CN112330021A - Network coordination control method of distributed optical storage system - Google Patents

Network coordination control method of distributed optical storage system Download PDF

Info

Publication number
CN112330021A
CN112330021A CN202011222997.7A CN202011222997A CN112330021A CN 112330021 A CN112330021 A CN 112330021A CN 202011222997 A CN202011222997 A CN 202011222997A CN 112330021 A CN112330021 A CN 112330021A
Authority
CN
China
Prior art keywords
time
data
real
historical
photovoltaic
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.)
Pending
Application number
CN202011222997.7A
Other languages
Chinese (zh)
Inventor
吕冬翔
胡秉晨
王焘
朱立宏
孙子路
李钊
贾子熙
钟豪
仇海波
赵彬涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Cetc Energy Co ltd
Original Assignee
CETC 18 Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by CETC 18 Research Institute filed Critical CETC 18 Research Institute
Priority to CN202011222997.7A priority Critical patent/CN112330021A/en
Publication of CN112330021A publication Critical patent/CN112330021A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • 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
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • 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
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Educational Administration (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Power Engineering (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)
  • Molecular Biology (AREA)

Abstract

A network coordination control method of a distributed optical storage system, the method comprising the steps of: acquiring and processing real-time photovoltaic data, real-time load data, historical photovoltaic data and historical load data of the distributed light storage system; inputting the processed real-time photovoltaic data, the processed real-time load data, the processed historical photovoltaic data and the processed historical load data into a long-time and short-time neural network model; the long-time neural network model outputs a photovoltaic power generation capacity predicted value and a load demand predicted value at a future moment; acquiring current charge data of each subsystem in the distributed optical storage system; inputting the current charge data, the photovoltaic power generation capacity predicted value and the load demand predicted value of each subsystem into a deep reinforcement learning model; and the deep reinforcement learning model performs optimized scheduling control on the distributed light storage system. The energy utilization efficiency of the system can be effectively improved, and the energy of each power supply module is balanced.

Description

Network coordination control method of distributed optical storage system
Technical Field
The invention belongs to the technical field of distributed energy scheduling control, and particularly relates to a network coordination control method of a distributed optical storage system.
Background
The distributed energy network is formed by connecting a large number of power generation, energy storage and power supply devices. Because there is the unmatched problem of power generation link photovoltaic and load power value, need store unnecessary electric quantity when photovoltaic cell power supply is more sufficient usually, let energy storage battery carry out energy supply when photovoltaic cell power supply is not enough, wherein the energy storage link plays the effect of peak clipping and valley filling. However, due to the random characteristics of the photovoltaic cell, the energy storage cell and the power load, the energy imbalance phenomenon easily occurs at each part of the energy network, so that the utilization rate of the system energy and the energy storage capacity is reduced, and the quality of electric energy is reduced.
At present, an artificial intelligence algorithm is one of the popular solutions to the above problems, and a value-based reinforcement learning algorithm such as Q-learning is widely used in the energy field due to its simple and practical characteristics. And the Q-learning enables the intelligent agent to interact with the environment according to the set reward rule and collects the interactive reward to update the Q-table until the optimal decision Q-table is found, and the training process is completed. However, since the basis of the selected action of the Q-learning algorithm in the training process is a discrete Q function, for a continuous problem, the method is difficult to handle a complex problem of continuous or high latitude, and a single non-prediction algorithm is difficult to make a long-term scheduling plan.
Disclosure of Invention
In order to solve the above problem, the present invention provides a network coordination control method for a distributed optical storage system, where the method includes:
acquiring and processing real-time photovoltaic data, real-time load data, historical photovoltaic data and historical load data of the distributed light storage system;
inputting the processed real-time photovoltaic data, the processed real-time load data, the processed historical photovoltaic data and the processed historical load data into a long-time and short-time neural network model;
the long-time neural network model outputs a photovoltaic power generation capacity predicted value and a load demand predicted value at a future moment;
acquiring current charge data of each subsystem in the distributed optical storage system;
inputting the current charge data, the photovoltaic power generation capacity predicted value and the load demand predicted value of each subsystem into a deep reinforcement learning model;
and the deep reinforcement learning model performs optimized scheduling control on the distributed light storage system.
Preferably, the acquiring and processing real-time photovoltaic data, real-time load data, historical photovoltaic data and historical load data of the distributed light storage system includes the steps of:
collecting real-time photovoltaic power generation data and a real-time load power sequence of an i node at the current moment and historical photovoltaic power generation data and a historical load power sequence of previous k moments in the distributed optical storage system;
regularization processing is respectively carried out on the real-time photovoltaic power generation data and the historical photovoltaic power generation data;
and respectively carrying out thermal coding processing on the real-time load power sequence and the historical load power sequence.
Preferably, the expression of the real-time photovoltaic power generation data of the i node at the current moment is as follows:
Gik(T)＝{pig(T)，pig(T-1)，...，pig(T-k)}；
the expression of the historical load power sequence at the first k moments is as follows:
Lik(T)＝{pio(T)，pio(T-1)，...，pio(T-k)}；
wherein, G isik(T) represents real-time photovoltaic power generation data of the i-node at the present time, Lik(T) represents the historical load power sequence at the k previous times, i is 1, 2, 3, k is 10, T represents the current time, p represents the power at time Tig0Representing the power from the moment of the present start to the kth time node, pio0Indicating the power of the node from time t to the kth time in the historical data.
Preferably, the expression after the real-time photovoltaic power generation data regularization is as follows:
Figure BDA0002762711870000031
the expression after the historical photovoltaic power generation data is regularized is as follows:
wherein,
Figure BDA0002762711870000032
representing the real-time photovoltaic power generation data after regularization,
Figure BDA0002762711870000033
representing the normalized historical photovoltaic power generation data, pig0Representing the power from the moment of the present start to the kth time node, pig *Representing the set maximum value, G, of the real-time photovoltaic power generation dataik(T) represents the inodeI is 1, 2, 3, k is 10, T represents the current time, and p represents the power at time T.
Preferably, the expression of the real-time load power sequence is:
Tk＝{T，T-1，T-2，...，T-k}；
the expression of the historical load power sequence is as follows:
Mk＝{M0，M1，M2，...，Mk}；
wherein T represents a time node, Tk represents a real-time load power sequence of a kth time node, k represents the number of the time nodes, and Mk represents a historical load power sequence of the kth time node.
Preferably, the step of inputting the processed real-time photovoltaic data, the real-time load data, the historical photovoltaic data and the historical load data into the long-term and short-term neural network model includes the steps of:
designing the long-time neural network model;
forming an input matrix by the real-time photovoltaic power generation data and the historical photovoltaic power generation data which are subjected to regularization processing and the real-time load power sequence and the historical load power sequence which are subjected to thermal coding processing;
and inputting the input matrix into the long-time and short-time neural network model.
Preferably, the expression of the photovoltaic power generation amount predicted value is:
Figure BDA0002762711870000034
the expression of the load demand predicted value is as follows:
Figure BDA0002762711870000041
wherein,
Figure BDA0002762711870000042
represents the predicted value of the photovoltaic power generation amount,
Figure BDA0002762711870000043
represents the predicted load demand value, j is 10, P represents the power at time T, and P represents the predicted load demand valueig0Representing the short-time power, p, from the moment of the present start to the kth time nodeio0The short-time power from the time T to the kth time node in the historical data is shown, and T represents the time node.
Preferably, the optimal scheduling control of the distributed light storage system by the deep reinforcement learning model comprises the following steps:
combining the current state of charge of each power module of the distributed light storage system, the photovoltaic power generation capacity predicted value and the load demand predicted value into state variables;
defining action variables and reward functions;
performing deep learning training by adopting a strategy-based near-end optimization strategy algorithm;
the circulation interacts with the environment, and collects corresponding state, action and advantage function, so as to find out and obtain the optimum action strategy pi by adopting a strategy gradient methodθ。
Preferably, the expression of the action variable is:
Action(T)＝{a1，a2，a3}|al|∈[ptmin，ptmax]；
wherein P represents the power at time T, a1, a2, a3 represent the operating variables at a certain time, and PtminRepresenting the minimum value of power at time t, ptmaxRepresenting the maximum power at time t.
Preferably, the expression of the reward function is:
Figure BDA0002762711870000044
where P denotes the power at time T, T denotes the current time, C is the compensation factor of the reward function, w1 denotes the coefficients of the reward function, and m denotes the process constants of the different scenarios in the reward function.
The network coordination control method of the distributed optical storage system is correspondingly applied to energy management of the distributed electric energy network in a mode of combining the neural network with deep learning, the energy utilization efficiency of the system can be effectively improved, and the energy of each power module is balanced.
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, and 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 diagram of a network coordination control method of a distributed optical storage system provided in the present invention;
fig. 2 is a schematic diagram of an electric energy network in a network coordination control method of a distributed optical storage system according to the present invention;
fig. 3 is a schematic diagram of a long-term and short-term neural network prediction model in a network coordination control method of a distributed optical storage system according to the present invention;
fig. 4 is a flowchart of a deep learning algorithm in the network coordination control method of the distributed optical storage system provided by the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings in conjunction with the following detailed description. It should be understood that the description is intended to be exemplary only, and is not intended to limit the scope of the present invention. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present invention.
As shown in fig. 1 to 4, in the embodiment of the present application, the present invention provides a network coordination control method of a distributed optical storage system, where the method includes the steps of:
s1: acquiring and processing real-time photovoltaic data, real-time load data, historical photovoltaic data and historical load data of the distributed light storage system;
s2: inputting the processed real-time photovoltaic data, the processed real-time load data, the processed historical photovoltaic data and the processed historical load data into a long-time and short-time neural network model;
s3: the long-time neural network model outputs a photovoltaic power generation capacity predicted value and a load demand predicted value at a future moment;
s4: acquiring current charge data of each subsystem in the distributed optical storage system;
s5: inputting the current charge data, the photovoltaic power generation capacity predicted value and the load demand predicted value of each subsystem into a deep reinforcement learning model;
s6: and the deep reinforcement learning model performs optimized scheduling control on the distributed light storage system.
When the network coordination control is carried out on the distributed light storage system, the real-time photovoltaic data, the real-time load data, the historical photovoltaic data and the historical load data of the distributed light storage system are firstly obtained and processed, then inputting the processed real-time photovoltaic data, the real-time load data, the historical photovoltaic data and the historical load data into a long-time and short-time neural network model, then the long-time neural network model outputs a photovoltaic power generation amount predicted value and a load demand predicted value at the future moment, then obtaining the current charge data of each subsystem in the distributed light storage system, inputting the current charge data of each subsystem, the photovoltaic power generation amount predicted value and the load demand predicted value into a deep reinforcement learning model, and then the deep reinforcement learning model performs optimized scheduling control on the distributed light storage system.
In this embodiment of the present application, the acquiring and processing the real-time photovoltaic data, the real-time load data, the historical photovoltaic data, and the historical load data of the distributed light storage system in step S1 includes the steps of:
collecting real-time photovoltaic power generation data and a real-time load power sequence of an i node at the current moment and historical photovoltaic power generation data and a historical load power sequence of previous k moments in the distributed optical storage system;
regularization processing is respectively carried out on the real-time photovoltaic power generation data and the historical photovoltaic power generation data;
and respectively carrying out thermal coding processing on the real-time load power sequence and the historical load power sequence.
When the real-time photovoltaic data, the real-time load data, the historical photovoltaic data and the historical load data of the distributed light storage system are obtained and processed, firstly, the real-time photovoltaic power generation data and the real-time load power sequence of an i node in the distributed light storage system at the current moment and the historical photovoltaic power generation data and the historical load power sequence of the previous k moments are collected, then the real-time photovoltaic power generation data and the historical photovoltaic power generation data are respectively subjected to regularization processing, and then the real-time load power sequence and the historical load power sequence are respectively subjected to thermal coding processing.
In this embodiment of the present application, the expression of the real-time photovoltaic power generation data of the i node at the current time is as follows:
Gik(T)＝{pig(T)，pig(T-1)，...，pig(T-k)}；
the expression of the historical load power sequence at the first k moments is as follows:
Lik(T)＝{pio(T)，pio(T-1)，...，pio(T-k)}；
wherein G isik(T) represents real-time photovoltaic power generation data of the i-node at the present time, Lik(T) represents the historical load power sequence at the k previous times, i is 1, 2, 3, k is 10, T represents the current time, p represents the power at time Tig0Representing the power from the moment of the present start to the kth time node, pio0Indicating the power of the node from time t to the kth time in the historical data.
In this embodiment of the present application, the expression after the real-time photovoltaic application data regularization is:
Figure BDA0002762711870000071
the expression after the historical photovoltaic power generation data is regularized is as follows:
wherein,
Figure BDA0002762711870000072
representing the real-time photovoltaic power generation data after regularization,
Figure BDA0002762711870000073
representing the normalized historical photovoltaic power generation data, pig0Representing the power from the moment of the present start to the kth time node, pig *Representing the set maximum value, G, of the real-time photovoltaic power generation dataik(T) represents real-time photovoltaic power generation data at the current time of the i-node, i is 1, 2, 3, k is 10, T represents the current time, and p represents power at the time of T.
In this embodiment of the present application, the expression of the real-time load power sequence is:
Tk＝{T，T-1，T-2，...，T-k}；
the expression of the historical load power sequence is as follows:
Mk＝{M0，M1，M2，...，Mk}；
wherein T represents a time node, Tk represents a real-time load power sequence of a kth time node, k represents the number of the time nodes, and Mk represents a historical load power sequence of the kth time node.
In an embodiment of the present application, the inputting the processed real-time photovoltaic data, the real-time load data, the historical photovoltaic data, and the historical load data into the long-term and short-term neural network model includes:
designing the long-time neural network model;
forming an input matrix by the real-time photovoltaic power generation data and the historical photovoltaic power generation data which are subjected to regularization processing and the real-time load power sequence and the historical load power sequence which are subjected to thermal coding processing;
and inputting the input matrix into the long-time and short-time neural network model.
In the embodiment of the present application, the expression of the photovoltaic power generation amount predicted value is:
Figure BDA0002762711870000081
the expression of the load demand predicted value is as follows:
Figure BDA0002762711870000082
wherein,
Figure BDA0002762711870000083
represents the predicted value of the photovoltaic power generation amount,
Figure BDA0002762711870000084
represents the predicted load demand value, j is 10, P represents the power at time T, and P represents the predicted load demand valueig0Representing the short-time power, p, from the moment of the present start to the kth time nodeio0The short-time power from the time T to the kth time node in the historical data is shown, and T represents the time node.
In an embodiment of the present application, the performing, by the deep reinforcement learning model, optimized scheduling control on the distributed light storage system includes:
combining the current state of charge of each power module of the distributed light storage system, the photovoltaic power generation capacity predicted value and the load demand predicted value into state variables;
defining action variables and reward functions;
performing deep learning training by adopting a strategy-based near-end optimization strategy algorithm;
the circulation interacts with the environment, and collects corresponding state, action and advantage function, so as to find out the optimal action by adopting a strategy gradient methodStrategy piθ。
In the embodiment of the present application, the expression of the action variable is:
Action(T)＝{a1，a2，a3}|al|∈[ptmin，ptmax]；
wherein P represents the power at time T, a1, a2, a3 represent the operating variables at a certain time, and PtminRepresenting the minimum value of power at time t, ptmaxRepresenting the maximum power at time t.
In the embodiment of the present application, the expression of the reward function is:
Figure BDA0002762711870000091
where P denotes the power at time T, T denotes the current time, C is the compensation factor of the reward function, w1 denotes the coefficients of the reward function, and m denotes the process constants of the different scenarios in the reward function.
The present application is described in detail below with specific examples.
As shown in fig. 1, the distributed optical storage system is formed by connecting 3 energy modules with power generation-energy storage-energy supply functions as nodes, the nodes are connected through power gateways, and the power gateways have a power transmission function and can continuously transmit power from one node to another node. In order to enable the system to operate efficiently and uniformly, the transmission power and direction of each power gateway need to be controlled, so that the energy of each node is complemented.
Therefore, a network coordination control method of a distributed optical storage system is provided, and the basic idea is as follows: the prediction model predicts photovoltaic power generation power values and load power values at 10 future moments on the basis of historical information, and then inputs the predicted power value sequence and the current system state information into a near-end strategy optimization (PPO) model of deep reinforcement learning to obtain control actions of 3 power gateways, wherein the actions tend to be optimal for a long term because the future predicted information is used.
The invention provides a network coordination control method of a distributed optical storage system, which specifically comprises the following steps:
(1) design of long-time and short-time neural network energy prediction model
The photovoltaic power generation power and the load power have time sequence correlation, and the photovoltaic load power value at the later moment can be predicted by the power values at the first 10 moments. And a long-term and short-term neural network for learning long-term dependence is adopted. Considering that photovoltaic power generation and load power values are continuous variables, and the dimension can be greatly changed. In order to overcome the problem that different dimensions have great influence on the prediction result, the input power value is expanded to the range of [0, 1] by adopting a regularization method.
Gik(T)＝{pig(T)，pig(T-1)，...，pig(T-k)}
Lik(T)＝{pio(T)，pio(T-1)，...，pio(T-k)}
i＝1，2，3 k＝10
Wherein G isik(T)，LikAnd (T) are power generation and load power value sequences of the current time and the k previous times of the node at the T time.
Figure BDA0002762711870000101
Figure BDA0002762711870000102
Wherein,
Figure BDA0002762711870000103
respectively, normalized photovoltaic power generation, load power sequence, pig *And pio *The maximum values of the photovoltaic power generation and the load power are set respectively.
In addition, because the time sequence and the task sequence are discrete sequences, the values of the sequences can only be discrete integer values, and the measurement distances of all the values in the sequences should be the same, the values of the discrete features are expanded into the Euclidean space in a thermal coding mode, and a certain value of the discrete features corresponds to a certain point of the Euclidean space.
Tk＝{T，T-1，T-2，...，T-k}
Mk＝{M0，M1，M2，...，Mk}
k＝10
Wherein, TkAnd MkRespectively time sequence and task number sequence, and respectively corresponding to hot code sequence after hot coding
Figure BDA0002762711870000104
And
Figure BDA0002762711870000105
regularizing the photovoltaic power generation power value
Figure BDA0002762711870000106
Load power value Lik(T) and thermally encoded time series
Figure BDA0002762711870000111
Sequence of task numbers
Figure BDA0002762711870000112
Composing an input matrix
Figure BDA0002762711870000113
Inputting the long and short time neural network model to obtain the prediction of the photovoltaic power generation power value and the load power value of each node at 10 moments in the future
Figure BDA0002762711870000114
And
Figure BDA0002762711870000115
Figure BDA0002762711870000116
Figure BDA0002762711870000117
the long-short time neural network model design adopts 3 layers of long-short time neural networks, each layer of long-short time neural network comprises 20 neurons, each neuron adopts a leak-relu activation function, and the output of each layer of LSTM network is connected with batch regularization operation so as to help the convergence of the neural networks.
(2) Design of coordination control algorithm based on deep reinforcement learning
On the basis of the prediction information, the deep reinforcement learning algorithm can make an action which accords with the optimal overall value in a future period of time at the current moment. Here we use the current state of charge SoC of each power module of the systemi(i 1, 2, 3), predicted power for 10 time instants
Figure BDA0002762711870000118
The combination is the state variable:
Figure BDA0002762711870000119
the control action is completed by the power gateway, and the transmission power of the power gateway ranges from [ p ]tminptmax]Within this range, the flow direction and the magnitude of the power can be continuously controlled. Defining action variables:
Action(T)＝{a1，a2，a3}|al|∈[ptmin，ptmax]。
the reward function comprehensively considers the predicted residual energy and the balance in the system under the action to provide judgment basis for the action. And the energy size of the power supply module with the minimum predicted residual energy in the network is used as a reward item. And meanwhile, introducing the variance of each power supply module as a penalty term:
Figure BDA0002762711870000121
wherein, w1And w2The weight coefficients are corresponding terms respectively, the first term is the predicted energy value of the energy minimum node in the electric energy network after 10 moments, CiThe current energy value of the i node is shown, and delta T is the time interval of adjacent moments; second term ε (SoC)1～3(T)) is the variance of the current state of charge of each node.
Based on the design, the deep reinforcement learning model is constructed, a near-end optimization strategy algorithm (PPO) based on a strategy is adopted for training, and the training process is shown in FIG. 3. Through continuous interaction with the environment and collection of corresponding state, action and advantage functions, the optimal action strategy pi is found and obtained by adopting a strategy gradient methodθ. In order to avoid that the action difference generated by the new strategy and the old strategy is too large, which causes the non-convergence of the training process, the change degree of the new strategy and the old strategy needs to be limited, a KL penalty term is introduced, and the coefficients are adaptively adjusted. And continuously updating the strategy until the optimal strategy in the running process of the system.
Through verification, the electric energy network shown in fig. 1 is controlled by using the coordination control algorithm, the generated power and the load power of each power module in the electric energy network can be accurately predicted, and the energy utilization efficiency and the balance performance are greatly improved.
The network coordination control method of the distributed optical storage system is correspondingly applied to energy management of the distributed electric energy network in a mode of combining the neural network with deep learning, the energy utilization efficiency of the system can be effectively improved, and the energy of each power module is balanced.
It is to be understood that the above-described embodiments of the present invention are merely illustrative of or explaining the principles of the invention and are not to be construed as limiting the invention. Therefore, any modification, equivalent replacement, improvement and the like made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. Further, it is intended that the appended claims cover all such variations and modifications as fall within the scope and boundaries of the appended claims or the equivalents of such scope and boundaries.

Claims (10)

1. A network coordination control method of a distributed optical storage system, the method comprising the steps of:
acquiring and processing real-time photovoltaic data, real-time load data, historical photovoltaic data and historical load data of the distributed light storage system;
inputting the processed real-time photovoltaic data, the processed real-time load data, the processed historical photovoltaic data and the processed historical load data into a long-time and short-time neural network model;
the long-time neural network model outputs a photovoltaic power generation capacity predicted value and a load demand predicted value at a future moment;
acquiring current charge data of each subsystem in the distributed optical storage system;
inputting the current charge data, the photovoltaic power generation capacity predicted value and the load demand predicted value of each subsystem into a deep reinforcement learning model;
and the deep reinforcement learning model performs optimized scheduling control on the distributed light storage system.
2. The network coordination control method of the distributed light storage system according to claim 1, wherein said acquiring and processing real-time photovoltaic data, real-time load data, historical photovoltaic data and historical load data of the distributed light storage system comprises the steps of:
collecting real-time photovoltaic power generation data and a real-time load power sequence of an i node at the current moment and historical photovoltaic power generation data and a historical load power sequence of previous k moments in the distributed optical storage system;
regularization processing is respectively carried out on the real-time photovoltaic power generation data and the historical photovoltaic power generation data;
and respectively carrying out thermal coding processing on the real-time load power sequence and the historical load power sequence.
3. The network coordination control method of the distributed light storage system according to claim 2, wherein the expression of the real-time photovoltaic power generation data of the i-node at the current time is as follows:
Gik(T)＝{pig(T)，pig(T-1)，...，pig(T-k)}；
the expression of the historical load power sequence at the first k moments is as follows:
Lik(T)＝{pio(T)，pio(T-1)，...，pio(T-k)}；
wherein, G isik(T) represents real-time photovoltaic power generation data of the i-node at the present time, Lik(T) represents the historical load power sequence at the k previous times, i is 1, 2, 3, k is 10, T represents the current time, p represents the power at time Tig0Representing the power from the moment of the present start to the kth time node, pio0Indicating the power of the node from time t to the kth time in the historical data.
4. The network coordination control method of the distributed light storage system according to claim 2, wherein the expression after the real-time photovoltaic development data regularization is:
Figure FDA0002762711860000021
the expression after the historical photovoltaic power generation data is regularized is as follows:
wherein,
Figure FDA0002762711860000022
representing the real-time photovoltaic power generation data after regularization,
Figure FDA0002762711860000023
the historical photovoltaic power generation data p after the abnormal table is normalizedig0Indicating the moment of time from now onPower to the kth time node, pig *Representing the set maximum value, G, of the real-time photovoltaic power generation dataik(T) represents real-time photovoltaic power generation data at the current time of the i-node, i is 1, 2, 3, k is 10, T represents the current time, and p represents power at the time of T.
5. The network coordination control method of the distributed optical storage system according to claim 2, wherein the expression of the real-time load power sequence is:
Tk＝{T，T-1，T-2，...，T-k}；
the expression of the historical load power sequence is as follows:
Mk＝{M0，M1，M2，...，Mk}；
wherein T represents a time node, Tk represents a real-time load power sequence of a kth time node, k represents the number of the time nodes, and Mk represents a historical load power sequence of the kth time node.
6. The network coordination control method of the distributed light storage system according to claim 1, wherein the step of inputting the processed real-time photovoltaic data, the real-time load data, the historical photovoltaic data and the historical load data into the long-term neural network model comprises the steps of:
designing the long-time neural network model;
forming an input matrix by the real-time photovoltaic power generation data and the historical photovoltaic power generation data which are subjected to regularization processing and the real-time load power sequence and the historical load power sequence which are subjected to thermal coding processing;
and inputting the input matrix into the long-time and short-time neural network model.
7. The network coordination control method of the distributed light storage system according to claim 1, wherein the expression of the photovoltaic power generation amount predicted value is as follows:
Figure FDA0002762711860000031
the expression of the load demand predicted value is as follows:
Figure FDA0002762711860000032
wherein,
Figure FDA0002762711860000033
represents the predicted value of the photovoltaic power generation amount,
Figure FDA0002762711860000034
represents the predicted load demand value, j is 10, P represents the power at time T, and P represents the predicted load demand valueig0Representing the short-time power, p, from the moment of the present start to the kth time nodeio0The short-time power from the time T to the kth time node in the historical data is shown, and T represents the time node.
8. The network coordination control method of the distributed light storage system according to claim 1, wherein the deep reinforcement learning model performs optimal scheduling control on the distributed light storage system, and comprises the following steps:
combining the current state of charge of each power module of the distributed light storage system, the photovoltaic power generation capacity predicted value and the load demand predicted value into state variables;
defining action variables and reward functions;
performing deep learning training by adopting a strategy-based near-end optimization strategy algorithm;
the circulation interacts with the environment, and collects corresponding state, action and advantage function, so as to find out and obtain the optimum action strategy pi by adopting a strategy gradient methodθ。
9. The network coordination control method of the distributed light storage system according to claim 8, wherein the expression of the action variable is:
Action(T)＝{a1，a2，a3}|al|∈[ptmin，ptmax]；
wherein P represents the power at time T, a1, a2, a3 represent the operating variables at a certain time, and PtminRepresenting the minimum value of power at time t, ptmaxRepresenting the maximum power at time t.
10. The network coordination control method of the distributed light storage system according to claim 8, wherein the expression of the reward function is:
Figure FDA0002762711860000041
where P denotes the power at time T, T denotes the current time, C is the compensation factor of the reward function, w1 denotes the coefficients of the reward function, and m denotes the process constants of the different scenarios in the reward function.
CN202011222997.7A 2020-11-05 2020-11-05 Network coordination control method of distributed optical storage system Pending CN112330021A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011222997.7A CN112330021A (en) 2020-11-05 2020-11-05 Network coordination control method of distributed optical storage system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011222997.7A CN112330021A (en) 2020-11-05 2020-11-05 Network coordination control method of distributed optical storage system

Publications (1)

Publication Number Publication Date
CN112330021A true CN112330021A (en) 2021-02-05

Family

ID=74317215

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011222997.7A Pending CN112330021A (en) 2020-11-05 2020-11-05 Network coordination control method of distributed optical storage system

Country Status (1)

Country Link
CN (1) CN112330021A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115840794A (en) * 2023-02-14 2023-03-24 国网山东省电力公司东营供电公司 Photovoltaic system planning method based on GIS (geographic information System) and RL (Link State) models
CN116388279A (en) * 2023-05-23 2023-07-04 安徽中超光电科技有限公司 Grid-connected control method and control system for solar photovoltaic power generation system
CN116722561A (en) * 2023-04-28 2023-09-08 国网河北省电力有限公司电力科学研究院 Reactive power compensation system
CN118017614A (en) * 2024-01-31 2024-05-10 江南大学 Distributed new energy storage regulation and control method and system
CN118350609A (en) * 2024-06-17 2024-07-16 山东欧通信息科技有限公司 Distributed energy optimization scheduling method and system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104104116A (en) * 2014-07-01 2014-10-15 杭州电子科技大学 Design method for photovoltaic microgrid supply-demand control system containing distributed energy sources
CN105914769A (en) * 2016-05-03 2016-08-31 易事特集团股份有限公司 Optical storage micro-grid energy management method and system in off-network state
CN111008728A (en) * 2019-11-01 2020-04-14 深圳供电局有限公司 Method for predicting short-term output of distributed photovoltaic power generation system
CN112101651A (en) * 2020-09-08 2020-12-18 中国电子科技集团公司第十八研究所 Electric energy network coordination control method and system and information data processing terminal

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104104116A (en) * 2014-07-01 2014-10-15 杭州电子科技大学 Design method for photovoltaic microgrid supply-demand control system containing distributed energy sources
CN105914769A (en) * 2016-05-03 2016-08-31 易事特集团股份有限公司 Optical storage micro-grid energy management method and system in off-network state
CN111008728A (en) * 2019-11-01 2020-04-14 深圳供电局有限公司 Method for predicting short-term output of distributed photovoltaic power generation system
CN112101651A (en) * 2020-09-08 2020-12-18 中国电子科技集团公司第十八研究所 Electric energy network coordination control method and system and information data processing terminal

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115840794A (en) * 2023-02-14 2023-03-24 国网山东省电力公司东营供电公司 Photovoltaic system planning method based on GIS (geographic information System) and RL (Link State) models
CN115840794B (en) * 2023-02-14 2023-05-02 国网山东省电力公司东营供电公司 Photovoltaic system planning method based on GIS and RL models
CN116722561A (en) * 2023-04-28 2023-09-08 国网河北省电力有限公司电力科学研究院 Reactive power compensation system
CN116388279A (en) * 2023-05-23 2023-07-04 安徽中超光电科技有限公司 Grid-connected control method and control system for solar photovoltaic power generation system
CN116388279B (en) * 2023-05-23 2024-01-23 安徽中超光电科技有限公司 Grid-connected control method and control system for solar photovoltaic power generation system
CN118017614A (en) * 2024-01-31 2024-05-10 江南大学 Distributed new energy storage regulation and control method and system
CN118350609A (en) * 2024-06-17 2024-07-16 山东欧通信息科技有限公司 Distributed energy optimization scheduling method and system

Similar Documents

Publication Publication Date Title
CN112330021A (en) Network coordination control method of distributed optical storage system
CN109347149A (en) Micro-capacitance sensor energy storage dispatching method and device based on depth Q value network intensified learning
CN111884213A (en) Power distribution network voltage adjusting method based on deep reinforcement learning algorithm
CN108964050A (en) Micro-capacitance sensor dual-layer optimization dispatching method based on Demand Side Response
CN112101651B (en) Electric energy network coordination control method, system and information data processing terminal
CN113572157A (en) User real-time autonomous energy management optimization method based on near-end policy optimization
CN115409431B (en) Distributed power resource scheduling method based on neural network
Li et al. Day-ahead optimal dispatch strategy for active distribution network based on improved deep reinforcement learning
CN116436003B (en) Active power distribution network risk constraint standby optimization method, system, medium and equipment
CN116667325B (en) Micro-grid-connected operation optimization scheduling method based on improved cuckoo algorithm
Georgiou et al. Implementing artificial neural networks in energy building applications—A review
CN117455183B (en) Comprehensive energy system optimal scheduling method based on deep reinforcement learning
Xing et al. Real-time optimal scheduling for active distribution networks: A graph reinforcement learning method
CN113344283B (en) Energy internet new energy consumption capability assessment method based on edge intelligence
Leo et al. Multi agent reinforcement learning based distributed optimization of solar microgrid
CN112084680B (en) Energy internet optimization strategy method based on DQN algorithm
Yang et al. Cloud-edge-end intelligence for fault-tolerant renewable energy accommodation in smart grid
CN117117989A (en) Deep reinforcement learning solving method for unit combination
CN112003279A (en) Method for evaluating new energy consumption capability of hierarchical micro-grid
CN116362504A (en) Optimal scheduling method for electric heating combined energy system, terminal equipment and storage medium
CN116345462A (en) Power system power flow calculation method for driving dynamic neural network by using physical equation
CN115313510A (en) Adaptive reactive compensation photovoltaic inverter control method and system
CN115411776A (en) Thermoelectric cooperative scheduling method and device for residential comprehensive energy system
CN114792974A (en) Method and system for energy optimization management of interconnected micro-grid
CN114583696A (en) Power distribution network reactive power optimization method and system based on BP neural network and scene matching

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: 20211125

Address after: 300384 No. 6 Huake No. 7 Road, Binhai New Area, Tianjin Binhai High-tech Industrial Development Zone

Applicant after: The 18th Research Institute of China Electronics Technology Group Corporation

Applicant after: CETC Energy Co.,Ltd.

Address before: 300384 No. 6 Huake No. 7 Road, Binhai New Area, Tianjin Binhai High-tech Industrial Development Zone

Applicant before: The 18th Research Institute of China Electronics Technology Group Corporation

TA01 Transfer of patent application right
TA01 Transfer of patent application right

Effective date of registration: 20220929

Address after: 300384 No. 6 Huake No. 7 Road, Binhai New Area, Tianjin Binhai High-tech Industrial Development Zone

Applicant after: CETC Energy Co.,Ltd.

Address before: 300384 No. 6 Huake No. 7 Road, Binhai New Area, Tianjin Binhai High-tech Industrial Development Zone

Applicant before: The 18th Research Institute of China Electronics Technology Group Corporation

Applicant before: CETC Energy Co.,Ltd.

CB02 Change of applicant information
CB02 Change of applicant information

Address after: 300384 No. 6 Huake No. 7 Road, Binhai New Area, Tianjin Binhai High-tech Industrial Development Zone

Applicant after: CETC Blue Sky Technology Co.,Ltd.

Address before: 300384 No. 6 Huake No. 7 Road, Binhai New Area, Tianjin Binhai High-tech Industrial Development Zone

Applicant before: CETC Energy Co.,Ltd.

RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210205