09 | 2022-09-30 | 国网上海市电力公司 | Transformer substation equipment comprehensive state evaluation and active early warning method based on digital twinning |
-
2022
- 2022-12-28 CN CN202211699940.5A patent/CN115936060B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107907235A (en) * | 2017-07-14 | 2018-04-13 | 国网浙江省电力公司湖州供电公司 | A kind of substation capacitors automatic temperature measurement prior-warning device and method |
| WO2019141040A1 (en) * | 2018-01-22 | 2019-07-25 | 佛山科学技术学院 | Short term electrical load predication method |
| CN109270442A (en) * | 2018-08-21 | 2019-01-25 | 西安工程大学 | High-voltage circuitbreaker fault detection method based on DBN-GA neural network |
| CN109284863A (en) * | 2018-09-04 | 2019-01-29 | 南京理工大学 | A kind of power equipment temperature predicting method based on deep neural network |
| CN110175386A (en) * | 2019-05-21 | 2019-08-27 | 陕西科技大学 | Substation Electric Equipment temperature predicting method |
| CN110580543A (en) * | 2019-08-06 | 2019-12-17 | 天津大学 | Power load prediction method and system based on deep belief network |
| CN110763929A (en) * | 2019-08-08 | 2020-02-07 | 浙江大学 | Intelligent monitoring and early warning system and method for convertor station equipment |
| CN111091143A (en) * | 2019-11-22 | 2020-05-01 | 国网新疆电力有限公司电力科学研究院 | Distribution transformer weight overload early warning method based on deep belief network and K-means clustering |
| CN111028486A (en) * | 2019-12-26 | 2020-04-17 | 杭州电力设备制造有限公司 | High-temperature early warning method, device, equipment and storage medium for electrical equipment |
| WO2021257128A2 (en) * | 2020-02-14 | 2021-12-23 | Cornell University | Quantum computing based deep learning for detection, diagnosis and other applications |
| WO2022062286A1 (en) * | 2020-09-24 | 2022-03-31 | 广东电网有限责任公司河源供电局 | Method and apparatus for controlling transformer, device, and storage medium |
| CN112910288A (en) * | 2020-12-08 | 2021-06-04 | 上海交通大学 | Over-temperature early warning method based on inverter radiator temperature prediction |
| CN113935463A (en) * | 2021-09-30 | 2022-01-14 | 南方电网数字电网研究院有限公司 | Microgrid controller based on artificial intelligence control method |
| CN115130369A (en) * | 2022-05-09 | 2022-09-30 | 国网上海市电力公司 | Transformer substation equipment comprehensive state evaluation and active early warning method based on digital twinning |
Non-Patent Citations (2)
| Title |
|---|
| 变电站设备温度预警系统研究与实现;吴波 等;《建筑电气》;第31卷(第11期);47-54 * |
| 基于LSTM算法的变电设备发热预测技术研究;苏宁 等;《山东电力技术》;第48卷(第06期);10-14 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115936060A (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115936060B (en) | Substation capacitance temperature early warning method based on depth deterministic strategy gradient | |
| US11409347B2 (en) | Method, system and storage medium for predicting power load probability density based on deep learning | |
| CN110929918B (en) | 10kV feeder fault prediction method based on CNN and LightGBM | |
| CN111079989B (en) | DWT-PCA-LSTM-based water supply amount prediction device for water supply company | |
| CN106600136B (en) | Power section out-of-limit control efficiency evaluation method | |
| CN112488452B (en) | Energy system management multi-time scale optimal decision method based on deep reinforcement learning | |
| CN112798961B (en) | Method for predicting remaining service life of power battery of electric automobile | |
| CN112651519A (en) | Secondary equipment fault positioning method and system based on deep learning theory | |
| CN113947186A (en) | Heat supply energy consumption circulation prediction method based on generation of countermeasure network | |
| CN113313306A (en) | Elastic neural network load prediction method based on improved wolf optimization algorithm | |
| CN107273971B (en) | Feed-forward neural network structure self-organization method based on neuron significance | |
| CN111651878B (en) | Large power grid static voltage stability optimization decision method and system considering situation assessment | |
| CN116880191A (en) | Intelligent control method of process industrial production system based on time sequence prediction | |
| CN114204546B (en) | Unit combination optimization method considering new energy consumption | |
| CN111414927A (en) | Method for evaluating seawater quality | |
| CN114722693A (en) | Optimization method of two-type fuzzy control parameter of water turbine regulating system | |
| CN114499409A (en) | Photovoltaic module hot spot fault diagnosis method | |
| CN112615364A (en) | Novel wide-area intelligent cooperative control method for power grid stability control device | |
| CN113159395A (en) | Deep learning-based sewage treatment plant water inflow prediction method and system | |
| CN111859780A (en) | Micro-grid operation optimization method and system | |
| CN115434878B (en) | Wind generating set temperature cluster control method, device, equipment and medium | |
| Wang et al. | Dynamic neural network based genetic algorithm optimizing for short term load forecasting | |
| CN113537627B (en) | Marine wind turbine generator set fault interval time prediction method oriented to operation and maintenance | |
| CN107423847A (en) | A kind of Methods of electric load forecasting of distributed intelligence power distribution network | |
| Zhou et al. | A Medium and Long Term Load Forecasting Method Based on BP Neural Network and S-Shaped Curve Fitting |
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 |