Electricity comes out of the socket! And how does it get there? Most of today's power stations are bad for the climate and the environment. So there is currently a switch to environmentally friendly power generation. New wind turbines are springing up everywhere, but households are increasingly not only consuming electricity, but also producing it: via solar panels and modern heating systems that produce electricity as well as heat. This is good for the environment, but it also creates new problems. electricity is very difficult to store, so at any given moment exactly as much electricity has to be produced as is currently being consumed. This used to be easy because only a few large power plants had to be controlled. But who controls the many small ones, if that is even possible! Solar power is available when the sun shines, wind power when the wind blows. But wait, if generation is difficult to control, can't consumption be adjusted? Yes, it is possible, that's exactly what is being researched. But the devices that consume and generate the electricity have to become smarter!

Many different devices have to co-ordinate and work together. Imagine that for a moment: The robot hoover negotiates with the neighbour's solar system when there is enough electricity for vacuuming, the washing machine waits until the evening to wash because the heating system can supply it with electricity then - in the evening, son needs warm water for his bath anyway. So she also promises the electric car across the road that it can charge itself.

How can you imagine something like that? The devices would then behave like autonomous beings. In this context, we speak of agents. Software ensures that the devices can behave smartly.

But what does that actually mean: software agent?

Scientists define the term agent, for example, as follows:

"An agent is anything that can perceive its environment through sensors and acts in that environment through actuators. [...] A human agent has eyes, ears and other organs as sensors and hands, feet, mouth and other body parts as actuators. A robotic agent could use cameras and infrared range detectors as sensors, [sic] and various motors as actuators. A software agent could accept key codes, file contents and network packets as sensory inputs and act in the environment by displaying something on the screen, printing files or sending network packets."

[RN04, p. 55] (see below) So let's imagine software that is an agent. What must such an agent be able to do?

• An agent perceives its environment, i.e. it observes.
• Those who observe also think about what they see. He draws conclusions.
• An agent can influence its environment. It must therefore pursue a specific goal.
• What if there are several agents in the environment? Then they also communicate with each other.

We have already seen how an agent can perceive its environment and how it influences it. So let's first deal with the question of how a software/agent can reason. To do this, each agent has a set of rules that it has been given by its developer. Since every agent observes its environment, it has a certain amount of knowledge. The agent is therefore able to analyse its model of reality based on past perceptions with regard to what-if questions. It is therefore possible to use the rules to make predictions about the future and the effects of its actions.
If an agent has the choice of what to do, then the question arises as to whether he is doing the right thing. As everyone knows, it is better to do the right thing than the wrong thing. But what does it mean to do the right thing? For an agent, doing the right thing means being as successful as possible. He therefore needs a way to measure his success. To do this, his developer provides him with a function to evaluate his performance. It is difficult to answer the question of what such an evaluation might look like. Imagine the following example: A hoover agent is supposed to vacuum the floor. Now you could measure the performance by looking at how much dirt is removed in 15 minutes. Then the agent could come up with the idea of maximising its own success by emptying out dirt that has already been vacuumed up in order to vacuum it up again. Perhaps penalty points for used electricity should be included in the evaluation. Expert knowledge is often required here. So now we know what drives an agent: Where am I? Where do I want to go? How do I get there?

Simulations with agents are an interesting area of application. Here, scenes from the real world are recreated in the computer with the help of agents. A well-known example is the simulation of fish in a swarm or the behaviour of ants(KinderUniversität). Simulations can be used to make statements about the dynamics of the underlying system. Simulations also have a number of advantages over real experiments:

• Since the agents and the simulation are software, many variants can be tried out in a short time with little effort.
• Interesting experiments can also be repeated as often as required.
• It is also possible to try out scenarios that do not (yet) exist in reality.

Let us now look at two examples of how such agents can contribute to improving the energy supply of the future.

There are many wind turbines in a wind farm and they produce electricity depending on how the wind is blowing. And this is where the problem lies. It often happens that gusts of wind hit individual wind turbines. They then generate more electricity for a moment than the weather forecast predicted. Such unforeseen fluctuations can have a negative impact on the quality of the electricity and must therefore be compensated by the energy supplier above a certain threshold. It would be much better if the windmills could co-ordinate with each other so that some windmills voluntarily slow down a little as soon as some of their colleagues are hit by a gust. This would even out the fluctuations in generation. However, this would require software agents in the windmills to negotiate this equalisation among themselves.

Electricity has been traded on the European Energy Exchange ( EEX) in Leipzig since 2000. It would be great if you could get involved with your solar system on the roof or your CHP (combined heat and power unit) in the basement. Unfortunately, a single small system alone produces far too little electricity. But if many systems work together, they produce enough electricity to be able to sell it on the exchange. To do this, however, they have to co-ordinate with the many electricity consumers and also with each other so that it is clear who can produce how much electricity and when.

In order for this electricity to be sold at a good price, another type of agent is also needed. These special agents do exactly what the stockbrokers on Wall Street do: they exchange offers and counter-offers for electricity with other stockbrokers until everyone has reached an agreement and the excess electricity can be sold at a price that everyone is happy with.

In the future, our devices may do all of this all by themselves and without us even noticing. And incidentally, the agents in the appliances will also automatically recognise how we like it best. The coffee machine will make our breakfast coffee on its own - later on Sundays, of course, because it has learnt that we sleep longer - and always in consultation with the egg cooker or the hoover or something - depending on how much electricity can be used at the time.