Protection device and method for monitoring an electrical energy supply grid

A protection device for monitoring an electrical energy supply grid, the protection device comprising:

• a measured value acquisition device for acquiring measured values which have been recorded at at least one measuring point in the energy supply grid and which indicate an electrical state of the energy supply grid at the at least one measuring point;
• and
• an evaluation device connected to said measured value acquisition device and being formed entirely as a single neural network having an input neuron layer, at least one intermediate neuron layer, and an output neuron layer;
• said evaluation device being configured to carry out a plurality of protection functions on a basis of the measured values and, as a result of the protection functions being carried out, to make a decision as to whether the energy supply grid is in a permissible or impermissible operating state;
• and
• said input neuron layer being configured to receive the measured values and/or values derived therefrom, said output neuron layer being configured to output a classification of operating states of the energy supply grid, and said neural network, for classifying the operating states, being trained to carry out all the protection functions together;
• wherein the protection device is a central protection device configured to receive measured values from sensors at a plurality of measuring points in the energy supply grid and to control switching devices at a plurality of points in the energy supply grid;
• and
• said neural network is trained to classify operating states of the energy supply grid using a one-stage learning method, and said neural network is trained using a deep learning method.

A method for monitoring an electrical energy supply grid, the method comprising:

• acquiring measured values at one or more measuring points in the energy supply grid, the measured values indicating an electrical state of the energy supply grid at the one or more measuring points;
• providing a protection device with an evaluation device being designed entirely in the form of a single neural network having an input neuron layer, at least one intermediate neuron layer, and an output neuron layer;
• transmitting the measured values to the protection device and receiving the measured values and/or values derived therefrom by the input neuron layer;
• carrying out a plurality of protection functions on a basis of the measured values by an evaluation device of the protection device and outputting with the output neuron layer a classification of operating states of the energy supply grid;
• and
• as a result of the protection functions being carried out, making a decision as to whether the energy supply grid is in a permissible operating state or in an impermissible operating state;
• and
• wherein at least one of the intermediate neuron layers is connected to an intermediate output neuron layer and outputting with the intermediate output neuron layer a result specific to a single protection function.

A method for monitoring an electrical energy supply grid, the method comprising:

• acquiring measured values at one or more measuring points in the energy supply grid, the measured values indicating an electrical state of the energy supply grid at the one or more measuring points;
• providing a protection device with an evaluation device being designed entirely in the form of a single neural network having an input neuron layer, at least one intermediate neuron layer, and an output neuron layer;
• transmitting the measured values to the protection device and receiving the measured values and/or values derived therefrom by the input neuron layer;
• carrying out a plurality of protection functions on a basis of the measured values by an evaluation device of the protection device and outputting with the output neuron layer a classification of operating states of the energy supply grid;
• and
• as a result of the protection functions being carried out, making a decision as to whether the energy supply grid is in a permissible operating state or in an impermissible operating state,
• wherein the protection device is configured for local operation at an installation location in a vicinity of the measuring point, and the neural network is trained to classify the operating states of the energy supply grid in a two-stage learning method before starting the monitoring;
• and
• in a first stage of the learning method, training a structure of the neural network from the input neuron layer to an intermediate neuron layer immediately upstream of the output neuron layer using a deep learning method;
• and
• in a second stage of the learning method, training a structure of the neural network from the intermediate neuron layer immediately upstream of the output neuron layer to the output neuron layer using a reinforcement learning method.