There is instability in the distributed energy storage cloud group end region on the power grid side. In order to avoid large-scale fluctuating charging and discharging in the power grid environment and make the capacitor components show a continuous and stable charging and discharging state, a hierarchical time-sharing configuration algorithm of distributed energy storage cloud group end region on the power grid side based on multi-scale and multi feature convolution neural network is proposed. Firstly, a voltage stability analysis model based on multi-scale and multi feature convolution neural network is constructed, and the multi-scale and multi feature convolution neural network is optimized based on Self-Organizing Maps (SOM) algorithm to analyze the voltage stability of the cloud group end region of distributed energy storage on the grid side under the framework of credibility. According to the optimal scheduling objectives and network size, the distributed robust optimal configuration control model is solved under the framework of coordinated optimal scheduling at multiple time scales; Finally, the time series characteristics of regional power grid load and distributed generation are analyzed. According to the regional hierarchical time-sharing configuration model of “cloud”, “group” and “end” layer, the grid side distributed energy storage cloud group end regional hierarchical time-sharing configuration algorithm is realized. The experimental results show that after applying this algorithm, the best grid side distributed energy storage configuration scheme can be determined, and the stability of grid side distributed energy storage cloud group end region layered time-sharing configuration can be improved.

In order to meet the growing demand for electric energy, China’s power construction scale has gradually increased, the power generation capacity has reached the first place in the world, and the installed capacity of power generation is far more than that of many developed countries, promoting the stable development of the national economy [

In order to enable the power grid to reduce operating costs, reduce losses and improve power supply security, many scholars have studied the hierarchical time-sharing configuration algorithm. Bai et al. [

To solve the problems in the above methods, a grid side distributed energy storage cloud group end region hierarchical time-sharing configuration algorithm based on multi-scale and multi feature convolution neural network is proposed. Multiscale and multi feature convolution neural network is a feedforward neural network, which is constructed according to the characteristics of biological vision and perception, including supervised learning and unsupervised learning, and can be applied to the field of recognition and classification. In the grid side distributed energy storage cloud group end regional environment, the energy storage and supply system is responsible for the aggregation and processing of electric energy signals, and can distribute them to lower level application components with the help of transmission channels, which can effectively improve the grid side distributed energy storage cloud group end regional layered time-sharing configuration.

Through the high-dimensional nonlinear mapping relationship between the electrical quantity and the voltage stability state, the voltage stability of the distributed energy storage cloud group end region at the power grid side can be analyzed. The deep structure of the multi-scale and multi feature convolution neural network can obtain more information, which is more intelligent than artificial, and can obtain the key characteristics of the input quantity [

Based on the stability analysis of multi-scale and multi feature convolution neural network, it is necessary to define the appropriate model structure of multi-scale and multi feature convolution neural network, and construct the correct mapping relationship between input and output. Finally, deep learning is carried out according to the input characteristics, so as to analyze the voltage stability of multi-scale and multi feature convolution neural network.

To analyze the voltage stability of multi-scale and multi feature convolution neural network, it is necessary to select the appropriate characteristic quantity as the input sample. Select the corresponding data from the measurement or simulation, input the initial data in the multi-scale and multi feature convolution neural network model, and the multi-scale and multi feature convolution neural network model is shown in

For the input sample matrix, the convolution layer realizes local learning through the convolution kernel of the multi-scale and multi feature convolution neural network model. The convolution calculation equation of the convolution kernel is:

In

The pooling layer uses the maximum pooling method, which is described by

In

The calculation equation of the whole connection layer is:

In

The full connection layer inputs the output value into the output layer, and the output layer uses

In

The convergence of multi-scale and multi feature convolution neural network affects its learning effect. The convergence is related to the initialization weight. SOM algorithm is applied to multi-scale and multi feature convolution neural network to preprocess the initialization weight, that is, preprocess the stability index to optimize the convergence of its multi-scale and multi feature convolution neural network.

SOM neural network includes two layers, specifically input layer and output layer. The input layer inputs samples into the network, and the selection of neurons in the output layer is based on the principle of survival of the fittest, so the input samples include multiple reaction areas.

The input sample of SOM neural network is set as

In

The multi-scale and multi feature convolutional neural network optimized by SOM algorithm cannot analyze the stability of the stable boundary. Therefore, this paper proposes a four element decision structure based on the credibility framework to improve the credibility of the stability analysis of the distributed energy storage cloud group end region layered time-sharing configuration on the power grid side.

After the test samples are processed by

In

Looking up the training samples, we can know the reliability thresholds

As shown in

Based on the above analysis method, the final analysis results of voltage stability in the cloud group end region of distributed energy storage on the power grid side are stability, missed instability, instability and misjudged instability. For the situation of stability and misjudgment of instability, analyze its stability. If the evaluation result is instability or missing judgment of instability, give an alarm in time, and inform the staff to adjust or make decisions on the voltage of the distributed energy storage cloud group end area on the power grid side.

According to the three time scales of day ahead, day in day rolling and real-time, the distributed energy storage grid is divided into three time periods for configuration optimization scheduling. During the day ahead optimization period, the configuration optimization of the distributed energy storage network is carried out by using the day ahead market price, the output of the distributed energy storage network, the interruptible load reduction, etc., to control the output status, load reduction and power purchase of the distributed energy storage network next day, and configure the optimization scheduling framework, as shown in

According to

The distributed energy storage grid has enough time in the previous stage to optimize the configuration and scheduling, so as to provide reference for the distribution of the distributed energy storage grid and the use and production of relevant resources in the next day [

The calculation process of optimal scheduling objective function is shown in

In

(1) Minimum objective function of power grid loss

The calculation process of the minimum objective function of power grid loss is shown in

In

(2) Peak shaving and grain filling objective function

The meaning of peak shaving and valley filling is to make the load curve as stable as possible. Through the calculation of load variance, the stable state of the load curve can be presented, and the minimum value of load variance is taken as the objective result of avoiding peak and filling insufficient sub objective function operation. The objective function is shown in

In

Network scale is a scoped index parameter, which is different from the weighting factor and correlation order. This coefficient index has a poor influence on wireless sensor networks, and the instruction significance derived from it can only be expressed in the shallow structure of the network. However, for the distributed energy storage and supply system, the network scale determines the electrical signal action intensity that the host element must bear in unit time [

With the increase of the coverage area of wireless sensor networks, the calculation results of the network scale will continue to increase. In this process, the energy storage and energy supply intensity of the distributed microgrid will gradually increase, so its ability to control the stability of the charging and discharging behavior of capacitive components is also continuously enhanced.

For the distributed robust optimal configuration control model, this paper adopts the multi-scale and multi feature convolution neural network to solve it. This method decomposes the solved objective function, forms each objective function into two main and sub problems, and solves them alternately, so as to obtain the optimal solution of the objective function. The solving steps are as follows:

Step 1: Given the value of a group of

Step 2: According to the worst scenario determined above, complete the solution of the two main problems and obtain the optimal solution; The new lower bound is described by the value of the objective function of the main problem [

Step 3: According to the optimal solution result, solve the corresponding sub problem, and obtain the objective function value result and the value of

Step 4: Set the convergence threshold of the solution algorithm, represented by

In the actual operation process, the load size of regional power grid and distributed power generation will change correspondingly with time [

An example of regional grid load curve is shown in

An example of distributed power timing curve is shown in

As shown in

The above process completes the analysis of the regional power grid load and the timing characteristics of distributed generation, exposes the key problems of distributed generation access to the power grid, and makes sufficient preparations for the subsequent implementation of the grid side distributed energy storage cloud group end region layered time-sharing configuration algorithm.

Based on the above analysis results of regional power grid load and distributed generation timing characteristics, from the perspective of cloud, group and end, the grid side distributed energy storage cloud group end regional hierarchical time-sharing configuration algorithm is realized, as shown below:

Aiming at maximizing the benefits of distributed power operators, build a regional layered time-sharing configuration model of the “cloud” layer

In

Aiming at minimizing the cost of regional power grid, a “group” layer regional hierarchical time-sharing configuration model

In

In the process of regional power grid configuration planning, after the distributed power supply is connected, it is easy to have “island” phenomenon, resulting in the complexity of

(1)

(2)

(3)

With the goal of maximizing user benefits, build a regional layered time-sharing configuration model of the “end” layer

In

Integrating

In order to test the application effect of the grid side distributed energy storage cloud group end region layered time-sharing configuration algorithm based on multi-scale and multi feature convolution neural network in this paper. PQVIEW is a multi component software system used to establish and analyze power quality and energy measurement databases. A regional power grid is randomly selected for the coordinated configuration experiment. In the experiment, the distributed power is connected to the side distributed energy storage network and placed on different nodes. The distributed power is one wind turbine with a capacity of 50 kW; One small gas turbine with capacity of 90 kW; A battery with a capacity of 250 kW, a rated capacity of 4.5 kW, an AC voltage of 380 V/50 Hz, and a total power of 0~3 kw for DC and AC loads. The relevant parameters of the power supply connected to the distributed energy storage grid are shown in

Serial number | Parameter name | Value |
---|---|---|

1 | Maximum charging and discharging power of mixed energy storage/MW | 5.5 |

2 | Mixed energy storage rated capacity/MW· H | 10 |

3 | Minimum technical output of wind turbine | 0 |

4 | Maximum technical output of wind turbine | 50 |

5 | Minimum technical output of small gas turbine | 0 |

6 | Maximum technical output of small gas turbine | 90 |

7 | Battery | 24 |

In the experiment, the grid side distributed energy storage cloud group end area hierarchical time-sharing configuration needs to analyze the load of each time period in a day, and take the total load calculated by the whole day standard as a reference to adjust the total load at other times. The load active power of the day ahead load and the load time-sharing configuration of the distributed energy storage cloud group end region on the grid side at each time before the day are shown in

According to

Load fluctuation degree/% | Power loss/MWh | Maximum voltage deviation/V |
---|---|---|

10 | 0.23 | 0.044 |

00 | 0.30 | 0.047 |

30 | 0.35 | 0.050 |

40 | 0.30 | 0.053 |

50 | 0.34 | 0.055 |

60 | 0.38 | 0.058 |

70 | 0.37 | 0.060 |

80 | 0.43 | 0.062 |

90 | 0.40 | 0.064 |

100 | 0.41 | 0.066 |

Average value | 0.351 | 0.0559 |

According to the test results in

In order to verify the applicability of the algorithm in this paper, the indirect and direct carbon emission results under different power output during the operation of the grid side distributed energy storage cloud group end region layered time-sharing configuration after the application of the algorithm in this paper are shown in

Output power/MW | Before control/10000 tons | After control/10000 tons | Difference/10000 tons | |||
---|---|---|---|---|---|---|

Indirect carbon emissions | Direct carbon emissions | Indirect carbon emissions | Direct carbon emissions | Indirect carbon emissions | Direct carbon emissions | |

10 | 226.4 | 216.6 | 95.8 | 88.5 | 130.6 | 128.1 |

20 | 229.4 | 223.2 | 103.1 | 89.1 | 126.3 | 134.1 |

30 | 226.5 | 224.9 | 90.4 | 87.5 | 36.1 | 137.4 |

40 | 224.7 | 218.5 | 105.1 | 91.3 | 119.6 | 127.2 |

50 | 229.2 | 221.7 | 103.6 | 94.5 | 125.6 | 127.2 |

Total value | 1136.2 | 1104.9 | 598 | 450.9 | 538.2 | 654 |

According to the test results in

Five kinds of insulation faults are set up in the experiment, including harmonic fault, arc anomaly, switching overvoltage, bus voltage fluctuation, and insulator dampness. The judgment sensitivity of the distributed energy storage cloud group end area layered time-sharing configuration on the power grid side to different fault types before and after using this algorithm is compared. Sensitivity is an indicator to evaluate system faults. The higher the sensitivity is, the earlier the fault is detected.

According to the analysis of

This method, reference [

It can be seen from the experimental results in

The fluctuation trend of the active power of the day ahead load and the load of the distributed energy storage cloud group end area layered time-sharing configuration on the grid side at each time before the day is similar; After the application of this algorithm, the best distributed energy storage configuration scheme on the grid side can be determined, which can ensure the utilization of electric energy, reduce the power loss of the network, and ensure the stability of voltage at the same time; After the algorithm control in this paper, under different output power, the results of indirect and direct carbon emissions decreased significantly, with the highest values of 1,051,000 and 945,000 tons. Therefore, this algorithm has good applicability and can effectively reduce carbon emissions; Compared with before the application of this algorithm, after the application of this algorithm, the grid side distributed energy storage cloud group end region layered time-sharing configuration has higher sensitivity to judge various fault types; It shows that after applying the algorithm in this paper, the fault diagnosis sensitivity of layered time-sharing configuration is high, which can improve the speed of analyzing the stability of distributed energy storage cloud group end region layered time-sharing configuration on the power grid side.

Some achievements have been made in the research of grid side distributed energy storage cloud group end region layered time-sharing configuration, but there are still some problems that have not been discussed in depth, and there are still research values in many aspects, including the following points:

(1) Optimize the scheduling of controllable energy efficiency loads among individual users. With the improvement of residents’ living standards, there will be many new devices for controllable loads, such as electric vehicles, intelligent devices, etc. How to centrally optimize and manage more controllable energy efficiency loads, so as to further reduce power costs and achieve the purpose of peak shaving and valley filling while ensuring users’ normal life and comfort.

(2) The evaluation index for the optimal dispatching of the cloud group end area of distributed energy storage on the power grid side can be reflected by the indoor temperature and water heater temperature difference before and after optimization. The evaluation of user satisfaction is relatively simple. Therefore, how to establish a more comprehensive user satisfaction evaluation index or system is very important and needs further research and discussion.

(3) The grid side distributed energy storage cloud group end area layered time-sharing configuration equipment used is the most widely commercialized electrochemical battery. However, there is no research on electromagnetic energy storage equipment in the future development prospects, such as superconducting and super capacitor energy storage equipment. How to coordinate and optimize the configuration of a variety of energy storage equipment is also a problem worth studying in the future.

The study was supported by

The authors declare that they have no conflicts of interest to report regarding the present study.