Table 1.

coordinated pair A ¼ ð Þ X; μA : Set X will be the universe of fuzzy set A, and μA will be the membership function of fuzzy set A. For each x∈X,real number μA xð Þ is the level or degree of the membership of element x in fuzzy set A; μA xð Þ will be

Maximisation of relation μð Þ g Xð Þ by applying the function of triangular fuzzy number zero [8].

μA xð Þ∈ð Þ 0; 1 It is not possible to identify whether x is a member of A,

In our case, g(X) expresses the deviation of the stable output balance which is why we seek to set it to zero. Number x∈X is selected arbitrarily from fuzzy set A, and μg is a function of fuzzy set A (where the admissible deviation is defined). It is obvious that for each x∈X, real number μ (g (X)) may be called the membership

We describe and compare the expressions x � g Xð Þ and μA Xð Þ� μð Þ� g Xð Þ . Then,

μð Þ¼ g Xð Þ ð Þ ΔP � j g Xð Þj =ΔP in case when

Expressions (33) and (34) and Figure 8 allow us to assume that ΔP¼0, by which the constraining condition is fulfilled (see μðð Þ¼ g Xð Þ 1). Eq. (32) is optimised by its subsequent minimisation or maximisation (μ¼0 ǎz 1). Fuzzy number "μ" will always be small, and we may achieve that using number a (therefore we maximise function f(X)). At this moment, we may say that we have solved the optimisation of our task for the purposes of other applications, for example, in order to minimise

Let us assume the fictitious smart city (intelligent area of Rohan Island) which consists of a complex of intelligent residential, administrative and public buildings

The energy concept of the area under consideration is clearly focused on local renewable energy sources (FV1, FV2, FV3, FV4 and FV5) assembled together with biomass and cogeneration systems (Figure 2), including TS-DS 22/0.4 kV power station, RMS, (Figure 2), located in the underground floor of KU02. This is a RES

while the value of μA xð Þ expresses the level, degree of the

g Xð Þ∈h i �ΔP;ΔP (33)

μð Þ¼ g Xð Þ 0 in case when g Xð Þ∉ h i �ΔP;ΔP (34)

μA xð Þ¼ 0 Element x is not a member of set A. μA xð Þ¼ 1 Element x is a member of set A.

membership of x in A.

interpreted the following way:

Zero and Net Zero Energy

Figure 8.

level or degree of element x in fuzzy set A.

μð Þ g Xð Þ can be expressed as follows:

the special-purpose function.

with a wide range of civic amenities (Figure 1).

3. Experiment

76

RES parameters in micro-networks (local RES) [8].


"Note: Wp [kWh / m<sup>2</sup> ] is the specific electricity consumption per floor area in m<sup>2</sup> , Wp, year [kWh / m<sup>2</sup> ], Wsp [kWh] is electricity consumption per year, PPV [kWp] is photovoltaic power."

#### Table 2.

Building types and their specific and total consumption.


#### Table 3.

Building types and their respective electricity generation per year [8].

electricity generation per year per type of facility including total electricity generation per year.

#### 3.1 Self-organising map

The aim of our experiment is to define a design solution for the system of sorting resources for randomly selected working days Monday and Thursday. Using mathematical analysis using the optimisation stochastic method—simulated annealing—we reached the design and evaluation of input parameters for the purpose of designing the software for the application of the programming language JAVA.

> We also apply the self-organising neural network (SOM)—the Kohonen network. The Kohonen network works analogously as a cluster or factor analysis. The aim is to reduce the input file by mapping it to a smaller number of clusters. Thus, we can imagine finding the spatial representation of complex data structures so that classes of similar vectors are defined by close neurons in each topology. After the network adaptation, the Kohonen map (Figure 11) is drawn out during active

Typical daily electricity consumption diagram and its standard (Monday).

Optimising Energy Systems in Smart Urban Areas DOI: http://dx.doi.org/10.5772/intechopen.85342

Typical daily electricity consumption diagram and its standard (Thursday).

Figure 9.

Figure 10.

Figure 11.

79

The Kohonen map of our experiment (cluster analysis).

The input parameters for the optimisation programme are hourly load prediction (obtained from the history of experimental scientific observation)—what will be the power consumption at a given time? To evaluate this data, a neural network was used to transmit and process information (data). The neural network was also used to implement and optimise the parameters and structure of the fuzzy model. In addition, the clustering method—cloud analysis of data—was used through data analysis. Several types of daily diagrams were created, and then grouped into "clusters", so that two objects of the same cluster were like two objects from different clusters. The result of the individual clusters was the so-called prototype. Prototypes, cost factors and constraints were input into the neural network, the number of power generators (sources), the number of hours we are functioning on, the cost factors for each generator, the predicted consumption for each hour of the time period and the weight w. A cluster analysis method was applied, and the annual history of electricity consumption has been artificially modelled to compare identified daytime patterns with a standard. The baseline standard used hourly patterns of consumption of the working days on Monday and Thursday in January 2019, where each hourly consumption was randomly modified using a random number generator with a normal probability distribution. This modelling was performed 260 times (the total number of Monday working days) and 260 times (the total number of Thursday working days) through a JAVA programme. In Figure 9, two examples of randomly modelled Monday daytime charts are selected, and two examples of randomly modelled daily charts are selected on Thursday. These are hourly consumption forecasts, that is, its standards derived from historical data Figure 10.

Optimising Energy Systems in Smart Urban Areas DOI: http://dx.doi.org/10.5772/intechopen.85342

Figure 9. Typical daily electricity consumption diagram and its standard (Monday).

Figure 10. Typical daily electricity consumption diagram and its standard (Thursday).

We also apply the self-organising neural network (SOM)—the Kohonen network. The Kohonen network works analogously as a cluster or factor analysis. The aim is to reduce the input file by mapping it to a smaller number of clusters. Thus, we can imagine finding the spatial representation of complex data structures so that classes of similar vectors are defined by close neurons in each topology. After the network adaptation, the Kohonen map (Figure 11) is drawn out during active

Figure 11. The Kohonen map of our experiment (cluster analysis).

electricity generation per year per type of facility including total electricity genera-

Building types and their respective electricity generation per year [8].

The aim of our experiment is to define a design solution for the system of sorting resources for randomly selected working days Monday and Thursday. Using

The input parameters for the optimisation programme are hourly load prediction (obtained from the history of experimental scientific observation)—what will be the power consumption at a given time? To evaluate this data, a neural network was used to transmit and process information (data). The neural network was also used to implement and optimise the parameters and structure of the fuzzy model. In addition, the clustering method—cloud analysis of data—was used through data analysis. Several types of daily diagrams were created, and then grouped into "clusters", so that two objects of the same cluster were like two objects from different clusters. The result of the individual clusters was the so-called prototype. Prototypes, cost factors and constraints were input into the neural network, the number of power generators (sources), the number of hours we are functioning on, the cost factors for each generator, the predicted consumption for each hour of the time period and the weight w. A cluster analysis method was applied, and the annual history of electricity consumption has been artificially modelled to compare identified daytime patterns with a standard. The baseline standard used hourly patterns of consumption of the working days on Monday and Thursday in January 2019, where each hourly consumption was randomly modified using a random number generator with a normal probability distribution. This modelling was performed 260 times (the total number of Monday working days) and 260 times (the total number of Thursday working days) through a JAVA programme. In Figure 9, two examples of randomly modelled Monday daytime charts are selected, and two examples of randomly modelled daily charts are selected on Thursday. These are hourly consumption forecasts, that is, its standards derived from

mathematical analysis using the optimisation stochastic method—simulated annealing—we reached the design and evaluation of input parameters for the purpose of designing the software for the application of the programming

tion per year.

Table 3.

language JAVA.

3.1 Self-organising map

Zero and Net Zero Energy

historical data Figure 10.

78

#### Zero and Net Zero Energy

dynamics, after resubmitting the training patterns, from which we can find a very well-defined massive cluster corresponding to Mondays and Thursdays.

Furthermore, by spreading propagation or active dynamics, we can extract the weight vectors from the configuration of the learned neural network, that is, searched day-type diagrams (Figures 12 and 13), where they compare with the appropriate standard.

The individual daily charts in the annual history in Figures 12 and 13 show that consumption patterns are quite different. Typical daily consumption patterns are basically like the relevant standards (Figures 12 and 13), as illustrated by the fact that the cluster analysis method is very effective.

Table 4 shows that the average and maximum tolerances range from approximately 0.1% to 0.5%. From this expression we can evaluate that the cluster analysis method is a very effective and high-quality method demonstrated by this experiment.

If we are to evaluate our experiment according to our specifications, we will assume a situation when we supply power to our smart area of "Rohanský ostrov" through the RES microgrid. The RES microgrid is equipped by eight power generators complemented with the low-voltage grid supplies and the installation of ACCUs. This is a combination of the following ways of power generation:

photovoltaics, cogeneration and biomass plus low-voltage supplies from the distri-

Another task and therefore the aim of the experiment was to design a unit commitment for the weekdays, Monday and Thursday, in January 2019. The hourly consumption forecast has been processed for Thursday (for this chapter we do not specify the hourly consumption of Monday's working day in terms of content)

Organisation system for power energy sources of the RES microgrid for the working day Thursday.

Initial operation of the temperature setting is based on its initial estimate and its subsequent increase to a value at which almost every failure is accepted during the first 10%. The principle of tuning the number of iterations is based on its initial estimate and subsequent increase to a value that does not reduce the resulting production cost to the amount of energy that covers the consumption of that period. The reference cost of electricity generation that covered the estimated consumption of the period was defined as a simplified solution. A simplified solution for that period consisted in the fact that all resources work at medium strength (see rela-

bution grid (see Table 5).

Typical daily diagram (TDD) compared to standard.

Optimising Energy Systems in Smart Urban Areas DOI: http://dx.doi.org/10.5772/intechopen.85342

(Table 5).

Table 4.

tionship (35)).

Table 5.

81

#### Figure 12.

Typical daily diagram working day Thursday.

Figure 13. Standard daily diagram working day Thursday.

dynamics, after resubmitting the training patterns, from which we can find a very

Furthermore, by spreading propagation or active dynamics, we can extract the

The individual daily charts in the annual history in Figures 12 and 13 show that consumption patterns are quite different. Typical daily consumption patterns are basically like the relevant standards (Figures 12 and 13), as illustrated by the fact

Table 4 shows that the average and maximum tolerances range from approximately 0.1% to 0.5%. From this expression we can evaluate that the cluster analysis method is a very effective and high-quality method demonstrated by this experiment. If we are to evaluate our experiment according to our specifications, we will assume a situation when we supply power to our smart area of "Rohanský ostrov" through the RES microgrid. The RES microgrid is equipped by eight power generators complemented with the low-voltage grid supplies and the installation of ACCUs. This is a combination of the following ways of power generation:

well-defined massive cluster corresponding to Mondays and Thursdays.

that the cluster analysis method is very effective.

appropriate standard.

Zero and Net Zero Energy

Figure 12.

Figure 13.

80

Typical daily diagram working day Thursday.

Standard daily diagram working day Thursday.

weight vectors from the configuration of the learned neural network, that is, searched day-type diagrams (Figures 12 and 13), where they compare with the


Table 4. Typical daily diagram (TDD) compared to standard.

photovoltaics, cogeneration and biomass plus low-voltage supplies from the distribution grid (see Table 5).

Another task and therefore the aim of the experiment was to design a unit commitment for the weekdays, Monday and Thursday, in January 2019. The hourly consumption forecast has been processed for Thursday (for this chapter we do not specify the hourly consumption of Monday's working day in terms of content) (Table 5).

Initial operation of the temperature setting is based on its initial estimate and its subsequent increase to a value at which almost every failure is accepted during the first 10%. The principle of tuning the number of iterations is based on its initial estimate and subsequent increase to a value that does not reduce the resulting production cost to the amount of energy that covers the consumption of that period. The reference cost of electricity generation that covered the estimated consumption of the period was defined as a simplified solution. A simplified solution for that period consisted in the fact that all resources work at medium strength (see relationship (35)).


#### Table 5. Organisation system for power energy sources of the RES microgrid for the working day Thursday.

$$P\_i(t) = \mathcal{C}(t) \frac{P\_i^C}{\sum\_i P\_i^C} P\_i^C = \frac{1}{2} \left( P\_i^{\max} - P\_i^{\min} \right) \tag{35}$$

our experiment, 4.2 MWh/year of surplus power would be 85%, which is 3.6 MWh/year. The intelligent urban area would be self-sufficient in terms of electricity consumption and would also generate 3.6 MWh of electricity per year into the 22-kV power grid. The smart area would be energy-efficient in this case, and 85% of the total volume of electricity produced would be commercial. With the transition to smart grids (Figure 2), it is assumed that the intelligent urban development of the Rohan embankment will behave like a power producer and be able to influence the energy market. Similarly to the today's use of automated exchange system to offset exchange rate differences, a decentralised network of autonomous buildings

When defining the unit commitment optimisation from RES by working day (Thursday) in 1-hour increments, we have achieved a further saving of

Department of Microenvironmental and Building Services Engineering,

© 2019 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium,

Czech Technical University in Prague, Prague, Czech Republic

\*Address all correspondence to: bohumir.garlik@fsv.cvut.cz

provided the original work is properly cited.

—power stations—can be created on the energy market.

Optimising Energy Systems in Smart Urban Areas DOI: http://dx.doi.org/10.5772/intechopen.85342

approximately 20%.

Author details

Bohumír Garlík

83

During the course of optimisation, we perform random settings of the state and output of a randomly selected RES generator within the microgrid using a random number generator. This defined procedure is done for each hour and each set period as well as for every iteration. This is done using a programme in JAVA source code.

```
// start hour cycle
for (int j = 2; j <= nt +1; j ++) {
     // start iteration cycle
     for (iter = 1; iter <= n; iter ++) {
              // state random generation
              // random choice of source
              i = ran (seed) * (ng - 1) +1;
              ij = (i-1) * (nt +1) + j;
              // random change of state
              if (x (ij) == 0) {
                         if (ran (seed) <= Ponoff) x (ij) = 1;
              Else
                         if (ran (seed) <= Ponoff) x (ij) = 0;
              // random choice of source
              i = ran (seed) * (ng - 1) +1;
              ij = (i-1) * (nt +1) + j;
              // random set of power
              p (ij) = rand (seed) * (Pmax (i) - Pmin (i)) + Pmin (i);
}
}
```
nt is the number of hours and ng is the number of sources. p(ij) the output of the ith generator and the jth hour. Further, x(ij) is the state of the ith source during the jth hour. Pmin and Pmax are power values with their limit in the ith generator. Here, nt is the number of hours, ng is the number of available sources, p (ij) is the power of the ith source and the jth hour, and x (ij) is the state of the ith source during the jth hour. Values Pmin and Pmax have their limit in the ith source. Ponoff is the parametrizable probability of a change of the source state. Function ran is set up by the random number generation within the interval of (0,1) with even probability distribution. The result of our experiment in terms of source organisation on the selected Thursday as defined by us is presented in Table 5. Worth mentioning is also the fact that the calculation time when using a laptop was 2 minutes and 30 seconds.

### 4. Conclusion

The total power load (consumption) of the intelligent urban area "Rohanské nábřeží" (Rohan Island) according to Table 2 is estimated at 21,000,757 kWh/year = 21 MWh/year. Total power generation from RES microgrid (Table 3) is 7,801,559 kWh/year = 7.8 MWh/year. At present β ¼ 0:6 of the total power consumption of the smart area which is 12,600,454 kWh/year. The installed distributed micro-network of RES will cover the power consumption of the urban smart area with 62% of electricity. The projected planned concept (ideal idea) is to have by 2020 a factor of 0.2, thus existing distribution rates can be optimised. In

#### Optimising Energy Systems in Smart Urban Areas DOI: http://dx.doi.org/10.5772/intechopen.85342

PiðÞ¼ <sup>t</sup> C tð Þ <sup>P</sup><sup>C</sup>

// start hour cycle

Zero and Net Zero Energy

for (int j = 2; j <= nt +1; j ++) { // start iteration cycle

Else

} }

4. Conclusion

82

for (iter = 1; iter <= n; iter ++) {

if (x (ij) == 0) {

// state random generation

// random choice of source i = ran (seed) \* (ng - 1) +1; ij = (i-1) \* (nt +1) + j; // random change of state

// random choice of source i = ran (seed) \* (ng - 1) +1; ij = (i-1) \* (nt +1) + j;

// random set of power

i ∑<sup>i</sup> P<sup>C</sup> i PC <sup>i</sup> <sup>¼</sup> <sup>1</sup> 2 Pmax <sup>i</sup> � <sup>P</sup>min i

During the course of optimisation, we perform random settings of the state and output of a randomly selected RES generator within the microgrid using a random number generator. This defined procedure is done for each hour and each set period as well as for every iteration. This is done using a programme in JAVA source code.

if (ran (seed) <= Ponoff) x (ij) = 1;

if (ran (seed) <= Ponoff) x (ij) = 0;

p (ij) = rand (seed) \* (Pmax (i) - Pmin (i)) + Pmin (i);

nt is the number of hours and ng is the number of sources. p(ij) the output of the ith generator and the jth hour. Further, x(ij) is the state of the ith source during the jth hour. Pmin and Pmax are power values with their limit in the ith generator. Here, nt is the number of hours, ng is the number of available sources, p (ij) is the power of the ith source and the jth hour, and x (ij) is the state of the ith source during the jth

hour. Values Pmin and Pmax have their limit in the ith source. Ponoff is the

parametrizable probability of a change of the source state. Function ran is set up by the random number generation within the interval of (0,1) with even probability distribution. The result of our experiment in terms of source organisation on the selected Thursday as defined by us is presented in Table 5. Worth mentioning is also the fact that the calculation time when using a laptop was 2 minutes and 30 seconds.

The total power load (consumption) of the intelligent urban area "Rohanské

kWh/year = 21 MWh/year. Total power generation from RES microgrid (Table 3) is 7,801,559 kWh/year = 7.8 MWh/year. At present β ¼ 0:6 of the total power consumption of the smart area which is 12,600,454 kWh/year. The installed distributed micro-network of RES will cover the power consumption of the urban smart area with 62% of electricity. The projected planned concept (ideal idea) is to have by 2020 a factor of 0.2, thus existing distribution rates can be optimised. In

nábřeží" (Rohan Island) according to Table 2 is estimated at 21,000,757

(35)

our experiment, 4.2 MWh/year of surplus power would be 85%, which is 3.6 MWh/year. The intelligent urban area would be self-sufficient in terms of electricity consumption and would also generate 3.6 MWh of electricity per year into the 22-kV power grid. The smart area would be energy-efficient in this case, and 85% of the total volume of electricity produced would be commercial. With the transition to smart grids (Figure 2), it is assumed that the intelligent urban development of the Rohan embankment will behave like a power producer and be able to influence the energy market. Similarly to the today's use of automated exchange system to offset exchange rate differences, a decentralised network of autonomous buildings —power stations—can be created on the energy market.

When defining the unit commitment optimisation from RES by working day (Thursday) in 1-hour increments, we have achieved a further saving of approximately 20%.
