CN113206507A - Three-phase load unbalance edge side treatment method and system - Google Patents
Three-phase load unbalance edge side treatment method and system Download PDFInfo
- Publication number
- CN113206507A CN113206507A CN202110522264.3A CN202110522264A CN113206507A CN 113206507 A CN113206507 A CN 113206507A CN 202110522264 A CN202110522264 A CN 202110522264A CN 113206507 A CN113206507 A CN 113206507A
- Authority
- CN
- China
- Prior art keywords
- intelligent
- reversing switch
- phase
- phase load
- intelligent reversing
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/26—Arrangements for eliminating or reducing asymmetry in polyphase networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/50—Arrangements for eliminating or reducing asymmetry in polyphase networks
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a three-phase load unbalance edge side treatment method and a three-phase load unbalance edge side treatment system, which comprise the following steps: acquiring three-phase current data output by a transformer in a low-voltage distribution station area; when the three-phase load unbalance calculated according to the three-phase current data does not meet a preset standard, calculating an intelligent reversing switch adjusting strategy according to the three-phase current data by adopting a reinforcement learning model constructed based on reinforcement learning, and sending the intelligent reversing switch adjusting strategy to the intelligent reversing switch; and the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjusting strategy, so that the real-time treatment of the unbalanced three-phase load is realized. The well-trained reinforcement learning model is deployed on the edge side of the low-voltage power distribution area, the high complexity and time calculation cost of an iterative algorithm are avoided, meanwhile, the actual environment is considered, low-voltage three-phase load unbalance management is achieved in the low-voltage power distribution area through edge calculation equipment, internet of things communication and an intelligent reversing switch, the real-time response speed is high, the safety is high, and the pressure of a cloud network is reduced.
Description
Technical Field
The invention relates to the field of low-voltage power distribution, in particular to a three-phase load unbalance edge side treatment method and a three-phase load unbalance edge side treatment system.
Background
The low-voltage distribution area is used as a terminal part of an electric power system and directly faces to power consumers, most of the household users are single-phase power, and the phenomenon of three-phase load imbalance in the distribution area is serious due to unbalanced load and different power utilization moments. The three-phase load unbalance of the distribution area can generate adverse effects on the power supply safety, the power supply quality and the economic operation of the distribution network, and is one of the main embodiments of the weak operation links of the distribution network.
At present, two different solutions are mainly provided for a three-phase load unbalance treatment method in a low-voltage distribution area. Firstly, adopt enhancement mode equipment, rely on traditional electric power electronic type SVG or SVG reactive compensation technique, combine ripe control algorithm, make it possess reactive compensation simultaneously and restrain the unbalanced function of three-phase. However, such devices are usually installed on the low voltage side of the distribution transformer for the purpose of centralized compensation. Although the quality of the electric energy at the outlet of the transformer can be effectively adjusted, the operation condition of the transformer is improved, the phenomenon of three-phase imbalance of uncertain distribution along the line is still not effectively adjusted, and the low-voltage condition still exists. And secondly, an intelligent reversing switch is adopted. The intelligent reversing switch is a switch device capable of automatically adjusting the phase of a power supply, the source end of the intelligent reversing switch is three-phase input, the outlet end of the intelligent reversing switch is single-phase, the operation condition is that only one phase working phase exists whenever, the other two phases are in a breaking state, and the potential switching can be realized according to a control instruction. The scheme fundamentally solves the problem of unbalanced three-phase load, and compared with the scheme of SVG reactive compensation, the scheme can effectively and quickly adjust the three-phase load state in a low-voltage distribution area.
Although the intelligent commutation switch technology realizes the online switching of the power utilization phase, in the practical application process, it is still difficult to coordinate and adjust the actions of a large number of intelligent commutation switches in the power distribution network. On one hand, the adjustment time is not suitable to be too long, so that the influence of a three-phase load unbalance state on the safety of the power distribution network is prevented, and the complexity of a coordination algorithm cannot be too high; on the other hand, the number of the adjusting reversing switches is reduced as much as possible, and the three-phase load balance state is achieved as soon as possible by the minimum adjusting number, so that the adjusting reliability is guaranteed, and the service life of the reversing switches is prolonged.
A large number of transformer areas exist in the actual operation of a power grid, if the operation state of each transformer area is transmitted to the cloud end in a centralized mode and is processed by the cloud end in sequence, the response time of three-phase unbalance management of the transformer areas is influenced, and hidden dangers also exist in the safety. The existing coordination adjustment method for the reversing switch mostly adopts an iterative algorithm for adjustment, the complexity and the calculation time of the algorithm are high, and the adjustment of the reversing switch with large scale in an actual low-voltage power distribution station area is difficult to meet.
Therefore, how to reduce the complexity of the three-phase load imbalance algorithm based on the intelligent reversing switch, the realization of short adjustment time and small adjustment quantity, and the main research problem of rapidly and stably solving the three-phase imbalance of the low-voltage distribution station area.
Disclosure of Invention
In view of the above, the embodiments of the present invention provide a method and a system for treating an edge side of three-phase load imbalance, so as to quickly and accurately implement real-time treatment of three-phase load imbalance.
The embodiment of the invention provides a three-phase load unbalance edge side treatment method, which comprises the following steps:
acquiring three-phase current data output by a transformer in a low-voltage distribution station area;
when the three-phase load unbalance calculated according to the three-phase current data does not meet a preset standard, calculating an intelligent reversing switch adjusting strategy according to the three-phase current data by adopting a reinforcement learning model constructed based on reinforcement learning, and sending the intelligent reversing switch adjusting strategy to the intelligent reversing switch;
and the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjusting strategy, so that the real-time treatment of the unbalanced three-phase load is realized.
In one embodiment, when a reinforcement learning model is constructed based on reinforcement learning, the current and the phase of an intelligent reversing switch are used as the environment state of the reinforcement learning, an intelligent reversing switch adjustment strategy is used as the action generated by the reinforcement learning, after the intelligent reversing switch adjustment is completed according to the intelligent reversing switch adjustment strategy output by the reinforcement learning, the newly calculated new three-phase load unbalance degree and the number of the intelligent reversing switches adjusted under the intelligent reversing switch adjustment strategy are counted, and the new three-phase load unbalance degree and the number of the intelligent reversing switches adjusted are used as the reward standard of the reinforcement learning.
In one embodiment, the intelligent commutation switch adjustment strategy includes phase currents to be switched by the intelligent commutation switches, and preferably, each intelligent commutation switch is represented by an action pair composed of 2 action values output by reinforcement learning, when the action pair is (0,0), the intelligent commutation switch maintains a current phase sequence, when the action pair is (1,0), the intelligent commutation switch is switched to an a phase sequence, when the action pair is (0,1), the intelligent commutation switch is switched to a B phase sequence, and when the action pair is (1,1), the intelligent commutation switch is switched to a C phase sequence.
In one embodiment, the step of using the new three-phase load unbalance and the adjusted intelligent reversing switch number as the reward standard for reinforcement learning comprises the following steps:
and defining the reward R of reinforcement learning according to the new three-phase load unbalance degree and the adjusted intelligent reversing switch number as follows:
wherein, beta represents the adjustment of the number T of the intelligent reversing switchesnThe weight occupied is sigma is a parameter for adjusting the magnitude of reward punishment, M is a preset unbalance standard, and rhoiIndicating a new three-phase load imbalance.
In one embodiment, when a DQN algorithm including a decision network and a target network is used to construct a reinforcement learning model, wherein the decision network and the target network both use fully connected networks, a Relu function is selected as an activation function, only a value of 0 or 1 is selected for an action that ensures output in the last layer, a Sigmoid function is selected as an activation function in the last layer, and a value of 0 is selected when a value output by a neuron is less than 0.5, and a value of 1 is selected when the value is greater than or equal to 0.5.
In one embodiment, in a Deep Q Network (DQN) algorithm, a value function of an environment state and an action pair is used as an action selection index, and an update process is as follows:
Q(s,a)＝r+γmaxQtarget(s',a)π(s,a)
wherein s represents the current environmental state, and s' represents the environmental state after the s state is transferred in the action a; q (s, a) and Qtarget(s', a) a value function representing pairs of environmental states and actions in the decision network and the target network; r is the reward value obtained by selecting the action a under the current environment state of the decision network; gamma is a discount quotation that represents the importance of future rewards.
In one embodiment, after the intelligent reversing switch adjustment strategy is obtained, the intelligent reversing switch adjustment strategy is required to be subjected to calculation and verification of the three-phase load unbalance degree, and when the calculation and verification result meets the unbalance degree standard range, the intelligent reversing switch adjustment strategy is sent to the intelligent reversing switch.
In one embodiment, when the calculation and verification result does not meet the unbalance degree standard range, condition reporting is carried out, expert personnel carry out remote operation adjustment on the intelligent reversing switch, and the adjustment is used as expert experience data for training the reinforcement learning model.
In one embodiment, after the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjusting strategy, the calculation and verification of the degree of unbalance of the three-phase load are carried out according to the adjusted new state of the intelligent reversing switch, and the real-time management of the unbalance of the three-phase load is completed.
The embodiment of the invention provides a three-phase load unbalance edge side treatment system, which comprises a transformer of a low-voltage distribution area, edge computing equipment arranged at the edge of the transformer, and a plurality of intelligent reversing switches for transmitting electric energy to a load, wherein the edge computing equipment is respectively in communication connection with the transformer and the intelligent reversing switches;
the edge computing equipment acquires three-phase current data output by the transformer;
when the three-phase load unbalance calculated according to the three-phase current data does not meet the preset standard, the edge calculation equipment adopts a reinforcement learning model constructed based on reinforcement learning to calculate an intelligent reversing switch adjustment strategy according to the three-phase current data, and then sends the intelligent reversing switch adjustment strategy to the intelligent reversing switch;
the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjusting strategy, and real-time treatment of unbalanced three-phase load is achieved.
According to the three-phase load unbalance edge side treatment method and system, the trained reinforcement learning model is deployed on the edge side of the low-voltage power distribution station area, the high complexity and time calculation cost of an iterative algorithm are avoided, meanwhile, the actual environment is considered, an edge calculation framework is adopted, low-voltage three-phase load unbalance treatment is locally achieved in the low-voltage power distribution station area through edge calculation equipment, an internet of things communication means and an intelligent reversing switch, the real-time response speed is high, the safety is high, and the pressure of a cloud network is reduced.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic structural diagram of a three-phase unbalanced load edge abatement system in one embodiment;
FIG. 2 is a flow chart of a three-phase unbalanced load edge treatment method in an embodiment;
FIG. 3 is a flow diagram illustrating training of reinforcement learning models in an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the detailed description and specific examples, while indicating the scope of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
In the face of the characteristics that the actual three-phase load unbalance management of a low-voltage distribution area requires low algorithm complexity, high real-time response, small number of adjustment reversing switches and the like, the embodiment provides a three-phase load unbalance edge side management method and a three-phase load unbalance edge side management system, combines the technologies of edge calculation, Internet of things, reinforcement learning and the like, dynamically monitors the three-phase load condition of the low-voltage distribution area in real time by utilizing a reinforcement learning model which is deployed by edge calculation equipment and is constructed based on reinforcement learning at the edge side of the area, and when the three-phase load unbalance occurs, the reinforcement learning model can issue an adjustment instruction to an intelligent phase-change switch in time by means of communication of the Internet of things, and completes the three-phase load unbalance management at the edge side of the low-voltage distribution area.
Fig. 1 is a schematic structural diagram of a three-phase load unbalance edge side treatment system in one embodiment. As shown in fig. 1, an embodiment provides a three-phase load imbalance edge-side remediation system in a low-voltage distribution grid scenario, including a transformer, an edge computing device, an intelligent diverter switch, and a unidirectional load. The edge computing device is a device which is deployed in the local area and has a computer function, and can run an algorithm model, read, calculate and store data, communicate with a cloud network, control the edge side device and the like. The transformer area transformer and the intelligent reversing switches of all branches are connected through communication means of the Internet of things such as RS485/WiFi/LORA/4G/5G, current values of phases A, B and C on the outlet side of the transformer and conduction states and current magnitudes of three-phase circuits of the intelligent reversing switches A, B and C of all branches can be obtained in real time, and each intelligent reversing switch is connected with each one-way load and is used for transmitting power to the one-way load through control.
And the edge computing equipment reads the current values of the A, B and C phases at the outlet sides of the transformer area at intervals of a certain time, and carries out three-phase load unbalance edge side treatment based on the current values of the A, B and C phases. Fig. 2 is a flowchart of a three-phase load imbalance edge-side treatment method in an embodiment. As shown in fig. 2, the three-phase load unbalance edge treatment method includes the following steps:
step 1, reading three-phase current data output by a transformer by edge computing equipment.
And 2, calculating whether the three-phase load unbalance degree at the moment meets a preset standard by the edge calculating equipment. If the preset standard is met, no processing is carried out.
In the embodiment, the following formula is adopted to calculate the three-phase load unbalance rho according to the read three-phase current data:
wherein, IA,IB,ICThe magnitude of the currents of A, B and C, IavRepresents the average of the three-phase currents, expressed as:
and the three-phase load unbalance rho is the maximum value of the unbalance of each phase current.
In the embodiment, the preset standard refers to the three-phase load unbalance degree of a low-voltage platform area according to the national power quality standard GB/T15543-2008 'power quality three-phase voltage allowable unbalance degree', and generally does not exceed 10%.
And 3, when the three-phase load unbalance degree does not meet the preset standard, calling a reinforcement learning model constructed based on reinforcement learning to calculate according to the three-phase current data to obtain an intelligent reversing switch adjustment strategy.
When the three-phase load unbalance degree calculated by the edge calculating equipment does not meet the preset standard, the low-voltage transformer area is considered to be in three-phase unbalance, and at the moment, a locally deployed reinforcement learning model is called to obtain an intelligent reversing switch adjustment strategy. And the edge computing equipment transmits the observed three-phase current data into the reinforcement learning model, and the reinforcement learning model generates corresponding intelligent reversing switch adjustment actions according to the trained strategy. After the adjustment action is generated, the edge computing equipment firstly simulates the three-phase load state after the adjustment action locally, and sends an action adjustment instruction to each intelligent reversing switch after the three-phase load requires; and if the three-phase load does not meet the requirement, remotely reporting, enabling the model to learn the state again, and updating the action selection strategy.
FIG. 3 is a flow diagram illustrating training of reinforcement learning models in an embodiment. As shown in fig. 3, the construction process of the reinforcement learning model is as follows:
s1, building a deep reinforcement learning model framework, and determining environmental states, intelligent body actions and reward standards of the environment to the actions in the reinforcement learning model;
s2, refining the framework of the reinforcement learning model by adopting a DQN (deep Q network) algorithm;
s3, training the hyperparameter in the DQN algorithm by using the three-phase data operated by the actual platform area;
s4, deploying the training converged DQN algorithm model to an edge computing gateway, and obtaining an optimal adjustment strategy through the DQN algorithm model when three-phase load imbalance occurs in the transformer area.
Preferably, in S1, building a deep reinforcement learning model framework, and determining the environment state, the agent action, and the reward criterion of the environment to the action in the reinforcement learning model comprises:
s11, taking the magnitude of three-phase current of each intelligent reversing switch A, B and C in the low-voltage distribution area as an environmental state S in reinforcement learning, and expressing as follows:
S＝{Ia1,Ib1,Ic1,Ia2,Ib2,Ic2...Ian,Ibn,Icn},
wherein Ian,Ibn,IcnAnd the current magnitude of the nth intelligent reversing switch A, B and C is represented.
S12, regarding the phase current to be switched by each intelligent phase-change switch in the low-voltage distribution substation area as an action a in reinforcement learning, the action is expressed as:
A＝{a10,a11,a20,a21,...an0,an1}
wherein a isn0,an1The action pair composed of (0,0) represents the action of the nth intelligent reversing switch, and an0,an1Can only take values of 0 or 1.
S13, according to the action output by the reinforcement learning model, after the adjustment of the reversing switches is completed, recalculating the three-phase unbalance degree of the platform area, and counting the number T of the adjustment reversing switches under the actionn. According to the actual treatment requirement, the treated three-phase load is not treatedAnd the balance degree and the number of the intelligent reversing switches are adjusted to serve as reward and punishment standards in the reinforcement learning model.
Specifically, the three-phase load unbalance is used as a main target of treatment to determine that a reward or a penalty is given, and the adjustment number of the intelligent reversing switch is used as a quality reference to determine the size of the reward or the penalty. When the treated three-phase load unbalance degree meets the specified requirements, the intelligent body in reinforcement learning is rewarded, and the smaller the number of the adjusted intelligent reversing switches is, the larger the reward is; similarly, when the treated three-phase unbalance degree does not meet the specified requirement, the intelligent body in reinforcement learning is given punishment, and the punishment is larger when the number of the adjusted intelligent reversing switches is smaller. Thus, for reward penalty R in reinforcement learning, it can be defined as:
wherein, β represents the weight occupied by the adjustment of the number of the reversing switches, σ is a constant and is used for adjusting the reward and punishment, M is a preset unbalance standard range, and is generally set to 10 according to the national electric energy quality standard GB/T15543-2008 "electric energy quality three-phase voltage allowable unbalance".
Preferably, in S2, the training of the hyper-parameters in the DQN algorithm by using the actual cell operation three-phase data includes:
s21, when a DQN algorithm containing a decision network and a target network is adopted to refine a frame of the reinforcement learning model, the neural network layers of the decision network and the target network adopt a full-connection mode, a Relu function is selected as an activation function, the value of an action A for ensuring output at the last layer is only 0 or 1, a Sigmoid function is selected as the activation function at the last layer, and when the value of the output of a neuron is less than 0.5, the value is 0, and when the value is more than or equal to 0.5, the value is 1.
Preferably, for action a ═ a10,a11,a20,a21,...an0,an1It is defined as follows: when (a)n0,an1) When equal to (0,0), it represents the nth intelligent reversing switchKeeping the current phase sequence; when (a)n0,an1) When the current is equal to (0,1), the nth intelligent reversing switch is switched to the phase sequence A; when (a)n0,an1) When the current is equal to (1,0), the nth intelligent reversing switch is switched to the phase B sequence; when (a)n0,an1) When the current is equal to (1,1), the nth intelligent reversing switch is switched to the phase sequence C; the rest remains the same.
S22, updating action selection strategy: namely, a value function Q (s, a) of a state and action pair in the DQN algorithm is used as an action selection index, and the updating process is as follows:
Q(s,a)＝r+γmaxQtarget(s',a)π(s,a)
wherein s represents the current environmental state, and s' represents the environmental state after the s state is transferred in the action a; q (s, a) and Qtarget(s', a) a value function representing pairs of environmental states and actions in the decision network and the target network; r is the reward value obtained by selecting the action a under the current environment state of the decision network; γ is a discount quotation indicating the importance of the future reward, and takes a value of (0, 1).
Preferably, in S3, the training of the hyper-parameters in the DQN algorithm by using the actual cell-running three-phase data includes the following processes:
s31, initializing a reinforcement learning model, building two neural networks of a decision network and a target network, initializing parameters, and determining a learning rate alpha, a maximum iterative training time K and a discount elicitor gamma;
and S32, selecting the action a for different station area states S according to the action selection strategy pi (S, a), searching for the action a, obtaining rewards, storing the states, the actions and the rewards searched each time into an experience pool as experience, and adding part of expert experience.
S33, the agent will randomly draw the experience in the memory pool to learn and follow the new action selection strategy in the training process.
S34, in the process of training the reinforcement learning model, the intelligent agent needs to obtain a better action selection strategy through a large amount of exploration and learning, in order to accelerate the learning process of the intelligent agent and ensure the learning quality, a semi-supervised training mode is adopted, and a part of proportion of expert experience (corresponding to the best action in a part of states) is artificially added into the memory pool to accelerate the learning of the intelligent agent. By the semi-supervised training mode, the convergence speed of the reinforcement learning model can be accelerated. The trained reinforcement learning is deployed to an edge computing gateway in the transformer area, when the gateway detects three-phase imbalance through computing, the state of the transformer area at the moment is transmitted into a reinforcement learning model, and the model automatically computes an intelligent reversing switch adjusting strategy.
And 4, the edge computing equipment calculates and verifies the adjustment strategy of the intelligent reversing switch, calculates whether the adjusted three-phase load unbalance degree meets the requirement or not, and transmits the adjustment strategy to each intelligent reversing switch through the Internet of things communication means after the requirement is met.
And 5, reporting the condition when the adjusted three-phase load unbalance does not meet the requirement, carrying out remote operation adjustment by related personnel, adding the intelligent reversing switch adjustment as expert experience into an experience pool for model training, retraining the model, and after the training is finished, carrying out on-line deployment on the model again.
After the adjustment strategy is issued, the intelligent reversing switch is adjusted according to the strategy rules, the state of the intelligent reversing switch needs to be uploaded to the edge computing equipment again, the edge computing equipment checks the state of the adjusted intelligent reversing switch, the three-phase unbalance degree of the platform area is recalculated, and the treatment of the three-phase load unbalance edge side of the platform area is completed.
The above-mentioned embodiments are intended to illustrate the technical solutions and advantages of the present invention, and it should be understood that the above-mentioned embodiments are only the most preferred embodiments of the present invention, and are not intended to limit the present invention, and any modifications, additions, equivalents, etc. made within the scope of the principles of the present invention should be included in the scope of the present invention.
Claims (10)
1. A three-phase load unbalance edge side treatment method is characterized by comprising the following steps:
acquiring three-phase current data output by a transformer in a low-voltage distribution station area;
when the three-phase load unbalance calculated according to the three-phase current data does not meet a preset standard, calculating an intelligent reversing switch adjusting strategy according to the three-phase current data by adopting a reinforcement learning model constructed based on reinforcement learning, and sending the intelligent reversing switch adjusting strategy to the intelligent reversing switch;
and the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjusting strategy, so that the real-time treatment of the unbalanced three-phase load is realized.
2. The three-phase load unbalance edge side treatment method according to claim 1, wherein when the reinforcement learning model is constructed based on reinforcement learning, the magnitude and phase of the current of the intelligent reversing switch are used as the environment state of the reinforcement learning, the adjustment strategy of the intelligent reversing switch is used as the action generated by the reinforcement learning, after the adjustment of the intelligent reversing switch is completed according to the adjustment strategy of the intelligent reversing switch output by the reinforcement learning, the newly calculated new three-phase load unbalance degree and the statistics of the number of the intelligent reversing switches adjusted under the adjustment strategy of the intelligent reversing switch are carried out, and the new three-phase load unbalance degree and the adjusted number of the intelligent reversing switches are used as the reward standard of the reinforcement learning.
3. The three-phase load imbalance edge-side governance method according to claim 1 or 2, wherein the intelligent commutation switch adjustment strategy includes phase currents to be switched by the intelligent commutation switches, and preferably, each intelligent commutation switch is represented by an action pair composed of 2 action values output by reinforcement learning, the intelligent commutation switch maintains a current phase sequence when the action pair is (0,0), the intelligent commutation switch is switched to the phase a sequence when the action pair is (1,0), the intelligent commutation switch is switched to the phase B sequence when the action pair is (0,1), and the intelligent commutation switch is switched to the phase C sequence when the action pair is (1, 1).
4. The three-phase load unbalance edge treatment method according to claim 2, wherein the step of using the new three-phase load unbalance degree and the adjusted intelligent reversing switch number as the reward criteria for reinforcement learning comprises the following steps:
and defining the reward R of reinforcement learning according to the new three-phase load unbalance degree and the adjusted intelligent reversing switch number as follows:
wherein, beta represents the adjustment of the number T of the intelligent reversing switchesnThe weight occupied is sigma is a parameter for adjusting the magnitude of reward punishment, M is a preset unbalance standard, and rhoiIndicating a new three-phase load imbalance.
5. The three-phase load unbalance edge side treatment method according to claim 1, characterized in that when a DQN algorithm including a decision network and a target network is used to construct the reinforcement learning model, wherein the decision network and the target network both use fully connected networks, the activation function selects the Relu function, and only takes a value of 0 or 1 for the action of ensuring output in the last layer, the activation function of the last layer selects the Sigmoid function, and when the value of the neuron output is less than 0.5, the value is 0, and when the value is greater than or equal to 0.5, the value is 1.
6. The three-phase load unbalance edge side treatment method according to claim 5, wherein in the DQN algorithm, a value function of an environment state and an action pair is adopted as an action selection index, and the updating process is as follows:
Q(s,a)＝r+γmaxQtarget(s',a)π(s,a)
wherein s represents the current environmental state, and s' represents the environmental state after the s state is transferred in the action a; q (s, a) and Qtarget(s', a) a value function representing pairs of environmental states and actions in the decision network and the target network; r is the reward value obtained by selecting the action a under the current environment state of the decision network; gamma is a discount quotation that represents the importance of future rewards.
7. The three-phase load unbalance edge side treatment method according to any one of claims 1 to 3, wherein after the intelligent reversing switch adjustment strategy is obtained, the intelligent reversing switch adjustment strategy is further subjected to calculation and verification of the three-phase load unbalance degree, and when a calculation and verification result meets an unbalance degree standard range, the intelligent reversing switch adjustment strategy is sent to the intelligent reversing switch.
8. The three-phase load unbalance edge side treatment method according to claim 7, wherein when the calculation and verification result does not satisfy the unbalance degree standard range, condition reporting is performed, an expert performs remote operation adjustment on the intelligent reversing switch, and the adjustment is used as expert experience data for training the reinforcement learning model.
9. The three-phase load unbalance edge side treatment method according to claim 1, wherein after the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjustment strategy, the calculation and verification of the three-phase load unbalance degree are performed according to the adjusted new state of the intelligent reversing switch, and the real-time treatment of the three-phase load unbalance is completed.
10. A three-phase load unbalance edge side treatment system comprises a transformer of a low-voltage distribution area, edge computing equipment arranged at the edge of the transformer and a plurality of intelligent reversing switches for transmitting electric energy to a load, wherein the edge computing equipment is respectively in communication connection with the transformer and the intelligent reversing switches; it is characterized in that the preparation method is characterized in that,
the edge computing equipment acquires three-phase current data output by the transformer;
when the three-phase load unbalance calculated according to the three-phase current data does not meet the preset standard, the edge calculation equipment adopts a reinforcement learning model constructed based on reinforcement learning to calculate an intelligent reversing switch adjustment strategy according to the three-phase current data, and then sends the intelligent reversing switch adjustment strategy to the intelligent reversing switch;
the intelligent reversing switch adjusts the state of the intelligent reversing switch according to the received adjusting strategy, and real-time treatment of unbalanced three-phase load is achieved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110522264.3A CN113206507B (en) | 2021-05-13 | 2021-05-13 | Three-phase load unbalance edge side treatment method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110522264.3A CN113206507B (en) | 2021-05-13 | 2021-05-13 | Three-phase load unbalance edge side treatment method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113206507A true CN113206507A (en) | 2021-08-03 |
| CN113206507B CN113206507B (en) | 2022-05-27 |
Family
ID=77030965
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110522264.3A Active CN113206507B (en) | 2021-05-13 | 2021-05-13 | Three-phase load unbalance edge side treatment method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113206507B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118157333A (en) * | 2024-05-13 | 2024-06-07 | 苏州火炎焱网络科技有限公司 | Intelligent monitoring method and system for distribution transformer area based on Internet of things technology |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110460072A (en) * | 2019-08-21 | 2019-11-15 | 天津大学 | Consider the real-time Management strategy of low-voltage platform area three-phase imbalance problem of phase index |
| CN110690718A (en) * | 2019-08-29 | 2020-01-14 | 国网新疆电力有限公司吐鲁番供电公司 | Three-phase load unbalance treatment control method |
| CN111200290A (en) * | 2020-03-16 | 2020-05-26 | 广东电网有限责任公司 | Intelligent control method of phase change switch for three-phase unbalance treatment of transformer area |
| CN112149347A (en) * | 2020-09-16 | 2020-12-29 | 北京交通大学 | Power distribution network load transfer method based on deep reinforcement learning |
| CN112186799A (en) * | 2020-09-22 | 2021-01-05 | 中国电力科学研究院有限公司 | Distributed energy system autonomous control method and system based on deep reinforcement learning |
-
2021
- 2021-05-13 CN CN202110522264.3A patent/CN113206507B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110460072A (en) * | 2019-08-21 | 2019-11-15 | 天津大学 | Consider the real-time Management strategy of low-voltage platform area three-phase imbalance problem of phase index |
| CN110690718A (en) * | 2019-08-29 | 2020-01-14 | 国网新疆电力有限公司吐鲁番供电公司 | Three-phase load unbalance treatment control method |
| CN111200290A (en) * | 2020-03-16 | 2020-05-26 | 广东电网有限责任公司 | Intelligent control method of phase change switch for three-phase unbalance treatment of transformer area |
| CN112149347A (en) * | 2020-09-16 | 2020-12-29 | 北京交通大学 | Power distribution network load transfer method based on deep reinforcement learning |
| CN112186799A (en) * | 2020-09-22 | 2021-01-05 | 中国电力科学研究院有限公司 | Distributed energy system autonomous control method and system based on deep reinforcement learning |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118157333A (en) * | 2024-05-13 | 2024-06-07 | 苏州火炎焱网络科技有限公司 | Intelligent monitoring method and system for distribution transformer area based on Internet of things technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113206507B (en) | 2022-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110535146B (en) | Electric power system reactive power optimization method based on depth determination strategy gradient reinforcement learning | |
| Hu et al. | Volt/VAr control in distribution systems using a time-interval based approach | |
| CN103746374B (en) | Containing the cyclization control method of many microgrids power distribution network | |
| CN106549394A (en) | Electric power idle work optimization system and method based on double fish-swarm algorithms | |
| JP3825171B2 (en) | Power distribution system control system | |
| Ulinuha et al. | Optimal scheduling of LTC and shunt capacitors in large distorted distribution systems using evolutionary-based algorithms | |
| CN116454926B (en) | Multi-type resource cooperative regulation and control method for three-phase unbalanced management of distribution network | |
| CN110163540B (en) | Power system transient stability prevention control method and system | |
| CN115313403A (en) | Real-time voltage regulation and control method based on deep reinforcement learning algorithm | |
| CN110932281A (en) | Multi-section cooperative correction method and system based on quasi-steady-state sensitivity of power grid | |
| JP3825172B2 (en) | Power flow calculator for distribution system control system. | |
| CN113206507B (en) | Three-phase load unbalance edge side treatment method and system | |
| CN108053066A (en) | The Forecasting Methodology of adjustable electric load in industrial production user's energy management system | |
| Zhang et al. | Deep reinforcement learning for load shedding against short-term voltage instability in large power systems | |
| CN114742283A (en) | Medium-voltage line loop closing current estimation and risk assessment method based on deep learning network | |
| CN113890039B (en) | Multi-terminal flexible direct-current power distribution network power flow scheduling optimization method | |
| CN111193295A (en) | Distribution network flexibility improvement robust optimization scheduling method considering dynamic reconfiguration | |
| CN109861254A (en) | The implementation method and device of load access micro-capacitance sensor | |
| JP3825173B2 (en) | Power distribution system control system | |
| CN116799780A (en) | Distributed photovoltaic bearing capacity assessment method, system and terminal for power distribution network | |
| CN117057623A (en) | Comprehensive power grid safety optimization scheduling method, device and storage medium | |
| CN110649633A (en) | Power distribution network reactive power optimization method and system | |
| CN107681666B (en) | A kind of rural power grids allocated radio low-voltage comprehensive processing method and device | |
| CN115632406A (en) | Reactive voltage control method and system based on digital-mechanism fusion drive modeling | |
| Wang et al. | Reinforcement learning based voltage control using multiple control devices |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220425 Address after: 325600 in talent Park, No. 268, Central Avenue, Yueqing Economic Development Zone, Yueqing City, Wenzhou City, Zhejiang Province Applicant after: Yueqing Yangtze River Delta Electrical Engineer Innovation Center Address before: 325600 No. 268, Central Avenue, Yueqing Economic Development Zone, Wenzhou City, Zhejiang Province Applicant before: Yueqing engineer Innovation Service Center |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |