Optimal power flow computation method based on multi-task deep learning

{
  "publication_number": "US11436494B1",
  "country": "US",
  "kind": "B1",
  "title": "Optimal power flow computation method based on multi-task deep learning",
  "abstract": "An optimal power flow computation method based on multi-task deep learning is provided, which is related to the field of smart power grids. The optimal power flow computation method based on multi-task deep learning includes: acquiring state data of a power grid at a certain dispatching moment, and amplifying collected data samples by means of sampling to acquire training data; applying an optimization method to acquire dispatching solutions of the power grid in different sampling states, and acquiring labels; designing a deep learning neural network model, learning feasibility and an optimal solution of an optimal power flow computation problem separately, and outputting a feasibility determination and an optimal solution prediction; simultaneously training, tasks of the feasibility determination and the optimal solution prediction in the optimal power flow computation problem; and determining whether there is a feasible dispatching solution, and outputting an optimal dispatching solution or an early warning.",
  "claims": [
    "1. An optimal power flow computation method based on a multi-task deep learning, comprising: step 1, acquiring state data of a power grid at a certain dispatching moment, and amplifying collected data samples by a sampling method to acquire training data; step 2, applying an optimization method to acquire dispatching solutions of the power grid in different sampling states, and acquiring labels of the training data; step 3, designing a deep learning neural network model, learning a feasibility and an optimal solution of an optimal power flow computation problem separately, and outputting a feasibility determination and an optimal solution prediction; step 4, simultaneously training, based on a multi-task learning framework, tasks of the feasibility determination and the optimal solution prediction in the optimal power flow computation problem to acquire a multi-task deep learning model, which is trained; and step 5, determining whether there is a feasible dispatching solution of the power grid according to an output result of the multi-task deep learning model, and outputting an optimal dispatching solution or an early warning of the power grid; wherein step 1 specifically comprises: in step 1.1, collecting the state data of the power grid at the certain dispatching moment, the state data including the number N bus of nodes, the number N gem of generators, the number N branch of lines, and a reactance x ij and a load power P D =[P d1, Pd 2,... P dNbus] between a node i and a node j in a system, p di, being a load power of a corresponding node i; in step 1.2, performing sampling and amplification on a node load powers p di, of nodes, which are collected, to acquire a sample {circumflex over (P)} D =[{circumflex over (P)} d1, {circumflex over (P)} d2,..., {circumflex over (P)} dNbus] and repeating the sampling and amplification for n times to acquire a training sample P D =[{circumflex over (P)} d1,{circumflex over (P)} d2,...,{circumflex over (P)} dNbus] the sampling method being used for uniform sampling with the node load powers p di, as a center as shown below: {circumflex over (P)} di ∈[(1− r di)× p di,(1+ r di)× p di] wherein r di, is a sampling parameter of the node i; and step 2 specifically comprises: taking the training sample {circumflex over (p)} di, as an input, outputting a corresponding feasibility sign {circumflex over (f)} G1; and a dispatching solution {circumflex over (P)} Gi =[{circumflex over (P)} g1, {circumflex over (P)} g2,..., {circumflex over (P)} gNgen] by means of a traditional optimal power flow solver based on the optimization method; repeating the above process for each of the training sample to acquire corresponding labels {circumflex over (P)} G =[{circumflex over (P)} G1, {circumflex over (P)} G2,..., {circumflex over (P)} Gn] and {circumflex over (F)} G =[{circumflex over (f)} G1, {circumflex over (f)} G2,..., {circumflex over (f)} Gn] of all of the training data; preprocessing the input {circumflex over (P)} D and an output {circumflex over (P)} G, of the training data to acquire a preprocessed input {circumflex over (P)} D and an output {tilde over (P)} G of the training data; inputting the preprocessed input {tilde over (P)} D, into the following deep learning neural network: h 0 ={tilde over (P)} D H i =ϕ(W ihi-1 +b i) wherein h 0 represents an original input of a neural network model, h i represents an output of a i hidden layer of a model, W i; represents a weight of the i hidden layer, b i represents a deviation of the i hidden layer and ϕ(⋅) represents an activation function; designing, aiming at the feasibility of the optimal power flow computation problem, the following output layer for learning: F G =ψ cls (W l cls h i-1 +b l cls)) wherein ψ cls represents the activation function, F g represents a predicted value of a feasibility sign, h i-1, represents an output of a last hidden layer, W l cls represents an output layer weight of a classification problem and b l cls represents an output layer deviation of the classification problem; and designing, aiming at the optimal solution of the optimal power flow computation problem, the following output layer for learning: P G =ψ reg (W l reg h i-1 +b l reg)) wherein ψ reg represents the activation function, P g represents a predicted value of the dispatching solution, h i-1, represents the output of the last hidden layer, W l reg represents an output layer weight of a regression problem and b l reg represents an output layer deviation of the regression problem.",
    "2. The optimal power flow computation method based on the multi-task deep learning according to claim 1, wherein the step 4 specifically comprises: setting a classification task, and measuring a difference between the predicted value F G and a true value {circumflex over (F)} G of the feasibility sign by means of a loss function: loss cls =L cls (F G,{circumflex over (F)} G) wherein L cls represents the loss function of the classification task; setting a regression task, and measuring a difference between the predicted value P G and a true value {tilde over (P)} G of the dispatching solution by means of the loss function: loss reg =L reg (P G,{tilde over (P)} G) wherein L reg represents the loss function of the classification task; acquiring a training loss of the multi-task deep learning model by means of a weighted summation: loss=ω cls ·loss cls +ω reg ·loss reg wherein ω cls and ω reg represent a weight of a classification task loss and a weight of a regression task loss respectively.",
    "3. The optimal power flow computation method based on the multi-task deep learning according to claim 2, wherein the step 5 specifically comprises: for given the state data of the power grid at any dispatching moment, outputting, by the multi-task deep learning model, the predicted value f G of the feasibility sign and a predicted value p G of the optimal solution; determining whether the predicted value p G is effective on the basis of the predicted value f G; if so, taking the predicted value p G of the optimal solution, which is outputted, as the optimal dispatching solution to guide a dispatching operation of the power grid; and otherwise, abandoning the predicted value p G, and outputting the early warning about the fact that the power grid cannot operate normally in a current system state."
  ],
  "description_excerpt": "The disclosure relates to the field of smart power grids and artificial intelligence, in particular to an optimal power flow computation method based on multi-task deep learning.\n\nThe optimal power flow computation method is the core for operation of a power system, and supports various applications of the power system, including economic dispatch, unit commitment, demand response, reliability evaluation, etc. Under the premise of abidance by physical constraints, the power system can achieve optimal generation dispatch by optimal power flow computation. Related to a mathematical programming, optimal power flow computation is diverse, generally including direct current (DC) optimal power flow computation and alternating current (AC) optimal power flow computation. Traditional solution methods are numerical algorithms based on an optimization theory, such as a gradient method, a Newton method and an interior point method. Generally speaking, the traditional methods can be better used for acquiring an optimal solution, and can meet requirements from a general power system. However, due to high time complexity and long computation processes of a numerical algorithm based on the optimization theory, it is difficult to meet the requirements of large-scale power systems, and real-time performance of solution cannot be guaranteed. How to guarantee solution effect of optimal power flow computation and shorten time spent during solution is one of main concerns in the field of smart power grids at present.",
  "cpc": [
    "G06N 3/08",
    "G06F 18/214",
    "G06F 18/241",
    "G06K 9/6256",
    "G06N 3/04",
    "G06N 3/045",
    "G06N 3/048",
    "G06N 3/0499",
    "G06N 3/09",
    "G06N 3/096",
    "H02J 13/00002",
    "H02J 13/12",
    "H02J 2103/30",
    "H02J 2103/35",
    "H02J 3/00",
    "H02J 4/00",
    "Y04S 10/50"
  ],
  "ipc": [
    "G06K 9/62",
    "G06N 3/04",
    "G06N 3/08",
    "H02J 13/00",
    "H02J 4/00"
  ],
  "assignees": [
    "Zhejiang Lab"
  ],
  "inventors": [
    "Gang Huang",
    "Longfei Liao",
    "Wei Hua"
  ],
  "filing_date": "2022-04-10",
  "publication_date": "2022-09-06",
  "grant_date": "2022-09-06",
  "priority_date": "2021-07-21",
  "application_number": "US-202217717121-A",
  "family_id": "77286868",
  "cited_by_count": 11
}