CN117875752A - Power system flexible operation domain assessment method based on self-organizing map decision tree - Google Patents
Power system flexible operation domain assessment method based on self-organizing map decision tree Download PDFInfo
- Publication number
- CN117875752A CN117875752A CN202311672051.4A CN202311672051A CN117875752A CN 117875752 A CN117875752 A CN 117875752A CN 202311672051 A CN202311672051 A CN 202311672051A CN 117875752 A CN117875752 A CN 117875752A
- Authority
- CN
- China
- Prior art keywords
- unit
- decision tree
- power system
- self
- cost
- 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
Links
- 238000003066 decision tree Methods 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 34
- 210000002569 neuron Anatomy 0.000 claims description 36
- 230000009194 climbing Effects 0.000 claims description 24
- 238000013138 pruning Methods 0.000 claims description 24
- 239000013598 vector Substances 0.000 claims description 20
- 238000013528 artificial neural network Methods 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 10
- 239000003245 coal Substances 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 230000001537 neural effect Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 230000001965 increasing effect Effects 0.000 claims description 5
- 239000010410 layer Substances 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 238000013145 classification model Methods 0.000 claims description 3
- 230000007423 decrease Effects 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000010248 power generation Methods 0.000 claims description 3
- 238000002360 preparation method Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012887 quadratic function Methods 0.000 claims description 3
- 239000002356 single layer Substances 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000010801 machine learning Methods 0.000 abstract description 3
- 238000004088 simulation Methods 0.000 abstract description 3
- 230000009471 action Effects 0.000 description 3
- 230000001174 ascending effect Effects 0.000 description 2
- 230000002860 competitive effect Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a power system flexible operation domain assessment method based on a self-organizing map decision tree, and belongs to the technical field of power system automation. The invention adopts the machine learning technology to excavate the flexible operation domain of the new energy power system, generates a large amount of training data through random time sequence simulation, utilizes the decision tree algorithm to extract the flexible boundary, and has the interpretation, so that the model shows the flexible operation boundary of the target power system, provides decision support for the scheduling operation of the supporting high-proportion new energy power system, and effectively improves the operation safety level of the power system.
Description
Technical Field
The invention belongs to the technical field of power system automation, and particularly relates to a power system flexible operation domain assessment method based on a self-organizing map decision tree.
Background
The power system flexibility is the ability to respond to predictable or unpredictable power ramp-up events, under the constraint of the power system at its safe and stable operating boundaries, with rapid response to large fluctuations in load and new energy. The prior art mainly relies on a robust optimization technique to solve the maximum uncertainty set of new energy output to define the flexible operation domain of the new energy station. However, for large-scale power transmission networks, the dimension of the optimization model is greatly increased, which presents greater challenges for the convergence and computation speed of the solution algorithm.
Therefore, at present, a power system flexible operation domain assessment method based on an ad hoc mapping decision tree needs to be designed to solve the problems.
Disclosure of Invention
The invention aims to provide a power system flexible operation domain assessment method based on a self-organizing map decision tree, which is used for solving the technical problems in the prior art, a machine learning technology is applied to mine a flexible operation domain of a new energy power system, a large amount of training data is generated through random time sequence simulation, a decision tree algorithm is utilized to extract a flexible boundary, and the decision tree algorithm has an interpretability, so that a model of the decision tree algorithm shows the flexible operation boundary of a target power system and provides decision support for scheduling operation of a high-proportion new energy power system.
In order to achieve the above purpose, the technical scheme of the invention is as follows:
a power system flexible operation domain assessment method based on a self-organizing map decision tree comprises the following steps:
s1, collecting a historical load curve and a wind-solar new energy output curve, and calculating an economic dispatch model to obtain a time sequence operation mode data set;
s2, training and constructing a self-organizing mapping neural network by taking an operation mode of adjacent time points as an input characteristic to obtain a cluster of adjacent time climbing scenes;
step S3, setting a fluctuation range of the wind and solar new energy, generating actual new energy output through random sampling, calculating an adjustable unit output adjustment strategy through optimal power flow, traversing all data samples in each climbing scene cluster, and summarizing to form a training data set;
step S4, training and constructing a corresponding power system flexible operation domain assessment model by utilizing a decision tree algorithm aiming at the cluster of each climbing scene and the corresponding data sample;
and S5, designating a current operation mode of the power system and a prediction operation mode of the next period, judging the cluster by using the self-organizing map neural network as input characteristics, calling a decision tree model corresponding to the cluster, and taking a splitting rule of the decision tree model as a flexible operation boundary of the prediction operation mode.
Further, in step S1, the economic dispatch model is as follows:
s1-1: objective function
The objective function is to minimize the system running cost, including the coal consumption cost caused by power generation and the start-stop cost caused by unit start-stop:
in the method, in the process of the invention,for the coal consumption cost of the unit i, P i,t For the output of the unit i in the period t, +.>For the start-up cost of the unit i>The shutdown cost of the unit i;
s1-2: the coal consumption cost function of the unit can be expressed by a quadratic function of the output:
wherein a is i ,b i ,c i The method comprises the steps of respectively obtaining a quadratic term coefficient, a primary term coefficient and a constant term of the unit cost;
s1-3: system power balance equation constraint
Wherein P is d,t For the load demand of the node d in the period t, N is the number of units and N L The number of nodes is the number of load nodes;
s1-4: hot standby inequality constraint
Wherein u is i,t For the start-stop state of the unit i in the period t, P i,max For the maximum output limit of the unit i, ρ is the hot standby limit;
s1-5: unit output inequality constraint
u i,t P i,min ≤P i ≤u i,t P i,max
Wherein P is i,max ,P i,min The maximum output limit and the minimum output limit of the unit i are respectively;
s1-6: unit climbing inequality constraint
-R d ≤P i,t -P i,t-1 ≤R u
Wherein R is d ,R u Respectively limiting the descending climbing speed and the ascending climbing speed of the unit;
s1-7: unit start-stop time constraint
In the formula, TS, TO are respectively the minimum shutdown time and the minimum startup time of the unit;
s1-8: start-stop expense constraint
Wherein H is i ,J i The single start-up cost and the single shut-down cost of the unit i are respectively,for the start-up cost of the unit i>The shutdown cost of the unit i;
s1-9: tidal current safety inequality constraint
P l,min ≤P l,t ≤P l,max
Wherein P is l,max ,P l,min Is the upper and lower safe limits of the tide of the line l.
Further, in step S2, the basic structure of the self-organizing map neural network is a single-layer network, all neurons are located on a plane, and a specific arrangement connection mode is adopted between the neurons; the input layer is positioned below the plane, each input port is connected with all neurons of the SOM network, and the input port is called a forward connection weight which can be adjusted through iteration; all the neurons are connected on the plane of the neural network, namely the lateral weight connection;
in the self-organizing map neural network, record w ij For the forward weights of j input ports directed to the ith neuron, let w i ＝(w i1 ，w i2 ，…，w in ) T The output of neuron i is in the following manner:
the purpose of using Euclidean distance is to measure the input vector x and the forward weight vector w on the neuron i The smaller the matching degree and the smaller the distance, the x and w i The higher the degree of matching;
the self-organizing map neural network training steps specifically comprise:
s2-1: initializing the weight value, and setting each w ij Assigning a random initial value;
s2-2: inputting a training sample vector x, and enabling the iteration times t=0;
s2-3: calculating Euclidean distance between x and all forward weight vectors;
d i ＝‖w i (t)-x‖
s2-4: the winning neural unit bmu is selected such that
d bmu ＝{d i }
S2-5: adjusting the corresponding forward link weights w on winning neural units bmu The adjustment is performed according to the following formula:
where α (t) is a learning factor, decreases with increasing iteration number, and ranges from (0, 1), N bmu (t) is a neighborhood of winning neuron nodes, the radius decreasing with increasing iteration numberIs small;
s2-6: after the weight connection value is adjusted, let t=t+1, return to step S2-3, continue iterating;
s2-7: repeating the iteration process, and stopping learning the training sample when the preset maximum iteration times are reached;
s2-8: when the network obtains unknown test samples, euclidean distance between the input vector and the forward connection weight vectors on all neurons is calculated, winning neurons are obtained, and sample classification is determined.
Further, in step S4, for each scene, a classification model based on a decision tree is constructed by taking the active power output of the new energy as an input feature and taking a 0/1 tag which meets the flexible climbing requirement as a prediction target; the decision tree is a prediction model for showing decision rules and classification results by a tree data structure, and the key point is that known data which looks disordered and disordered is converted into a tree model capable of predicting unknown data; in the decision tree, each path from the root node to the leaf node represents a rule of decision, wherein each branch node represents a selection among a plurality of alternatives, each leaf node represents a decision, and the essence of the leaf node represents a mapping relation between the object attribute and the object value;
the algorithm training steps of the decision tree are as follows:
step 41, data preparation: preparing a labeled training dataset, each data comprising a plurality of features and corresponding target variables;
step 42, feature selection: selecting a measurement index according to the selected characteristics, and selecting the best characteristics on the current data set as a root node; the selection metric may use information gain or kene purity;
step 43, tree construction: dividing the data set into different subsets according to the selected features; each subset corresponds to a branch; repeating steps 42 and 43 for each subset until a termination condition is met;
step 44, termination condition: setting different termination conditions according to different requirements;
step 45, recursion construction: dividing the subsets in a recursive manner until a termination condition is met;
step 46, leaf node processing: when the termination condition is reached, marking the current node as a leaf node, and setting the class or regression value of the node as the majority class or average value of the samples contained in the node;
step 47, pruning: pruning is carried out on the generated decision tree; pruning is achieved by pre-pruning or post-pruning; pre-pruning refers to stopping splitting operations in advance in the process of building a tree, and post-pruning refers to building a complete decision tree and then reducing the complexity of the tree through pruning operations.
A computer storage medium having stored thereon a computer program which when executed performs a power system flexible run domain assessment method based on an ad hoc mapping decision tree as described above.
Compared with the prior art, the invention has the following beneficial effects:
the method has the advantages that the machine learning technology is applied to mine the flexible operation domain of the new energy power system, a large amount of training data is generated through random time sequence simulation, the decision tree algorithm is utilized to extract the flexible boundary, and the model shows the flexible operation boundary of the target power system due to the interpretability of the decision tree algorithm, so that decision support is provided for supporting the dispatching operation of the high-proportion new energy power system, and the operation safety level of the power system is effectively improved.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Fig. 2 is a flowchart of a SOM network training procedure according to the present invention.
FIG. 3 is a schematic diagram of a decision tree training process according to the present invention.
Detailed Description
For the purpose of making the technical solution and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and examples. It should be understood that the particular embodiments described herein are illustrative only and are not intended to limit the invention, i.e., the embodiments described are merely some, but not all, of the embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present invention. It is noted that relational terms such as "first" and "second", and the like, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions.
Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.
As shown in fig. 1, a power system flexible operation domain assessment method based on a self-organizing map decision tree is provided, which comprises the following steps:
s1: collecting a historical load curve and a wind-solar new energy output curve, and calculating an economic dispatch model to obtain a time sequence operation mode data set;
s2: training and constructing a self-organizing mapping neural network by taking the operation mode of the adjacent time points as input characteristics to obtain a cluster of adjacent time climbing scenes;
s3: setting a fluctuation range of wind and light new energy sources, generating actual new energy source output through random sampling, calculating an adjustable unit output adjustment strategy through optimal power flow, traversing all data samples in each climbing scene cluster, and summarizing to form a training data set;
s4: aiming at the cluster of each climbing scene and the corresponding data sample, training and constructing a corresponding power system flexible operation domain assessment model by utilizing a decision tree algorithm;
s5: when the method is applied, the current operation mode of the power system and the prediction operation mode of the next period are designated as input features, the self-organizing map neural network is utilized to judge the cluster to which the power system belongs, then a decision tree model corresponding to the cluster is called, and the splitting rule of the decision tree model is used as a flexible operation boundary of the designated prediction operation mode.
Further, the step S1 includes: collecting a historical load curve and a wind-solar new energy output curve, and calculating an economic dispatch model to obtain a time sequence operation mode, wherein the time sequence operation mode specifically comprises the following steps of:
the economic dispatch model is as follows:
s1-1: objective function
The objective function is to minimize the system running cost, including the coal consumption cost caused by power generation and the start-stop cost caused by unit start-stop:
in the method, in the process of the invention,for the coal consumption cost of the unit i, P i,t For the output of the unit i in the period t, +.>For the start-up cost of the unit i>The shutdown cost of the unit i.
S1-2: the coal consumption cost function of the unit can be expressed by a quadratic function of the output:
wherein a is i ,b i ,c i The two coefficients are respectively a quadratic term coefficient, a first term coefficient and a constant term of the unit cost.
S1-3: system power balance equation constraint
Wherein P is d,t For the load demand of the node d in the period t, N is the number of units and N L Is the number of load nodes.
S1-4: hot standby inequality constraint
Wherein u is i,t For the start-stop state of the unit i in the period t, P i,max For the maximum output limit of the unit i, ρ is the hot standby limit.
S1-5: unit output inequality constraint
u i,t P i,min ≤P i ≤u i,t P i,max
Wherein P is i,max ,P i,min The maximum output limit and the minimum output limit of the unit i are respectively.
S1-6: unit climbing inequality constraint
-R d ≤P i,t -P i,t-1 ≤R u
Wherein R is d ,R u The descending climbing speed limit and the ascending climbing speed limit of the unit are respectively defined.
S1-7: unit start-stop time constraint
In the formula, TS, TO are respectively the minimum shutdown time and the minimum startup time of the unit.
S1-8: start-stop expense constraint
Wherein H is i ,J i The single start-up cost and the single shut-down cost of the unit i are respectively,for the start-up cost of the unit i>The shutdown cost of the unit i.
S1-9: tidal current safety inequality constraint
P l,min ≤P l,t ≤P l,max
Wherein P is l,max ,P l,min Is the upper and lower safe limits of the tide of the line l.
In the above formula, the variables are defined as follows:
further, the step S2 includes: training and constructing a self-organizing map neural network by taking an operation mode of adjacent time points as an input characteristic to obtain a cluster of adjacent time climbing scenes, wherein the cluster comprises the following concrete steps:
the basic structure of the Self-Organized Map (SOM) is a single-layer network, all neurons are located on a plane, and the neurons can be arranged in a specific arrangement connection manner, such as according to a rectangle or a hexagon. The input layer is below the plane, each input port has connections to all neurons of the SOM network, called forward connection weights, which can be adjusted by iteration. The individual neurons in the plane of the neural network employ full connections, called side-weight connections, so that side-suppression can be constructed in the network and cause competing behavior in the network.
In SOM network, record w ij For the forward weights of j input ports directed to the ith neuron, let w i ＝(w i1 ，w i2 ，…，w in ) T The output of neuron i may be in the following manner:
the purpose of using Euclidean distance is to measure the input vector x and the forward weight vector w on the neuron i The smaller the matching degree and the smaller the distance, the x and w i The higher the degree of matching.
The SOM network stores all characteristic information of input samples in forward weight connection vectors of the network through training and learning processes, after learning is completed, single neurons are only sensitive to a certain type of input samples, the neurons are distributed in a competitive layer, relative positions of the neurons are unchanged in the working stage of the network, under the condition that distances among the input samples are very small, the positions of best matched neurons corresponding to the samples are very close to each other in the competitive layer, the mode vectors stored on the neurons form relatively fixed clusters, and one cluster usually does not correspond to one neuron. The clusters in the SOM network have adjacent relations, and the adjacent clusters have higher correlation than non-adjacent clusters, so that the interpretation and visualization of the clustering result are facilitated. However, the SOM itself has a certain limitation, firstly, the SOM needs to select network parameters in advance, and lacks a specific objective function in the network learning process, generally sets a maximum iteration number of training, and in addition, the SOM network cannot guarantee convergence due to the random initialization weight.
As shown in fig. 2, the SOM network training step specifically includes:
s2-1: initializing the weight value, and setting each w ij Assigning a smaller random initial value;
s2-2: inputting a training sample vector x, and enabling the iteration times t=0;
s2-3: calculating Euclidean distance between x and all forward weight vectors;
d i ＝‖w i (t)-x‖
s2-4: the winning neural unit bmu (best matching unit) is selected such that
d bmu ＝{d i }
S2-5: adjusting the corresponding forward link weights w on winning neural units bmu The adjustment is performed according to the following formula:
where α (t) is a learning factor, decreases with increasing iteration number, and ranges from (0, 1), N bmu (t) is a neighborhood of winning neuron nodes, the radius decreasing as the number of iterations increases;
s2-6: after the weight connection value is adjusted, let t=t+1, return to step S2-3, continue iterating;
s2-7: and repeating the iteration process, and stopping learning the training sample when the preset maximum iteration times are reached.
S2-8: when the network obtains unknown test samples, euclidean distance between the input vector and the forward connection weight vectors on all neurons is calculated, winning neurons are obtained, and sample classification is determined.
Further, the step S3 includes: setting a fluctuation range of wind and light new energy sources, generating actual new energy source output through random sampling, calculating an adjustable unit output adjustment strategy through optimal power flow, traversing all data samples in each climbing scene cluster, and summarizing to form a training data set.
Further, the step S4 includes: aiming at the cluster of each climbing scene and the corresponding data sample, a corresponding power system flexible operation domain evaluation model is built by utilizing the decision tree algorithm training, specifically:
aiming at each scene, a classification model based on a decision tree is constructed by taking the active power output of the new energy as an input characteristic and taking a 0/1 label which meets the flexible climbing requirement as a prediction target. The Decision Tree (Decision Tree) is a prediction model for showing Decision rules and classification results in a Tree data structure, and the key point is to convert known data which appears unordered and disordered into a Tree model capable of predicting unknown data. In the decision tree, each path from the root node (attribute that contributes most to the final classification result) to the leaf node (final classification result) represents a rule for a decision, where each branch node represents a choice between multiple alternatives, and each leaf node represents a decision, which essentially represents a mapping between an object attribute and an object value. As a supervised learning algorithm, decision trees are mainly used for classification problems, are suitable for classification and continuous input and output of variables, are one of the most widely used and practical methods for inductive reasoning, and can learn and train data from given examples and predict invisible situations.
The algorithm training flow chart of the decision tree is shown in fig. 3, and the specific steps are as follows:
step 1: data preparation: a labeled training dataset is prepared and each data contains its multiple features and corresponding target variables (class or continuous values).
Step 2: feature selection: and selecting the best feature on the current data set as the root node according to the selected feature selection metric. The selection metric may use information gain, kene purity, etc.
Step 3: building a tree: the data set is divided into different subsets according to the selected features. One for each subset. For each subset, steps 2 and 3 are repeated until a termination condition is met.
Step 4: termination condition: different termination conditions may be set according to different requirements. Common termination conditions are: the maximum depth is reached, the number of samples on the node is less than a certain threshold, the samples on the node belong to the same class, etc.
Step 5: and (3) recursion construction: the sub-set is further partitioned in a recursive manner until a termination condition is met.
Step 6: leaf node processing: when the termination condition is reached, the current node is marked as a leaf node and the class or regression value of that node is set to the majority class or average of the samples contained by that node.
Step 7: pruning: to avoid overfitting, pruning operations may be performed on the generated decision tree. Pruning may be achieved by pre-pruning or post-pruning. Pre-pruning refers to stopping splitting operations in advance in the process of building a tree, and post-pruning refers to building a complete decision tree first and then reducing the complexity of the tree through pruning operations.
According to the above steps, a complete decision tree model can be finally generated, as shown in fig. 3.
Further, the step S5 includes: when the method is applied, the current operation mode of the power system and the prediction operation mode of the next period are designated as input features, the self-organizing map neural network is utilized to judge the cluster to which the power system belongs, then a decision tree model corresponding to the cluster is called, and the splitting rule of the decision tree model is used as a flexible operation boundary of the designated prediction operation mode.
The above is a preferred embodiment of the present invention, and all changes made according to the technical solution of the present invention belong to the protection scope of the present invention when the generated functional effects do not exceed the scope of the technical solution of the present invention.
Claims (5)
1. The power system flexible operation domain assessment method based on the self-organizing map decision tree is characterized by comprising the following steps of:
s1, collecting a historical load curve and a wind-solar new energy output curve, and calculating an economic dispatch model to obtain a time sequence operation mode data set;
s2, training and constructing a self-organizing mapping neural network by taking an operation mode of adjacent time points as an input characteristic to obtain a cluster of adjacent time climbing scenes;
step S3, setting a fluctuation range of the wind and solar new energy, generating actual new energy output through random sampling, calculating an adjustable unit output adjustment strategy through optimal power flow, traversing all data samples in each climbing scene cluster, and summarizing to form a training data set;
step S4, training and constructing a corresponding power system flexible operation domain assessment model by utilizing a decision tree algorithm aiming at the cluster of each climbing scene and the corresponding data sample;
and S5, designating a current operation mode of the power system and a prediction operation mode of the next period, judging the cluster by using the self-organizing map neural network as input characteristics, calling a decision tree model corresponding to the cluster, and taking a splitting rule of the decision tree model as a flexible operation boundary of the prediction operation mode.
2. The power system flexible operation domain assessment method based on self-organizing map decision tree according to claim 1, wherein in step S1, an economic dispatch model is as follows:
s1-1: objective function
The objective function is to minimize the system running cost, including the coal consumption cost caused by power generation and the start-stop cost caused by unit start-stop:
in the method, in the process of the invention,for the coal consumption cost of the unit i, P i,t For the output of the unit i in the period t, +.>For the start-up cost of the unit i>The shutdown cost of the unit i;
s1-2: the coal consumption cost function of the unit can be expressed by a quadratic function of the output:
wherein a is i ,b i ,c i The method comprises the steps of respectively obtaining a quadratic term coefficient, a primary term coefficient and a constant term of the unit cost;
s1-3: system power balance equation constraint
Wherein P is d,t For the load demand of the node d in the period t, N is the number of units and N L The number of nodes is the number of load nodes;
s1-4: hot standby inequality constraint
Wherein u is i,t For the start-stop state of the unit i in the period t, P i,max For the maximum output limit of the unit i, ρ is the hot standby limit;
s1-5: unit output inequality constraint
u i,t P i,min ≤P i ≤u i,t P i,max
Wherein P is i,max ,P i,min The maximum output limit and the minimum output limit of the unit i are respectively;
s1-6: unit climbing inequality constraint
-R d ≤P i,t -P i,t-1 ≤R u
Wherein R is d ,R u Respectively the lower parts of the unitsA hill climbing rate limit and an uphill climbing rate limit;
s1-7: unit start-stop time constraint
In the formula, TS, TO are respectively the minimum shutdown time and the minimum startup time of the unit;
s1-8: start-stop expense constraint
Wherein H is i ,J i The single start-up cost and the single shut-down cost of the unit i are respectively,for the start-up cost of the unit i>The shutdown cost of the unit i;
s1-9: tidal current safety inequality constraint
P l,min ≤P l,t ≤P l,max
Wherein P is l,max ,P l,min Is the upper and lower safe limits of the tide of the line l.
3. The flexible operation domain evaluation method of electric power system based on self-organizing map decision tree according to claim 2, wherein in step S2, the basic structure of self-organizing map neural network is a single-layer network, all neurons are located on a plane, and a specific arrangement connection mode is adopted between neurons; the input layer is positioned below the plane, each input port is connected with all neurons of the SOM network, and the input port is called a forward connection weight which can be adjusted through iteration; all the neurons are connected on the plane of the neural network, namely the lateral weight connection;
in the self-organizing map neural network, record w ij For the forward weights of j input ports directed to the ith neuron, let w i ＝(w i1 ,w i2 ,…,w in ) T The output of neuron i is in the following manner:
the purpose of using Euclidean distance is to measure the input vector x and the forward weight vector w on the neuron i The smaller the matching degree and the smaller the distance, the x and w i The higher the degree of matching;
the self-organizing map neural network training steps specifically comprise:
s2-1: initializing the weight value, and setting each w ij Assigning a random initial value;
s2-2: inputting a training sample vector x, and enabling the iteration times t=0;
s2-3: calculating Euclidean distance between x and all forward weight vectors;
d i ＝‖w i (t)-x‖
s2-4: the winning neural unit bmu is selected such that
d bmu ＝{d i }
S2-5: adjusting the corresponding forward link weights w on winning neural units bmu The adjustment is performed according to the following formula:
where α (t) is a learning factor, decreases with increasing iteration number, and ranges from (0, 1), N bmu (t) is a neighborhood of winning neuron nodes, the radius decreasing as the number of iterations increases;
s2-6: after the weight connection value is adjusted, let t=t+1, return to step S2-3, continue iterating;
s2-7: repeating the iteration process, and stopping learning the training sample when the preset maximum iteration times are reached;
s2-8: when the network obtains unknown test samples, euclidean distance between the input vector and the forward connection weight vectors on all neurons is calculated, winning neurons are obtained, and sample classification is determined.
4. The power system flexible operation domain assessment method based on the self-organizing map decision tree according to claim 3, wherein in step S4, for each scene class, a classification model based on the decision tree is constructed by taking the active power output of the new energy as an input characteristic and taking a 0/1 label which meets the flexible climbing requirement as a prediction target; the decision tree is a prediction model for showing decision rules and classification results by a tree data structure, and the key point is that known data which looks disordered and disordered is converted into a tree model capable of predicting unknown data; in the decision tree, each path from the root node to the leaf node represents a rule of decision, wherein each branch node represents a selection among a plurality of alternatives, each leaf node represents a decision, and the essence of the leaf node represents a mapping relation between the object attribute and the object value;
the algorithm training steps of the decision tree are as follows:
step 41, data preparation: preparing a labeled training dataset, each data comprising a plurality of features and corresponding target variables;
step 42, feature selection: selecting a measurement index according to the selected characteristics, and selecting the best characteristics on the current data set as a root node; the selection metric may use information gain or kene purity;
step 43, tree construction: dividing the data set into different subsets according to the selected features; each subset corresponds to a branch; repeating steps 42 and 43 for each subset until a termination condition is met;
step 44, termination condition: setting different termination conditions according to different requirements;
step 45, recursion construction: dividing the subsets in a recursive manner until a termination condition is met;
step 46, leaf node processing: when the termination condition is reached, marking the current node as a leaf node, and setting the class or regression value of the node as the majority class or average value of the samples contained in the node;
step 47, pruning: pruning is carried out on the generated decision tree; pruning is achieved by pre-pruning or post-pruning; pre-pruning refers to stopping splitting operations in advance in the process of building a tree, and post-pruning refers to building a complete decision tree and then reducing the complexity of the tree through pruning operations.
5. A computer storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed, performs the power system flexible operation domain assessment method based on the self-organizing map decision tree according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311672051.4A CN117875752A (en) | 2023-12-06 | 2023-12-06 | Power system flexible operation domain assessment method based on self-organizing map decision tree |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311672051.4A CN117875752A (en) | 2023-12-06 | 2023-12-06 | Power system flexible operation domain assessment method based on self-organizing map decision tree |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117875752A true CN117875752A (en) | 2024-04-12 |
Family
ID=90593632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311672051.4A Pending CN117875752A (en) | 2023-12-06 | 2023-12-06 | Power system flexible operation domain assessment method based on self-organizing map decision tree |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117875752A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118094398A (en) * | 2024-04-26 | 2024-05-28 | 深圳市云之声科技有限公司 | Power supply evaluation method based on Internet of things |
-
2023
- 2023-12-06 CN CN202311672051.4A patent/CN117875752A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118094398A (en) * | 2024-04-26 | 2024-05-28 | 深圳市云之声科技有限公司 | Power supply evaluation method based on Internet of things |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230080737A1 (en) | Federated Learning-Based Regional Photovoltaic Power Probabilistic Forecasting Method and Coordinated Control System | |
| JP5888640B2 (en) | Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program | |
| CN109886473B (en) | Watershed wind-solar water system multi-objective optimization scheduling method considering downstream ecology | |
| Uzlu et al. | Estimates of hydroelectric generation using neural networks with the artificial bee colony algorithm for Turkey | |
| CN111144663B (en) | Ultra-short-term wind power prediction method for offshore wind farm considering output fluctuation process | |
| CN106251001A (en) | A kind of based on the photovoltaic power Forecasting Methodology improving fuzzy clustering algorithm | |
| CN114792156B (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN102270309A (en) | Short-term electric load prediction method based on ensemble learning | |
| CN104636801A (en) | Transmission line audible noise prediction method based on BP neural network optimization | |
| CN105447509A (en) | Short-term power prediction method for photovoltaic power generation system | |
| CN117875752A (en) | Power system flexible operation domain assessment method based on self-organizing map decision tree | |
| Wani et al. | Cluster based approach for mining patterns to predict wind speed | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| CN116341717A (en) | Wind speed prediction method based on error compensation | |
| CN116826710A (en) | Peak clipping strategy recommendation method and device based on load prediction and storage medium | |
| CN109886488B (en) | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag | |
| Huang et al. | Short-term load forecasting based on a hybrid neural network and phase space reconstruction | |
| CN108694475B (en) | Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model | |
| Shen et al. | Short-term load forecasting of power system based on similar day method and PSO-DBN | |
| CN117196019B (en) | New Anjiang model parameter calibration method based on improved self-adaptive genetic algorithm | |
| CN116559975A (en) | Multi-step long weather prediction method based on multi-element time sequence diagram neural network | |
| CN110059871A (en) | Photovoltaic power generation power prediction method | |
| Yang et al. | Multistep wind speed forecasting using a novel model hybridizing singular spectrum analysis, modified intelligent optimization, and rolling elman neural network | |
| CN116089847B (en) | Distributed adjustable resource clustering method based on covariance agent | |
| CN116632842B (en) | Clustering characteristic-based method and system for predicting distribution type photovoltaic load probability of platform |
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 |