CN115049098B - Data-driven-based optimal energy flow random optimization method for electric-water combined system - Google Patents
Data-driven-based optimal energy flow random optimization method for electric-water combined system Download PDFInfo
- Publication number
- CN115049098B CN115049098B CN202210389338.5A CN202210389338A CN115049098B CN 115049098 B CN115049098 B CN 115049098B CN 202210389338 A CN202210389338 A CN 202210389338A CN 115049098 B CN115049098 B CN 115049098B
- Authority
- CN
- China
- Prior art keywords
- water
- electric
- power
- flow
- load
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 200
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000005457 optimization Methods 0.000 title claims abstract description 29
- 238000004364 calculation method Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims description 64
- 238000013528 artificial neural network Methods 0.000 claims description 43
- 238000013135 deep learning Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 20
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012614 Monte-Carlo sampling Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 8
- 238000000605 extraction Methods 0.000 abstract description 2
- 238000009826 distribution Methods 0.000 description 12
- 238000005070 sampling Methods 0.000 description 4
- 230000004075 alteration Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000000342 Monte Carlo simulation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000011282 treatment Methods 0.000 description 1
- 238000003809 water extraction Methods 0.000 description 1
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
-
- 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
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Software Systems (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Computer Hardware Design (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Medical Informatics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Geometry (AREA)
- Educational Administration (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a data-driven-based optimal energy flow random optimization method of an electric-water combined system, which comprises the following steps of: 1) Establishing an electricity-water combined energy flow model; 2) And calculating the pipeline flow, the node water head and the branch power predicted value of the electric-water combined system by using the electric-water combined energy flow model. The invention provides a data-driven-based power-water system optimal energy flow random optimization method, which aims at fully considering the characteristic that the connection between a deterministic optimization process and a probability characteristic extraction process is weak, and can effectively solve the problem of overlong power-water system optimal energy flow calculation time.
Description
Technical Field
The invention relates to the technical field of comprehensive energy, in particular to a data-driven electric-water combined system optimal energy flow random optimization method.
Background
The energy is a resource which is indistinct from human production and life, and is an important material foundation related to national economy; with the advancement of energy revolution and the realization of dual carbon targets, the unique advantages of integrating energy systems (INTEGRATED ENERGY SYSTEM, IES) to realize the mutual economy and comprehensive utilization of multiple energy sources are rapidly developed and widely applied.
Although in the traditional sense the power system and the water delivery and distribution system are not coupled to each other. But in fact the two systems are closely related to each other. The water distribution system is characterized in that the power system consumes a large amount of water for power production, and the water delivery and distribution system consumes a large amount of power for water extraction, treatment and distribution. This interdependence is commonly referred to as water energy, and with the continued development and advancement of research into integrated energy systems, the concept of an electro-water combination system (INTEGRATED ELECTRICITY-WATER SYSTEM, electro-water combination system) has also been sequentially proposed.
However, because both electrical and water loads have uncertainties, these uncertainties will have an impact on the integrated power-water system optimal energy flow (CC-OPWF) model. The existing random optimization method for solving IEWS optimal energy flows usually adopts Monte-Card sampling to extract probability characteristics, and solves the problems of repeated energy flow calculation and low calculation efficiency, and how to reduce the solving time while guaranteeing the optimal energy flow calculation solving precision becomes a problem to be solved urgently.
Disclosure of Invention
The invention aims to provide a data-driven-based optimal energy flow random optimization method for an electric-water combined system, which comprises the following steps of:
1) Establishing an electricity-water combined energy flow model;
the electric-water combined energy flow model is obtained by training a load sample and a control variable.
The step of establishing the electricity-water combined energy flow model comprises the following steps:
1.1 Basic data of the electric-water combined system is obtained, monte Carlo sampling is carried out, n groups of load samples, control variables and reservoir flow Q tank are obtained, and a training data set is established;
the load sample comprises water load power P w and electric load power P D; the control variables include generator output P gen;
The basic data of the electric-water combined system comprises the number N Wbus of nodes in a water network, the number N Wbranch of water network branches, the number N bus of nodes in a power network, the number N branch of power network branches, the number N gen of generators, the output P gen of the generators, the water load power P w, the electric load power P D and the flow Q tank of a reservoir; wherein the water load power Electric load power/>Generator output/> The water load power of the N Wbus water network node, the electric load power of the N bus power network node and the output of the N gen generator are respectively shown.
1.2 Carrying out deterministic combined power flow calculation by using the training data set to obtain the state quantity of the electric-water combined system and taking the state quantity as the label information of the training data; the state quantity comprises pipeline flow, node water head and branch power value; the branch power is calculated through a direct current power flow model; the pipeline flow, the node water head and the branch power value are calculated by the formulas (1) - (2):
Δhp,ij＝hi-hj＝kp,ijqp,ij|qp,ij|n-1 (1)
Pij＝Bij(δi-δj) (2)
Wherein h i、hj represents the water head of the nodes i and j; k p,ij is the coefficient of friction of the pipeline in terms of the loss of path; q p,ij is the pipe flow; Δh p,ij represents the pipe flow; p ij is the branch power value; b ij is susceptance; delta i、δj is the voltage phase angle of nodes i, j;
1.3 Training the deep learning neural network by using a training data set with state quantity, and establishing an electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the state quantity.
The electric-water combined energy flow model comprises an electric-water combined energy flow model for fitting the relation between a load sample and a control variable and pipeline flow, an electric-water combined energy flow model for fitting the relation between the load sample and a node water head, and an electric-water combined energy flow model for fitting the relation between the load sample and the control variable and branch power.
The output of the electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the pipeline flow is the inversely normalized pipeline flow;
The output of the electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the node water head is the inverse normalized node water head;
The output of the electro-water joint energy flow model used to fit the relationship between load samples, control variables and branch power is the inverse normalized branch power.
The training of the deep learning neural network comprises the following steps:
1.3.1 Pre-processing the training data to obtain pre-processed electric load power P Dy, water load power P wy, generator output P geny, reservoir flow Q tank, pipeline flow, node water head and branch power values; the preprocessing includes normalization.
1.3.2 Inputting the preprocessed electric load power P Dy, the water load power P wy, the generator output P geny and the reservoir flow Q tank into the deep learning neural network as shown in formulas (3) - (4);
h0＝{PDy;Pwy;Pgeny;Qtank} (3)
Wherein h 0 represents an original input layer of the deep learning neural network; h i represents the output of the ith hidden layer of the model; w i is the weight of the i-th hidden layer; b i is the bias of the ith hidden layer; Φ () is an activation function; h i-1 represents the output of the i-1 th hidden layer of the model;
1.3.3 Calculating the loss function of the deep learning neural network, and completing the training of the deep learning neural network when the loss function reaches the minimum value and is not changed along with the training times.
The loss function is set as a state quantity predicted valueThe square of the difference from the state quantity true value P branch.
2) And calculating the pipeline flow, the node water head and the branch power predicted value of the electric-water combined system by using the electric-water combined energy flow model.
The step of calculating the predicted values of the pipeline flow, the node water head and the branch power of the electric-water combined system comprises the following steps:
2.1 Initializing an uncertainty boundary, namely setting an upper uncertainty boundary and a lower uncertainty boundary of pipeline flow, node water head and branch power to be zero, and initializing iteration times k=1;
2.2 Acquiring constraint conditions and network states of the combined electric-water system;
2.3 Taking the current uncertainty boundary as a condition, solving the certainty optimization parameters of the electric-water combined system to obtain a load sample and a control variable value;
2.4 Inputting the load sample and the control variable value into an electric-water combined energy flow model to obtain predicted values of pipeline flow, node water head and branch power;
2.5 Counting probability characteristics of the pipeline flow, the node water head and the branch power predicted value, and correcting the uncertainty boundary according to the probability characteristics obtained by counting and the current uncertainty boundary to obtain a new uncertainty boundary;
The uncertainty boundary is calculated as follows:
Where λ U represents the uncertainty upper boundary; alpha is a safety margin factor between 0 and 1; epsilon is the quantile under a given security probability; s represents the true value of the state variable; Characterizing a predicted value of a state variable;
the relationship between the true value and the predicted value of the state variable satisfies the following equation:
wherein: delta represents the prediction error of the load;
2.6 Judging whether the new uncertainty boundary meets the convergence condition, if so, outputting the predicted values of the pipeline flow, the node water head and the branch power, otherwise, returning to the step 3), and enabling the iteration times k=k+1; the convergence condition is that the iteration number k is larger than k max and max (delta f (k))＜ε;kmax is the maximum iteration number; epsilon is convergence accuracy.
Δf(k)＝f(k+1)-f(k) (8)
Wherein: f (k+1) represents the uncertainty boundary value obtained in the (k+1) th iteration, and f (k) represents the uncertainty boundary value obtained in the (k) th iteration; Δf (k) represents the difference between this uncertainty boundary value and the value of the uncertainty boundary of the last iteration.
The invention provides a random optimization method of an electric power-water system based on data driving, which aims at fully considering the characteristic that the relation between a deterministic optimization process and a probability feature extraction process is weaker, and can effectively solve the problem of overlong calculation time of the optimal energy flow of the electric power-water system.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a flow chart of a conventional method;
FIG. 3 is a diagram of a system architecture of an exemplary method of the present invention;
FIG. 4 is a diagram of a neural network training process;
FIGS. 5 (a) - (c) are graphs of neural network flow, head, branch power prediction accuracy;
FIGS. 6 (a) - (c) are graphs of flow, head, and branch power accuracy versus time for two methods;
FIG. 7 is a graph of time versus sample number for two methods of calculation.
Detailed Description
The present invention is further described below with reference to examples, but it should not be construed that the scope of the above subject matter of the present invention is limited to the following examples. Various substitutions and alterations are made according to the ordinary skill and familiar means of the art without departing from the technical spirit of the invention, and all such substitutions and alterations are intended to be included in the scope of the invention.
Example 1:
Referring to fig. 1 to 5, the data-driven electric-water combined system optimal energy flow random optimization method comprises the following steps:
1) Establishing an electricity-water combined energy flow model;
the electric-water combined energy flow model is obtained by training a load sample and a control variable.
The step of establishing the electricity-water combined energy flow model comprises the following steps:
1.1 Basic data of the electric-water combined system is obtained, monte Carlo sampling is carried out, n groups of load samples, control variables and reservoir flow Q tank are obtained, and a training data set is established;
the load sample comprises water load power P w and electric load power P D; the control variables include generator output P gen;
The basic data of the electric-water combined system comprises the number N Wbus of nodes in a water network, the number N Wbranch of water network branches, the number N bus of nodes in a power network, the number N branch of power network branches, the number N gen of generators, the output P gen of the generators, the water load power P w, the electric load power P D and the flow Q tank of a reservoir; wherein the water load power Electric load power/>Generator output/> The water load power of the N Wbus water network node, the electric load power of the N bus power network node and the output of the N gen generator are respectively shown.
1.2 Carrying out deterministic combined power flow calculation by using the training data set to obtain the state quantity of the electric-water combined system and taking the state quantity as the label information of the training data; the state quantity comprises pipeline flow, node water head and branch power value; the branch power is calculated through a direct current power flow model; the pipeline flow, the node water head and the branch power value are calculated by the formulas (1) - (2):
Δhp,ij＝hi-hj＝kp,ijqp,ij|qp,ij|n-1 (1)
Pij＝Bij(δi-δj) (2)
Wherein h i、hj represents the water head of the nodes i and j; k p,ij is the coefficient of friction of the pipeline in terms of the loss of path; q p,ij is the pipe flow; Δh p,ij represents the pipe flow; p ij is the branch power value; b ij is susceptance; delta i、δj is the voltage phase angle of nodes i, j; n is a calculation coefficient.
1.3 Training the deep learning neural network by using a training data set with state quantity, and establishing an electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the state quantity.
The electric-water combined energy flow model comprises an electric-water combined energy flow model for fitting the relation between a load sample and a control variable and pipeline flow, an electric-water combined energy flow model for fitting the relation between the load sample and a node water head, and an electric-water combined energy flow model for fitting the relation between the load sample and the control variable and branch power.
The output of the electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the pipeline flow is the inversely normalized pipeline flow;
The output of the electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the node water head is the inverse normalized node water head;
The output of the electro-water joint energy flow model used to fit the relationship between load samples, control variables and branch power is the inverse normalized branch power.
The training of the deep learning neural network comprises the following steps:
1.3.1 Pre-processing the training data to obtain pre-processed electric load power P Dy, water load power P wy, generator output P geny, reservoir flow Q tank, output label pipeline flow, node water head and branch power value; ; the preprocessing includes normalization.
1.3.2 Inputting the preprocessed electric load power P Dy, the water load power P wy, the generator output P geny and the reservoir flow Q tank into the deep learning neural network as shown in formulas (3) - (4);
h0＝{PDy;Pwy;Pgeny;Qtank} (3)
Wherein h 0 represents an original input layer of the deep learning neural network; h i represents the output of the ith hidden layer of the model; w i is the weight of the i-th hidden layer; b i is the bias of the ith hidden layer; Φ () is an activation function;
1.3.3 Calculating the loss function of the deep learning neural network, and completing the training of the deep learning neural network when the loss function reaches the minimum value and is not changed along with the training times.
The loss function is set as a state quantity predicted valueThe square of the difference from the state quantity true value P branch.
2) And calculating the pipeline flow, the node water head and the branch power predicted value of the electric-water combined system by using the electric-water combined energy flow model.
The step of calculating the predicted values of the pipeline flow, the node water head and the branch power of the electric-water combined system comprises the following steps:
2.1 Initializing an uncertainty boundary, namely setting an upper uncertainty boundary and a lower uncertainty boundary of pipeline flow, node water head and branch power to be zero, and initializing iteration times k=1;
2.2 Acquiring constraint conditions and network states of the combined electric-water system;
2.3 Taking the current uncertainty boundary as a condition, solving the certainty optimization parameters of the electric-water combined system to obtain a load sample and a control variable value;
2.4 Inputting the load sample and the control variable value into an electric-water combined energy flow model to obtain predicted values of pipeline flow, node water head and branch power;
2.5 Counting probability characteristics of the pipeline flow, the node water head and the branch power predicted value, and correcting the uncertainty boundary according to the probability characteristics obtained by counting and the current uncertainty boundary to obtain a new uncertainty boundary;
The uncertainty boundary is calculated as follows:
Where λ U represents the uncertainty upper boundary; alpha is a safety margin factor between 0 and 1; epsilon is the quantile under a given security probability; s represents the true value of the state variable; Characterizing a predicted value of a state variable;
the relationship between the true value and the predicted value of the state variable satisfies the following equation:
wherein: delta represents the prediction error of the load;
2.6 Judging whether the new uncertainty boundary meets the convergence condition, if so, outputting the predicted values of the pipeline flow, the node water head and the branch power, otherwise, returning to the step 3), and enabling the iteration times k=k+1; the convergence condition is that the iteration number k is larger than k max and max (delta f (k))＜ε;kmax is the maximum iteration number; epsilon is convergence accuracy.
Δf(k)＝f(k+1)-f(k) (8)
Wherein: f (k+1) represents the uncertainty boundary value obtained in the (k+1) th iteration, and f (k) represents the uncertainty boundary value obtained in the (k) th iteration; Δf (k) represents the difference between this uncertainty boundary value and the value of the uncertainty boundary of the last iteration.
Example 2:
a data-driven electric-water combined system optimal energy flow random optimization method. The main steps are divided into two stages, wherein the first stage is obtained through offline training, and updating in subsequent calculation is not needed.
The first stage: step 1: acquiring a load sample and a control variable value as training data based on Monte Carlo simulation; step 2: calculating the electricity-water combined power flow to obtain corresponding load state variables as tag information of training data; step 3: training the deep learning neural network, setting super parameters such as the number of hidden layers, the number of neurons, the learning rate, an activation function and the like, and training the training sample and the label information obtained in the steps to obtain the neural network capable of fitting the relation between the training sample and the label information.
And a second stage: step 1: initializing an uncertainty boundary, and acquiring constraint conditions and network states of the electric-water combined system; step 2: solving a deterministic optimization problem influenced by an uncertainty boundary to obtain an initial control variable value; step 3: substituting the load sample obtained by Monte Carlo sampling and the initial control variable value obtained in the step 2 into a neural network in the first stage, and outputting a predicted value of a state variable of the load sample; step 4: and counting probability characteristics of the load sample state variable predicted value, and correcting an uncertainty boundary of the load sample state variable according to the obtained probability characteristics. And (3) repeating the steps 2-4 until the uncertainty boundary meets the convergence condition, and outputting the magnitudes of the control variable and the load state variable value. The specific implementation steps are as follows:
first stage neural network training process
1.1 Obtaining training samples
Collecting basic data of a combined electric-water system, wherein the basic data comprises the number N Wbus of nodes in a water network; the number of branches N Wbranch, the number of nodes N bus in the power grid, the number of branches N branch, the number of generators N gen, the generator output P gen, the water load power P w and the electric load power P D, wherein
The Monte Carlo is sampled n times to obtain n groups of load power values P w、PD and control variable values P gen which meet normal distribution; and carrying out deterministic combined power flow calculation according to the obtained load sample and the control variable value to obtain n groups of pipeline flow, node water head and branch power values.
1.2 Neural network construction
To ensure that the training produces more accurate results, separate feed-forward neural networks are trained for pipeline flow, node head, and branch power values. The training process is as follows:
Firstly, preprocessing training data to obtain preprocessed electric load power P Dy, water load power P wy, generator output P geny, reservoir flow Q tank, output label pipeline flow, node water head and branch power value.
The preprocessed P Dy、Pwy、Pgeny、Qtank is then input into the deep learning neural network as follows:
h0＝{PDy;Pwy;Pgeny;Qtank} (1)
Where h 0 represents the original input layer of the deep learning neural network, h i represents the output of the ith hidden layer of the model, w i is the weight of the ith hidden layer, b i is the bias of the ith hidden layer, and Φ () is the activation function; taking a neural network for predicting the pipeline flow as an example, the loss function is set as a pipeline flow predicted value Square of the difference from the true value P branch of the pipe flow:
and the step-by-step reduction process of the loss function, namely the process of continuously training the fitting input-output relationship by the neural network, wherein the neural network training is considered to be completed when the loss function reaches the minimum value and is not changed along with the training times through continuous training of the deep learning neural network.
1.3 Data inverse normalization
Further, the predicted value of the pipeline flow obtained in the step 1.2The method is the data after pretreatment, if an actual value is required to be obtained, the data is subjected to inverse normalization to obtain an inverse normalized actual predicted value/>
Solving uncertainty boundaries in the second stage
2.1 Initializing uncertainty boundaries
Setting the upper uncertainty boundary and the lower uncertainty boundary of the pipeline flow, the node water head and the branch power to be zero, and initializing the iteration times k=1; and obtaining the constraint condition and the network state of the electric-water combined system.
2.2 Solving deterministic optimization results
And (3) solving the deterministic optimization problem of the electric-water combined system by taking the uncertainty boundary initialized in the step (2.1) as a condition to obtain values of control variables such as node load of the reservoir, active output of the generator and the like.
2.3 Solving the probabilistic load flow
And (2) extracting n groups of electrical load samples p and water load samples q meeting normal distribution by Monte Carlo, and inputting the control variable values obtained in the step (2.2) into the neural network obtained by training in the first stage to obtain predicted values of pipeline flow, node water head and branch power.
2.4 Correction of uncertainty boundaries
And (3) counting probability characteristics of pipeline flow, node water head and branch power predicted values, and further correcting the uncertainty boundary according to the probability characteristics obtained through statistics and the initial uncertainty boundary to obtain a new uncertainty boundary.
Example 3:
as shown in fig. 3, taking an electricity-water combined system formed by an IEEE-14 node electric power system and a 7 node water network system as an example, an optimal energy flow random optimization method of the electricity-water combined system based on data driving is as follows:
1) Inputting basic data and initializing
1.1 Inputting basic data
Basic parameters of a water network are input, and an upper limit value kmax of iteration times, convergence accuracy epsilon, water network constraint conditions and power grid constraint conditions of the method are set.
Wherein, the water network node parameter, the branch parameter, the water pump parameter, the reservoir parameter and the cost data are respectively given in tables 1, 2, 3, 4 and 5.
Water load of water distribution network of 17 nodes of meter
| Node numbering | Elevation/m | Water load (m 3/h) | Minimum head (m) |
| 1 | 15 | 0 | 0 |
| 2 | 28.2 | 0 | 5 |
| 3 | 15.4 | 241.12 | 0 |
| 4 | 13.6 | 296.41 | 2 |
| 5 | 18.8 | 390.21 | 2 |
| 6 | 19.2 | 230.3 | 2 |
| 7 | 18.9 | 465.44 | 0 |
Branch data of water distribution network of 27 nodes of table
| Head-end node | End node | Length/m | Diameter (mm) | Coefficient of friction | Loss factor |
| 3 | 4 | 450 | 500 | 0.2 | 0 |
| 4 | 5 | 590 | 500 | 0.2 | 0 |
| 6 | 7 | 600 | 600 | 0.2 | 0 |
| 2 | 5 | 550 | 500 | 0.2 | 0 |
| 4 | 6 | 650 | 500 | 0.2 | 0 |
| 5 | 7 | 650 | 600 | 0.2 | 0 |
Water pump parameter of water distribution network of 37 nodes of table
| Head-end node | End node | Initial water head | Internal resistance coefficient | Head index | Efficiency of | Minimum rotational speed | Maximum rotation speed |
| 1 | 7 | 80 | 0.000001544 | 2 | 0.9 | 0 | 1.5 |
Parameters of water distribution net reservoir of 4 7 nodes of meter
| Node | Elevation (m) | Flow rate | Initial head (m) | Maximum head (m) | Minimum head (m) | Maximum in-out flow (m 3/s) |
| 2 | 28.2 | 0 | 20 | 5 | 70 | 0.14 |
Table 57 node distribution network cost data
| Electricity price (dollar/kWh) | Water price (dollar/ton) |
| 0.048 | 0.826 |
1.2 Initializing uncertainty boundaries)
The upper uncertainty boundary and the lower uncertainty boundary of the flow, the water head and the branch power are set to 0.
1.3 Solving deterministic equivalence problems bounded by uncertainty boundaries)
And solving a deterministic equivalence problem limited by the uncertainty boundary according to the initial uncertainty boundary to obtain the active output P gen of the generator and the node load Q tank of the reservoir.
1.4 Monte Carlo sampling
The Monte Carlo extracts 5000 groups of electric load samples p and water load samples q with the original load of the Monte Carlo as expectations and the variance of the electric load samples p and the water load samples q being 5% of expected values.
1.5 Neural network prediction
According to the neural network obtained by offline training, the obtained P gen,Qtank, P and q are input into the neural network to obtain the predicted values of pipeline flow, node water head and branch power values
1.6 Probability characteristics of statistical predictions
And counting the upper and lower bounds of the obtained predicted pipeline flow, the upper and lower bounds of the water head, and the upper and lower bounds of the branch power.
1.7 Correction of uncertainty boundaries
And calculating to obtain a new uncertainty boundary by the initial uncertainty boundary and the probability characteristics of the statistically obtained predicted value.
2) Offline training of neural networks
2.1 Obtaining training samples and preprocessing
The electric load sample p and the water load sample q obtained in 1.4 are used for randomly constructing 5000 groups of reservoir node loads q tank and generator active force p gen influenced by uncertainty boundaries to jointly form training data of a neural network:
Substituting training data into an electric-water combined system to solve deterministic power flow to obtain 5000 groups of pipeline flow q state, node water head h state and branch power value p state, and taking the 5000 groups of pipeline flow q state and the node water head h state as training labels:
the training data preprocessing process is as follows:
Wherein, For the preprocessed training data, x mean is the mean value of the training data and x std is the standard deviation of the training data.
The training label pretreatment process comprises the following steps:
Wherein, For the preprocessed training label, y mean is the average value of the training label, and y std is the standard deviation of the training label
2.2 Neural network training
The loss function is set as:
In order to improve accuracy, q state,hstate,pstate is used as a training label, three independent forward feedback neural networks are obtained through training, the number of hidden layers is set to be 3, an activation function of each hidden layer adopts a 'relu' function with good convergence performance, and the learning rate, namely the gradient update rate, is set to be 1e-4.
3) Experimental effect
The effectiveness and superiority of the method of the invention are compared and verified by taking an electric-water combined system consisting of an IEEE-14 power system and a 7-node water distribution network system shown in the figure 3 as a simulation object through the following two methods. The method provided by the invention is different from the traditional method in that complex repeated probability power flow calculation is avoided, and the neural network is used for efficiently and quickly replacing the original calculation. Method 1: and carrying out Meng Ka sampling and probability power flow calculation for 5000 times by adopting a traditional random optimization method. Method 2: by adopting the method of the invention, the sample is still carried out for 5000 times Meng Ka, but the neural network is used for replacing the original complex load flow calculation.
(1) Comparison of calculated results
Figure 6 shows a comparison of the calculation accuracy of the two calculation methods. Based on the results obtained in method 1, table 9 shows the relative errors of the results calculated in method 2.
TABLE 9 analysis of calculation result errors for electro-water coupling systems
The comprehensive comparison shows that the calculation results of the two methods are basically consistent, and the relative errors are in an acceptable range, so that the correctness and the effectiveness of the random optimization method (method 2) based on the neural network provided by the invention are verified.
(2) Calculating a time contrast
Table 10 gives a comparison of the calculation times for the two methods.
Table 10 comparison of calculated time for two methods of electro-water coupling system
| Project | Method 1 | Method 2 |
| Calculation time(s) | 53 | 18 |
As can be seen from table 10, compared with the conventional random optimization method (method 1), the data-driven random optimization method (method 2) provided by the invention has significantly reduced calculation time, which verifies the superiority of the data-driven random optimization method (method 2) provided by the invention. The change of calculation time along with the sampling times of the Monte Carlo in the two random optimization methods is further shown in the figure 7, and it can be seen that the time of the method is not obviously increased along with the increase of the sampling times, but the time required by the traditional method is greatly increased, and compared with the traditional method, the time of the method is reduced by nearly 84%, and the effect is obvious.
Claims (8)
1. The data-driven electric-water combined system optimal energy flow random optimization method is characterized by comprising the following steps of:
1) Establishing an electricity-water combined energy flow model;
2) Calculating the pipeline flow, the node water head and the branch power predicted value of the electric-water combined system by using an electric-water combined energy flow model;
The step of establishing the electricity-water combined energy flow model comprises the following steps:
1.1 Basic data of the electric-water combined system is obtained, monte Carlo sampling is carried out, n groups of load samples, control variables and reservoir flow Q tank are obtained, and a training data set is established;
the load sample comprises water load power P w and electric load power P D; the control variables include generator output P gen;
1.2 Carrying out deterministic combined power flow calculation by using the training data set to obtain the state quantity of the electric-water combined system and taking the state quantity as the label information of the training data; the state quantity comprises pipeline flow, node water head and branch power value;
1.3 Training the deep learning neural network by using a training data set with state quantity, and establishing an electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the state quantity;
the step of calculating the predicted values of the pipeline flow, the node water head and the branch power of the electric-water combined system comprises the following steps:
2.1 Initializing an uncertainty boundary, namely setting an upper uncertainty boundary and a lower uncertainty boundary of pipeline flow, node water head and branch power to be zero, and initializing iteration times k=1;
2.2 Acquiring constraint conditions and network states of the combined electric-water system;
2.3 Taking the current uncertainty boundary as a condition, solving the certainty optimization parameters of the electric-water combined system to obtain a load sample and a control variable value;
2.4 Inputting the load sample and the control variable value into an electric-water combined energy flow model to obtain predicted values of pipeline flow, node water head and branch power;
2.5 Counting probability characteristics of the pipeline flow, the node water head and the branch power predicted value, and correcting the uncertainty boundary according to the probability characteristics obtained by counting and the current uncertainty boundary to obtain a new uncertainty boundary;
the calculation mode of the uncertainty boundary is shown in the formula (3) and the formula (4):
Where λ U represents the uncertainty upper boundary; alpha is a safety margin factor between 0 and 1; epsilon is the quantile under a given security probability; s represents the true value of the state variable; Characterizing a predicted value of a state variable;
the relationship between the true value and the predicted value of the state variable satisfies:
wherein: delta represents the prediction error of the load;
2.6 Judging whether the new uncertainty boundary meets the convergence condition, if so, outputting the predicted values of the pipeline flow, the node water head and the branch power, otherwise, returning to the step 2.3), and enabling the iteration times k=k+1; the convergence condition is that the iteration number k is larger than k max and max (delta f (k))＜ε;kmax is the maximum iteration number; epsilon is convergence accuracy;
Δf(k)＝f(k+1)-f(k) (6)
Wherein: f (k+1) represents the uncertainty boundary value obtained in the (k+1) th iteration, and f (k) represents the uncertainty boundary value obtained in the (k) th iteration; Δf (k) represents the difference between this uncertainty boundary value and the value of the uncertainty boundary of the last iteration.
2. The method for optimizing optimal power flow random of a data-driven combined electric-water system according to claim 1, wherein the combined electric-water power flow model is trained by load samples and control variables.
3. The data-driven electric-water combined system optimal energy flow random optimization method according to claim 1, wherein basic data of the electric-water combined system comprises the number of nodes N Wbus in a water network, the number of water network branches N Wbranch, the number of nodes N bus in a power network, the number of power network branches N branch, the number of generators N gen, the generator output P gen, the water load power P w, the electric load power P D and the reservoir flow Q tank; wherein the water load powerElectric load power P D＝[PD1,PD2,…,PDNbus ], generator output/> The water load power of the N Wbus water network node, the electric load power of the N bus power network node and the output of the N gen generator are respectively shown.
4. The data-driven electric-water joint system optimal power flow random optimization method according to claim 1, wherein the electric-water joint power flow model comprises an electric-water joint power flow model for fitting a relation between a load sample and a control variable and a pipeline flow, an electric-water joint power flow model for fitting a relation between the load sample and the control variable and a node water head, and an electric-water joint power flow model for fitting a relation between the load sample and the control variable and a branch power.
5. The data-driven electric-water combined system optimal energy flow random optimization method according to claim 4, wherein the output of the electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the pipeline flow is the inversely normalized pipeline flow;
The output of the electric-water combined energy flow model for fitting the relation among the load sample, the control variable and the node water head is the inverse normalized node water head;
The output of the electro-water joint energy flow model used to fit the relationship between load samples, control variables and branch power is the inverse normalized branch power.
6. The method for optimizing energy flow random of a data-driven combined electric-water system according to claim 4, wherein the training the deep learning neural network comprises:
1) Preprocessing training data to obtain preprocessed electric load power P Dy, water load power P wy, generator output P geny, reservoir flow Q tank, pipeline flow, node water head and branch power values;
2) Inputting the preprocessed electric load power P Dy, the water load power P wy, the generator output P geny and the reservoir flow Q tank into a deep learning neural network as shown in formulas (3) - (4);
h0＝{PDy;Pwy;Pgeny;Qtank} (1)
Wherein h 0 represents an original input layer of the deep learning neural network; h i represents the output of the ith hidden layer of the model; h i-1 represents the output of the i-1 th hidden layer of the model; w i is the weight of the i-th hidden layer; b i is the bias of the ith hidden layer; Φ () is an activation function;
3) And calculating a loss function of the deep learning neural network, and completing training of the deep learning neural network when the loss function reaches a minimum value and is not changed along with the training times.
7. The method for optimizing energy flow random based on data driven combined electric and water system according to claim 6, wherein the preprocessing comprises normalization.
8. The method for optimizing power flow in a data driven electric-water combined system according to claim 6, wherein the loss function is set as a state quantity predictive valueThe square of the difference from the state quantity true value P branch.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210389338.5A CN115049098B (en) | 2022-04-13 | 2022-04-13 | Data-driven-based optimal energy flow random optimization method for electric-water combined system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210389338.5A CN115049098B (en) | 2022-04-13 | 2022-04-13 | Data-driven-based optimal energy flow random optimization method for electric-water combined system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115049098A CN115049098A (en) | 2022-09-13 |
| CN115049098B true CN115049098B (en) | 2024-06-07 |
Family
ID=83157519
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210389338.5A Active CN115049098B (en) | 2022-04-13 | 2022-04-13 | Data-driven-based optimal energy flow random optimization method for electric-water combined system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115049098B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108734391A (en) * | 2018-05-08 | 2018-11-02 | 重庆大学 | Electric-gas integrated energy system probability energy flow computational methods based on storehouse noise reduction autocoder |
| CN109117951A (en) * | 2018-01-15 | 2019-01-01 | 重庆大学 | Probabilistic Load Flow on-line calculation method based on BP neural network |
| WO2019200662A1 (en) * | 2018-04-20 | 2019-10-24 | 东北大学 | Stability evaluation and static control method for electricity-heat-gas integrated energy system |
| CN110428185A (en) * | 2019-08-08 | 2019-11-08 | 河海大学 | Electric-thermal based on pseudo- measurement model interconnects integrated energy system robust state estimation method |
| CN112531716A (en) * | 2020-11-19 | 2021-03-19 | 重庆大学 | Unified per unit calculation method for mixed energy flow of electricity-water interconnection system |
-
2022
- 2022-04-13 CN CN202210389338.5A patent/CN115049098B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109117951A (en) * | 2018-01-15 | 2019-01-01 | 重庆大学 | Probabilistic Load Flow on-line calculation method based on BP neural network |
| WO2019200662A1 (en) * | 2018-04-20 | 2019-10-24 | 东北大学 | Stability evaluation and static control method for electricity-heat-gas integrated energy system |
| CN108734391A (en) * | 2018-05-08 | 2018-11-02 | 重庆大学 | Electric-gas integrated energy system probability energy flow computational methods based on storehouse noise reduction autocoder |
| CN110428185A (en) * | 2019-08-08 | 2019-11-08 | 河海大学 | Electric-thermal based on pseudo- measurement model interconnects integrated energy system robust state estimation method |
| CN112531716A (en) * | 2020-11-19 | 2021-03-19 | 重庆大学 | Unified per unit calculation method for mixed energy flow of electricity-water interconnection system |
Non-Patent Citations (2)
| Title |
|---|
| Dingyi Liu 等.Data-Driven Optimal Power-water Flow: A linearize Machine Learning Approach.《 2023 IEEE Sustainable Power and Energy Conference (iSPEC)》.2024,第1-7页. * |
| 电-水耦合系统双线性加权最小二乘状态估计;崔镜心 等;《重庆大学学报》;20230509;第1-22页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115049098A (en) | 2022-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Probabilistic charging power forecast of EVCS: Reinforcement learning assisted deep learning approach | |
| CN109214708B (en) | Electric power system risk assessment method based on cross entropy theory optimization support vector machine | |
| Liu et al. | Real-time emergency load shedding for power system transient stability control: A risk-averse deep learning method | |
| CN106022614A (en) | Data mining method of neural network based on nearest neighbor clustering | |
| CN113141008B (en) | Data-driven power distribution network distributed new energy consumption capability assessment method | |
| CN111401757A (en) | Multi-region power load time sequence scene modeling method considering time-space correlation | |
| CN109255726A (en) | A kind of ultra-short term wind power prediction method of Hybrid Intelligent Technology | |
| CN115688579A (en) | Basin multi-point water level prediction early warning method based on generation of countermeasure network | |
| El-Sayed et al. | Fast initialization methods for the nonconvex economic dispatch problem | |
| CN110133447B (en) | Power distribution network fault section positioning method based on complementary relaxation constraint model | |
| CN116306864A (en) | Method for training deep learning model of power system | |
| CN115714382A (en) | Active power distribution network real-time scheduling method and device based on security reinforcement learning | |
| Yi et al. | Intelligent prediction of transmission line project cost based on least squares support vector machine optimized by particle swarm optimization | |
| CN109066651B (en) | Method for calculating limit transmission power of wind power-load scene | |
| Wang | Risk assessment of stochastic spinning reserve of a wind‐integrated multi‐state generating system based on a cross‐entropy method | |
| CN111415068A (en) | Power distribution decision modeling method based on relevance of transformation measures and loss load index | |
| CN109840335A (en) | Based on the radial forging pit prediction optimization method for strengthening T-S fuzzy neural network | |
| Zhang et al. | Planning of electric vehicle charging stations and distribution system with highly renewable penetrations | |
| CN104281978B (en) | A kind of Calculation of Available Transfer Capability method based on Probabilistic Load Flow | |
| CN105529714A (en) | Normal distribution combination characteristic-based rapid probabilistic power flow calculation method | |
| CN115049098B (en) | Data-driven-based optimal energy flow random optimization method for electric-water combined system | |
| Wang et al. | A prediction model for ultra-short-term output power of wind farms based on deep learning | |
| Tang et al. | Traffic flow anomaly detection based on robust ridge regression with particle swarm optimization algorithm | |
| CN114692729A (en) | New energy station bad data identification and correction method based on deep learning | |
| CN110829434B (en) | Method for improving expansibility of deep neural network tidal current model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |